a16z Partner Sees AI Consumer Builders Monetizing Luxury Software
a16z
Anish Acharya argues that model labs will not capture all downstream application value as high compute costs drive high consumer willingness to pay for specialized software.
## Background
This video was published by the a16z YouTube channel on August 26, 2026. According to the page description, host Jen Kha and Andreessen Horowitz General Partner Anish Acharya discuss the future of AI, exploring the evolving model landscape, open-source models, consumer AI applications, personal agents, and the changing economics of AI software startups.
## TL;DR
Anish Acharya argues that the AI landscape is shifting from a single-winner dynamic into a multi-winner ecosystem where specialized open-weight and frontier models coexist to serve distinct domain needs. While basic model capabilities function as underlying primitives, the real value creation and high margins are shifting downstream to application-layer products and consumer AI tools. With consumer and enterprise willingness to pay reaching historic highs, founders are leveraging lower marginal compute costs and automated agentic loops to build high-ticket 'luxury software' and highly profitable micro-SaaS businesses.
## Atmosphere & style
The conversation carries a warm, collegial, and highly practical venture capital atmosphere, filled with lighthearted banter, insider offsite references, and shared personal anecdotes. The dynamic between Jen and Anish is conversational yet analytical, balancing high-level macroeconomic analysis with grounded product teardowns. It feels like an authentic peek into a candid partner discussion, combining sharp business optimism with actionable insights for the next generation of software builders.
## Key takeaways
- [01:20](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=80s) OpenAI, Anthropic, and xAI are competing closely at the model layer, indicating that the future will likely feature multiple tier-one winners rather than a winner-take-all monopoly.
- [02:49](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=169s) Fears of an AI financial bubble may be overstated; persistent high demand for GPUs and tight compute supply point to structural scarcity.
- [03:20](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=200s) Enterprise software stocks overreacted to early AI hype because core enterprise software accounts for a minor portion of total IT spend, and replacing critical systems carries unacceptable operational risks.
- [05:06](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=306s) Traditional moats like scale, brand, and network effects remain intact, but integration moats are severely threatened as AI coding agents make data migration easier.
- [07:53](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=473s) Open-weight models fine-tuned with domain-specific RL allow startups to build localized, compounding operational moats without relying solely on single monolithic model providers.
- [10:02](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=602s) Frontier AI labs are vertically integrating downward into inference and compute infrastructure rather than expanding upward into specialized application layers.
- [13:00](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=780s) Application-layer software creates significant value by aggregating multiple distinct AI models and tailoring raw intelligence primitives to vertical domain workflows.
- [17:21](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1041s) High marginal compute costs, a lack of central app-store distribution, and primitive text interfaces have posed major headwinds for consumer AI adoption.
- [18:46](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1126s) AI coding tools are giving rise to a new era of digitally native entrepreneurs building high-margin micro-SaaS businesses generating $100k to $1M in revenue.
- [22:16](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1336s) Personal AI agents generate compounding retention and pricing power as they accumulate continuous long-term user context over time.
- [27:35](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1655s) Consumer AI is entering a renaissance driven by emotional/interpersonal model capabilities, big tech's risk aversion in social domains, and high consumer willingness to pay.
- [29:35](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1775s) Startups are embracing 'luxury software' models, charging premium prices up to hundreds or thousands of dollars per month to deliver step-function value.
- [30:44](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1844s) Today's early-stage founders lean more toward researchers than MBAs, bringing higher technical sophistication but often requiring advice on business mechanics.
- [32:20](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1940s) Large seed rounds are no longer inherently fatal, as capital today directly translates into training models, compute capacity, and product breadth.
- [34:26](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=2066s) Because major social platforms actively block piggyback distribution, AI startups targeting SMEs rely heavily on word-of-mouth growth to capture a booming wave of new business formation.
## Detailed walkthrough
### [00:02](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=2s) Consumer AI Agents and the "Jeans Secret"
Host Jen Kha kicks off the conversation by introducing Andreessen Horowitz (a16z) General Partner Anish Acharya, noting that the two were together earlier in the week at a GP offsite. Jen reveals that Anish has already been putting consumer AI to practical test using GrokBot to handle his shopping. Anish admits with a laugh that he is about to share some "protected IP": his preference for Frame jeans.
Anish details his test drive of GrokBot, highlighting its "resourcefulness" as the defining characteristic of this new wave of consumer tools. Before going to bed, he snapped a picture of his current jeans, set a budget cap of $500, and instructed the bot to buy a similar pair. By the morning, the agent had independently researched options, found an ideal match with the same fit and a different wash, executed the transaction using his credit card, and had the shipping confirmation ready.
```
- Task Given: Find and buy jeans based on a photo of current pair.
- Constraint: Maximum budget of $500.
- Execution: Autonomous research, product selection, credit card transaction, and shipping confirmation overnight.
```
Jen shares her own weekend plans to unleash a personal bot on a two-year backlog of items she promised her husband she would sell. Both agree that consumer adoption hinges on combining these underlying technical capabilities with intuitive product architecture and resourcefulness.
---
[01:20](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=80s) The Multi-Winner Model Race
When Jen poses the core question of which AI company will emerge as the undisputed winner three years down the line, Anish firmly places himself in the "many winners" camp. He points out how rapidly the model ecosystem shifts, noting that xAI transformed from a non-contender into a tier-one model provider alongside OpenAI and Anthropic in just a matter of weeks, shifting the frontier from a two-horse race to a three-horse race.
```
- Landscape Evolution: Transitioned from a two-player dynamic to a three-horse race with xAI's rapid rise.
- OpenAI Momentum: Recent resurgence driven by new model releases, the Codex harness, and the ChatGPT desktop app.
- Anthropic Dynamics: Shifted from apparent absolute market dominance to facing heavy developer competition, alongside plans to go public later in the year.
```
Jen notes that developer sentiment—especially on platforms like X—acts as a leading indicator of underlying market shifts. Developers operate as "fair-weather fans," migrating rapidly toward whichever model provides the highest performance and best token economics at any given moment. With Anthropic facing recent pushback around token usage limits while preparing to go public, and OpenAI regaining strong momentum through specialized tooling, both labs continue to experience massive growth side-by-side rather than competing in a winner-take-all monopoly.
---
[02:49](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=169s) The Macro AI Economy: Infinite Demand vs. Constrained Supply
Setting up the overarching presentation, Anish outlines three core themes for the discussion: macro market dynamics, the application layer as the "productization of the intelligence primitive," and the recent renaissance in consumer AI applications. Addressing broader market sentiment, Anish challenges the prevailing narrative surrounding an AI financial bubble. While acknowledging that bubble risks are heavily debated, he argues that the under-discussed scenario is that the market may actually be "insufficiently optimistic."
```
- Infrastructure Signal: Hourly rental prices for non-cutting-edge GPUs (such as the Nvidia B200) are increasing rather than experiencing normal hardware deflation.
- Market Dynamics: High demand coupled with tight compute supply points to structural scarcity rather than speculative overbuild.
- Enterprise Spend Context: Enterprise software accounts for only 8% to 12% of total IT spend, leaving substantial room for expansion.
```
Referencing the sharp SaaS market drawdown in February—where software stocks dipped 30% to 40% before rebounding by 40%—Anish argues that public markets overreacted. Because enterprise software spend represents a modest 8% to 12% of total enterprise budgets, companies will not suddenly replace established mission-critical systems like payroll or CRM with custom "vibe-coded" internal tools. The cost savings of replacing existing enterprise software are capped, while the operational and compliance risks of errors in core systems remain practically unlimited. Most enterprise tasks demand strict precision that raw base models alone cannot guarantee.
[05:06](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=306s) Traditional Moats, Integration Risks, and Enterprise Architecture
The conversation opens with a discussion on how the market environment for SaaS software is shifting as the initial hype recedes. High levels of stock-based compensation (SBC) previously masked underlying financial realities for many software-as-a-service companies. Now, these companies are forced to either accelerate product innovation or face decline. However, the market outlook is less bleak than initially feared, despite pressing existential questions around competitive moats.
Addressing the widespread belief that AI eliminates moats entirely, the speakers reference the framework codified in Hamilton Helmer's book *Seven Powers*. Most classic moats—such as network effects, scale distribution, and brand power—remain largely unaffected by cheap, abundant intelligence. For instance, coding agents will not diminish Nike's brand, nor do they undermine Instagram, whose primary value resides in its underlying user network rather than the technical complexity of building its app.
* **Exposed Moats:** The "integration moat" is severely threatened. Enterprise software legacy systems like SAP have historically been notoriously complex and risky to migrate to or from, creating locked-in dependencies. Coding agents now make software integration and data migration drastically easier, posing existential risks to system integrators (SIs) and global system integrators (GSIs).
* **Enterprise Job Functions:** Corporate functions divide into *alpha-creating* roles (product, sales, engineering, research) with unbounded upside, and *supporting* roles (legal, HR, finance) with bounded upside.
* **Economic Model Selection:** Enterprise workloads with unbounded upside logically justify paying top-dollar for frontier models (e.g., GPT-5/6, Claude 3.5, or Grok), where even a single IQ point increase drives outsized business value. Conversely, functions with bounded upside—such as closing financial books accurately—benefit from open-weight models fine-tuned with reinforcement learning to optimize cost efficiency.
[07:53](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=473s) Why Open-Weight Models Are Essential for Startups
The discussion shifts to the strategic role of open-source and open-weight models, referencing an insightful post by Decagon founder Jesse Zhang. For many high-growth startups, open-source AI is not merely a cost-saving measure, but the only viable architecture for localizing, fine-tuning, and maintaining complete operational control over their software infrastructure.
AI models increasingly display distinct behavioral archetypes and specialized focus areas that exist in tension with one another. Some models exhibit highly literal, methodical tendencies (e.g., GLM-4 series), executing exact instructions without extrapolation. Other models behave more creatively and presumptuously, functioning like open-ended brainstorming partners. Successful enterprise application architectures deliberately orchestrate multiple specialized model types rather than relying on a single monolithic intelligence.
* **Domain-Specific Fine-Tuning:** By applying reinforcement learning (RL) and domain-specific reasoning traces to open-weight models, companies create a compounding moat tailored to their unique customer dataset.
* **Specialization vs. Generality Trade-off:** Tailoring a model for domain mastery naturally reduces its general-purpose capabilities. For instance, legal AI platforms like Harvey or customer support platforms like Decagon build domain-dominant models that excel at their specific vertical tasks even if they lose performance in abstract logic or mathematics.
[10:02](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=602s) Model Integration vs. Infrastructure Expansion
Reflecting on how rapidly the market has evolved, Anish Acharya highlights the contrast between early expectations and reality. In early 2024, there was widespread panic that frontier AI labs would vertically integrate upward into the application layer. A prime example occurred when Anthropic released its legal plugin for Claude—which was essentially a collection of skill files, or structured long prompts—causing massive market sell-offs in legacy legal platforms like Thomson Reuters. However, instead of moving up into applications, major AI labs have vertically integrated *downward* into inference and compute infrastructure. Because inference workloads are highly homogeneous, building massive scale at the bottom of the value chain makes financial and operational sense, whereas moving into the application layer requires navigating highly fragmented pricing, packaging, and high operational expenditure (OpEx).
Acharya refutes the idea that AI models are becoming simple, interchangeable commodities. Rather than converging into identical commodities, models are developing distinct comparative advantages at the domain level:
* **Knowledge Work Specialization:** Models like OpenAI's GPT series excel at knowledge work, supported by tailored product harnesses (such as the ChatGPT desktop app) designed specifically for spreadsheets, slide decks, and document generation.
* **Software Engineering Specialization:** Platforms like Claude Code focus intensely on software development, providing specialized terminal interfaces, code planning tools, and automated testing features.
* **Diverse "Model Personalities":** Similar to the "Big Five" personality traits in psychology, different tasks require different cognitive traits. Accounting problems demand precision and "neuroticism," whereas creative design challenges benefit from high "openness."
---
[13:00](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=780s) Aggregating Models at the Application Layer
Because individual models possess distinct strengths and non-overlapping training datasets, aggregating multiple models within a single application often yields an outcome far greater than the sum of its parts. Anish uses Expedia as an illustrative metaphor: just as consumers prefer a single platform that aggregates flights from United, Delta, and Southwest rather than visiting each airline's site individually, users benefit from software shells that orchestrate multiple AI models under one roof.
* **Multi-Model Coding Workflows:** Developer tools like Cursor leverage high-end frontier models for complex planning and system design, while routing execution to faster, lower-cost models.
* **Multi-Modal Creative Suites:** Production platforms combine specialized audio models (like ElevenLabs) with image and video generation models (like Black Forest Labs' FLUX) inside a unified creative workspace.
* **Adversarial Research & Decision-Making:** Applications query multiple models simultaneously across distinct datasets to cross-examine responses adversarially, using a secondary model to reconcile discrepancies and reach a consensus.
Because AI research labs are structurally incentivized to promote their proprietary models, independent application layers hold a unique strategic advantage: they can remain model-agnostic and deliver true best-of-breed intelligence to end users.
---
[14:20](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=860s) Intelligence as a Primitive and the Application Layer Value
Raw AI intelligence operates as an infrastructure primitive, directly comparable to cloud compute primitives like AWS. Just as Salesforce built a massive business by transforming AWS cloud computing into specialized CRM software that delivers concrete economic outcomes, successful AI startups transform raw model capabilities into tailored industry solutions.
Raw frontier models lack the deep domain workflows, regulatory compliance frameworks, and specialized user experiences required by enterprise verticals. Specialized applications—such as Harvey in the legal sector—bridge this gap by converting baseline model intelligence into high-value economic outcomes tailored to the specific procurement and operational habits of target industries.
Here is a detailed, faithful walkthrough based on the transcript segment provided.
---
[14:57](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=897s) Enterprise Automation, Business Loops, and AI as Industries
Anish Acharya outlines how different market segments approach AI integration based on their specific ambitions and operational goals. For example, most credit unions do not buy AI to cut their workforce in half; rather, they want to double their headcount while maintaining an economically performant business. The primary opportunity for the application layer is delivering products tailored to these distinct expectations.
Anish highlights a major evolutionary shift in AI usage: transitioning from simple prompt engineering to placing models into continuous execution "loops." While the term "agent" is frequently overused, it fundamentally describes a model running inside a loop equipped with tools, memory, and context. Coding provides the clearest illustration of this dynamic in practice today:
1. A software bug is reported.
2. The agent automatically reproduces the bug.
3. A fix is generated and verified.
4. If it is a low-risk change, the system integrates and ships the fix autonomously (and may notify the customer).
5. If it is high-risk, a human engineer steps in to review it.
Beyond software engineering, these autonomous loops can streamline core business operations like price optimization and procurement. The most ambitious variant is the cross-cutting "business loop"—where a model evaluates macro-level operations and suggests strategic shifts (such as recommending opening a new physical branch in Tijuana).
Referencing a framework from Marc Andreessen, Anish emphasizes that founders and investors must evaluate AI capabilities as broad *industries* rather than narrow *markets*. In the coding ecosystem, diverse products serve different tiers of the stack: tools like Cursor expose the raw underlying mechanics directly to professional developers, whereas platforms like Replit provide accessible abstraction layers for non-technical small business owners.
- **Key Distinction**: Most regional institutions (e.g., credit unions) view AI as an expansion multiplier to scale operations, not a tool for headcount reduction.
- **Agent Definition**: A model wrapped inside an operational loop integrated with external tools and long-term memory.
- **Stack Spectrum**: Developer-centric primitives (e.g., Cursor) vs. non-technical abstraction layers (e.g., Replit).
---
[17:21](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1041s) Navigating Bottlenecks in Consumer AI
Turning to consumer AI, Anish suggests that despite years of premature declarations, the industry may finally be entering "consumer's quarter." However, consumer AI has faced three structural headwinds:
1. **High Marginal Costs**: Historically, consumers resist paying recurring software fees. Unlike traditional software with near-zero marginal distribution costs, high compute usage drives up costs significantly. Anish notes a personal example: an app he built to browse his X (formerly Twitter) timeline cost **$250 to onboard a single new user**. Even with zero customer acquisition costs (CAC), unit economics for free mass-market apps fail under those conditions. The recent rapid cost reductions of open-weight models are beginning to fix this.
2. **Lack of Native Distribution**: AI lacks a centralized, unified marketplace comparable to the mobile App Store. As a result, consumer AI builders must build their distribution channels from scratch alongside their core product, mirroring the early Web 2.0 era.
3. **Interface Limitations**: Current AI interactions remain in a "DOS era" dominated by text prompts and command-line interfaces. Achieving widespread adoption requires developing the intuitive "Windows era" of AI UX through deliberate product and design craft.
---
[18:46](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1126s) The Rise of "Mom-and-Pop SaaS" and Personal Agents
Despite structural challenges, two consumer AI categories are showing strong early traction: consumer-facing coding agents and personal automation agents.
Coding tools are enabling a new archetype: the digitally native entrepreneur. A decade ago, young people aspiring to build internet businesses without coding skills were limited to becoming content creators (such as YouTube creators). Today, non-technical individuals can use AI coding agents to launch niche software products generating **$100,000 to $1,000,000 in annual revenue**. While these independent "mom-and-pop SaaS" businesses may not fit traditional venture capital profiles, they represent a meaningful economic shift.
Concurrently, personal agents are evolving. Reflecting on open-source agent breakthroughs like OpenInterpreter (referred to in passing around early open-agent moments like "OpenClaw"), Anish observes that early iterations functioned like Homebrew Computer Club projects—appealing primarily to developers rather than average consumers. The next phase of consumer AI depends on bridging this gap, making personal agents accessible everyday utilities.
- **Developer vs. Non-Developer Monetization**: Shift from ad-funded creator models to self-built micro-SaaS generating $100k–$1M ARR.
- **Historical Parallel**: Early agent frameworks mirror the 1970s Homebrew Computer Club era—technically impressive, developer-centric, and awaiting consumer packaging.
[19:46](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1186s) Redefining Consumer: When the Plumber Uses GrokBot
The discussion opens with a shift away from the traditional definitions of consumer software, reflecting on how tools are crossing lines between personal utility and business operations. Having built companies during the previous era of consumer apps, the hosts examine how blurry the line between consumer and enterprise has become. A practical example is raised: when a local plumber uses an AI tool like GrokBot or ChatGPT to completely transform and run their business end-to-end, is that considered a consumer application or an enterprise application? Historically, consumer software was largely driven by product-led growth (PLG) tactics or monetized through entertainment.
Anish Acharya provides a pragmatic investment rule for defining consumer software: if you cannot justify acquiring a customer through a traditional direct sales force—which typically requires an Annual Contract Value (ACV) of at least $15,000—you must acquire them through marketing. Under this threshold, even a small business owner like a plumber is classified as a consumer. While acknowledging that productivity tools are expanding into everyday life, Anish notes that entertainment will remain a massive driver of consumer AI. Many consumers ultimately prefer spending time rather than just saving it. He points to AI-native entertainment platforms like Character.AI, as well as the surging popularity of short-form generative dramas originating in Asia, as key indicators of where consumer engagement is heading.
- **Customer Acquisition Rule:** If an ACV is below $15,000, acquiring customers through direct sales isn't economically viable, making marketing-driven acquisition necessary—classifying small business owners like plumbers as consumers.
- **Entertainment vs. Productivity:** Consumer applications lean heavily toward entertainment because most everyday users care more about engaging experiences than optimizing productivity.
- **Emerging Content Formats:** Generative and AI-assisted short-form video dramas, widely adopted across Asian markets, represent an emerging wave of consumer AI entertainment.
[22:16](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1336s) Town Demo: Personal Agents & Managing Chaos
The conversation transitions into personal productivity agents, highlighting an application named Town, an investment made by a16z partner Alex Rampell. The hosts encourage founders and consumers alike to use these products firsthand to build intuition around how fast AI tools evolve. Town functions as a personal agent by connecting directly to a user's inbox to handle noise, scrub unwanted subscriptions, and surface critical action items.
The core advantage of tools like Town lies in compounding memory and contextual awareness. Anish draws an analogy between an AI agent and a human workplace: on day one, an AI agent is like a new hire who needs time to get up to speed. By day 30, after absorbing a month of preferences, context, and feedback, it acts like a tenured employee—making accurate assumptions and executing decisions independently. This growing contextual history creates strong user retention and pricing power, as switching away from a fully personalized agent becomes increasingly difficult.
The hosts share personal experiences with inbox overload, noting how high-volume personal email accounts with tens of thousands of unread messages can be safely put on "autopilot." Agents like Town do not just clean the inbox; they proactively optimize usage—sending summaries, highlighting essential communications, and suggesting ways to save credits or eliminate unnecessary recurring costs.
- **Compounding Context:** Agents improve through continuous memory accumulation, transitioning from generic helpers to personalized operators over a 30-day period.
- **Retention & Pricing Power:** The value delivered to users increases over time, driving high product retention and allowing companies to expand their pricing remit.
- **Autopilot Management:** Personal agents take over chaotic administrative loops, handling everything from email prioritization to subscription auditing without requiring manual daily triage.
[23:45](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1425s) The Consumer Quality-of-Life Loop
Building on the concept of automated workflows, the speakers outline a broader framework for how AI agents will shape daily human life. Just as enterprise AI creates autonomous loops for software development and business operations, consumer AI is beginning to execute around fundamental personal life loops: health, personal finance, shopping, family, and friendships.
In these loops, personal circumstances constantly evolve, demanding continuous information gathering, decision-making, and execution. Platforms like OpenAI are investing heavily in personal health and financial intelligence, while numerous startups focus on automated personal shopping. Anish emphasizes that past technology super-cycles show that roughly 80% of the economic surplus created by platform shifts is ultimately passed down to mass-market consumers through dramatic improvements in quality of life.
- **Core Life Loops:** Personal AI agents target four primary consumer domains: health, money, family/friendships, and self-improvement.
- **Value Surplus:** Historically, around 80% of the value generated by major technological platform shifts directly benefits end consumers through enhanced daily efficiency and quality of life.
<ElicitationsGroup message="Explore related topics from this transcript segment:">
<Elicitation label="Analyze AI agent business models and retention" query="Explain how memory compounding in AI agents creates retention and pricing power, using examples from the transcript."/>
<Elicitation label="Compare consumer AI vs. enterprise AI acquisition strategies" query="What are the differences between consumer and enterprise AI acquisition strategies based on the $15k ACV rule discussed in the transcript?"/>
</ElicitationsGroup>
Here is a detailed, faithful walkthrough covering the provided transcript segment.
---
[24:42](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1482s) Personal Agents: One Dominant Platform vs. Specialized Bots
Jen Kha opens the discussion by asking about the architecture of personal assistants, noting that existing products seem to point toward a single assistant managing a person's life context. However, with so many options available, she asks whether the market will consolidate into one dominant platform for personal time and life management, or if it will resemble an operating system where multiple specialized bots talk to each other and configure things on the back end.
Anish Acharya frames his answer around the concept of "comparative vantage point." He explains that the characteristics someone wants from a Certified Financial Planner (CPA/CFA) are entirely different from what they want from a party planner. Because the surface area of human life is so broad, specialized context matters. While there are certainly overlapping pieces of context, he points to **GrokBots** as a great product illustration of how this multi-agent model works in practice: multiple bots are pointed in slightly different directions, all coordinating behind the scenes to deliver a globally optimal outcome for the user.
---
[25:26](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1526s) Value Accretion: App Layer vs. Model Labs
Jen transitions the conversation to the dynamic between model builders and application developers. She asks whether application-layer startups can successfully compete against frontier model labs, and whether the model labs will allow economic value to accrete downstream to applications instead of capturing it all themselves.
Anish argues that critics often underestimate the complexity of product packaging, pricing, and how end customers actually prefer to buy intelligence. The way a teenager wants to consume an intelligence primitive is fundamentally different from how a marketing executive at a credit union wants to buy it. Because end-user needs are so heterogeneous, it makes far less sense for frontier labs to move up into the specialized application layer than it does for them to move down into inference optimization.
Anish contrasts today's landscape with the market in 2023. Back when the industry operated under the assumption of "one model to rule them all," model labs possessed absolute leverage and could have theoretically squeezed app developers by capturing 100% of their gross margin over time. Today, however, because developers have access to multiple competitive options at every point along the Pareto frontier, model labs can no longer dictate terms or absorb all the downstream value.
---
[26:34](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1594s) Investment Criteria and Traction Thresholds
Jen asks Anish about a16z's current investment strategy, specifically inquiring whether they will still invest in companies with zero revenue given how fast teams are making progress today.
Anish clarifies that while they occasionally take a small "call option" on exceptional, experienced founders at a pre-product or "pre-everything" round, the vast majority of their investments are in companies showing clear signs of breakout traction:
* **Live Product Requirement:** Anish notes that it has become essentially disqualifying for a founder not to demo a live, working product during a pitch, regardless of stage, because building software has become so trivial.
* **Evaluation Framework:** Once a startup demonstrates statistically significant product and sales signals, his framework focuses on extrapolating those metrics to assess whether the entry valuation and implicit risks make sense.
* **Product Velocity:** The speed at which teams iterate and deploy updates remains one of the core metrics the firm measures carefully.
---
[27:35](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1655s) The Consumer AI Renaissance and Big Tech Blunders
Jen notes that consumer software has been "unloved" by venture capital for a long time, but asks if we are now witnessing a massive reversion as value shifts heavily toward applications.
Anish wholeheartedly agrees, declaring it a **"renaissance for being a consumer builder."** He highlights several factors driving this shift:
* **Emotional and Interpersonal Intelligence:** For 40 years, technology only boosted human intellect and productivity. For the first time, AI provides a primitive capable of operating in emotional and interpersonal domains—allowing users to have conversations with models like Claude or OpenAI and actually feel emotional connections.
* **Structural Blind Spots of Big Tech:** incumbent tech giants and frontier labs are culturally incapable of launching certain high-engagement consumer products. For example, Google could never launch an AI companion product that disagrees with the user or contains sexual innuendo due to thousands of internal compliance committees designed to prevent risk. This leaves a massive greenfield for unconstrained startups.
* **Shift in Consumer Willingness to Pay:** Consumers are excited to discover and download new software in a way not seen since the launch of the iPhone App Store in 2009. Crucially, while consumers in 2009 balked at paying 99 cents for an app, today's consumer AI users are willing to pay up to $200 per month for software that delivers real value—giving rise to a era of high-margin, "luxury" consumer software.
Jen jokes that the "autists in San Francisco" have been waiting a very long time for this exact moment. She then segues into a audience question from Michelle regarding the underlying unit economics, gross margins, compute access pressures, and capital moats for this new generation of AI application startups.
Here is the detailed, faithful walkthrough of the provided transcript section:
[29:35](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1775s) The New Economics of AI Apps: Margins, Compute & "Luxury Software"
Jen Kha opens the segment by highlighting how crucial compute access and capital have become as moats in the current environment. Given how expensive it is to stay competitive, she asks Anish Acharya how a venture firm like Andreessen Horowitz (a16z) thinks about the economics of underwriting returns in AI application companies today.
Anish responds that the conversation around software margins has become far more nuanced than it used to be. Pointing to a blog post written by their colleague David Ulevitch, Anish explains that in this current cycle, it is actually rational for companies to trade away traditional gross margins in order to deliver a much wider product surface. The silver lining balancing out higher compute costs is an unprecedented consumer and enterprise willingness to pay for high-value AI products.
To help founders capture this value, Anish describes an exercise a16z routinely conducts with them. Historically, $20 per month was considered the pricing ceiling for consumer software. Anish asks founders to imagine: *"What's the $200-a-month SKU of your product? In fact, what's the $2,000-a-month SKU?"* He envisions the emergence of "luxury software"—the software equivalent of a Hermès Birkin bag. Customers are proving eager to pay premium prices if the software delivers step-function value, making high compute costs a worthwhile trade-off despite the current "fog of war" surrounding margins.
Jen lightheartedly teases Anish about his upscale analogy, joking that he is dropping "Birkin bag framed jeans" references despite being a guy she usually sees dressed strictly in classic tech-founder quarter-zip sweaters. Anish laughs off the banter, clarifying that beneath the quarter-zip, he is simply being a good steward of venture capital.
- **David Ulevitch's thesis**: Software gross margins are no longer a static metric; trading margin for broader AI capabilities is often a rational strategic choice.
- **Pricing ceilings shift**: Moving beyond the classic $20/month consumer subscription model to explore $200/month and $2,000/month "luxury software" tiers.
---
[30:44](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1844s) The Evolving Founder Archetype: Less MBAs, More Researchers
Jen pivots the conversation to founder archetypes. She recalls a chat with Anish from roughly five years prior, when software and tech infrastructure had reached a level of maturity that attracted diverse backgrounds—such as Google program managers spinning out to start companies. She asks Anish what types of founders are emerging from the woodwork to build AI applications today: Are they technical researchers, product managers, or something else?
Anish observes a clear shift: today’s founders include **"less MBAs, more researchers."** While today's early-stage founders may exhibit lower business sophistication, their technical sophistication is dramatically higher. Anish notes that business acumen can easily be taught or picked up through advisors, but deep technical capability cannot be bootstrapped overnight.
Because these younger, research-heavy founders do not carry preconceived notions about standard software limitations, they accomplish extraordinary things. Senior founders who struggle in this product cycle often fail because they are too far removed from the underlying technology and remain anchored to historic limits. By contrast, younger technical founders operate on the assumption that anything is technically possible.
- **Ben Horowitz's insight**: Anish references an a16z partner offsite where Ben Horowitz remarked that in prior tech cycles, the biggest startup risk was that "ideas were too big." In the current AI cycle, the single biggest risk is that "ideas are too small."
---
[32:20](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1940s) Why Giving Seed Startups Massive Capital Isn't Fatal Anymore
Jen brings up another major theme from their recent firm offsite: the changing playbook around seed funding sizes. In past tech cycles, conventional wisdom held that handing a seed startup too much money—such as $20 million to $100 million—would ruin the company. Founders would attempt to pursue too many ideas simultaneously without the human talent required to execute them, making capital constraints a helpful forcing function for focus.
Anish confirms that while this was historical venture wisdom, the math has changed. Previously, a startup simply could not hire enough talented engineers to build across $20 million worth of product surface area at once. Today, capital can directly be translated into model training, compute scale, and product trade-offs. A team raising $100 million today can deploy that capital productively to deliver a fundamentally different, higher-leverage value proposition than the exact same team could achieve with only $20 million.
While determining the optimal seed round size involves navigating a nuanced landscape, Anish considers it an "embarrassment of riches." He notes he would much rather deal with capital deployment questions today than face the market dynamics of five years ago, when fintech companies were indirectly subsidizing unviable customer models through poor underwriting without a clear path to profitability.
Jen agrees, citing **Chris Dixon's venture framework**: investors always prefer facing a supply problem rather than a demand problem. In the current AI market, consumer and enterprise demand is abundantly high. While the industry works to fix the supply constraints (compute, models, infrastructure), the overwhelming pull of demand guarantees the supply side will catch up.
- **Historical capital constraint**: Oversized seed rounds used to fracture focus due to team bandwidth limits.
- **Current capital dynamics**: Capital directly powers compute, model fine-tuning, and expanded product surface area.
- **Chris Dixon's rule**: Prioritize markets with surplus demand over those with surplus supply; fixing supply issues is a much better problem to have.
---
[34:05](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=2045s) Go-to-Market Realities for Small and Medium Businesses (SMEs)
Jen introduces a listener question submitted by "Mosfa" regarding how AI adoption varies across different business sectors. The user points out that unlike large enterprises, small and medium-sized businesses (SMEs) face less friction when adopting AI because they require far less organizational change management.
Jen notes that while she agrees with the general premise, SMEs can sometimes present unexpected hurdles because small business owners have ingrained habits that require deliberate effort to rewire. She turns to Anish to ask how he views the evolving go-to-market (GTM) playbook for AI startups selling into the SME market.
[34:26](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=2066s) Modern Go-to-Market for SMEs and the New Business Renaissance
When examining how the go-to-market playbook for targeting small and medium enterprises (SMEs) has evolved in the AI era, existing businesses are still reached through the standard, historical channels. However, marketing modern products presents a distinct challenge because established platforms like Instagram, TikTok, and X have grown hyper-vigilant. Having witnessed how prior companies built massive social networks on top of existing platforms, these tech giants now actively prevent startups from piggybacking on their reach to launch new distribution channels. As a result, founders are forced to rely on the ultimate organic network effect: genuine word of mouth driven by exceptional product quality.
The most exciting segment of the current market lies in the extraordinary surge of new business formation, which is hitting an all-time high outside of the temporary peak observed during COVID-19. This boom is powered by an entirely new demographic of entrepreneurs. Rather than traditional business owners—like a 55-year-old local plumber—these new founders are often 25-year-olds who previously might have become YouTube creators. Armed with AI tools, they are now building bespoke SaaS solutions tailored specifically for their local neighborhoods, cities, or high school networks.
- Existing SMEs are still acquired through conventional legacy channels, but user acquisition strategies must account for platform lock-in.
- Major platforms (Instagram, TikTok, X) actively block startups from exploiting their networks to create new distribution layers.
- "Word of mouth" has re-emerged as the primary, indispensable organic growth mechanism for AI startups.
- New business formation rates are near historic highs, second only to the peak seen during the COVID-19 pandemic.
- The new archetype of SME founders consists of younger builders (e.g., 25-year-olds) building localized software applications instead of pursuing traditional creator pathways.
> [02:49](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=169s) productization of the intelligence primitive — Anish Acharya
> [25:26](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1526s) one model to rule them all — Anish Acharya
> [27:35](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1655s) renaissance for being a consumer builder — Anish Acharya
> [27:35](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1655s) autists in San Francisco — Jen Kha
> [29:35](https://www.youtube.com/watch?v=zEZ0rQ8Ef-Y&t=1775s) What's the $200-a-month SKU of your product? In fact, what's the $2,000-a-month SKU? — Anish Acharya
Andreessen Horowitz General Partner Anish Acharya argues that the AI ecosystem is entering a multi-winner foundation model phase while economic value shifts decisively toward specialized application layers and a consumer renaissance. He contends that persistent compute scarcity, emerging autonomous execution loops, and expanding customer willingness to pay challenge prevailing fears of an AI market bubble or model commoditization.
Macro Compute Scarcity and Multi-Winner Frontier Models (1:20)
Acharya disputes the idea that a single model provider will monopolize the market, noting that xAI quickly expanded the frontier into a three-horse race alongside OpenAI and Anthropic. Developer loyalty remains fluid, serving as a leading indicator as practitioners rapidly switch providers based on performance and token economics. At the macro level, Acharya challenges prevailing bubble narratives, suggesting the market may be insufficiently optimistic about structural demand. He cites rising hourly rental rates for non-cutting-edge GPUs, such as the Nvidia B200, as evidence of tight physical supply rather than speculative overcapacity.
Addressing enterprise risk, Acharya highlights that software comprises only 8% to 12% of total IT budgets. Consequently, companies will not hastily replace mission-critical systems like payroll or CRM with internally generated code. The cost savings from replacing existing software are strictly capped, whereas the operational, compliance, and accuracy risks of enterprise errors remain practically unbounded.
Enterprise Moats and Vertical Model Specialization (5:06)
Applying Hamilton Helmer's Seven Powers framework, Acharya argues that traditional moats like network effects, scale distribution, and brand power remain insulated from generative AI. However, enterprise integration moats face severe erosion: automated coding agents dramatically lower data migration and integration friction, posing existential threats to legacy system integrators. Enterprise workloads split functionally between alpha-creating tasks with unbounded upside (such as engineering and sales) and supporting tasks with bounded upside (such as accounting).
This division dictates model economics: unbounded functions justify high expenditures on frontier models where incremental intelligence yields outsized returns, whereas bounded functions favor open-weight models fine-tuned via reinforcement learning. Contrary to early predictions that frontier labs would vertically swallow application software, major labs have integrated downward into inference infrastructure due to the homogeneity of compute workloads. This leaves fragmented, vertical application layers—such as Harvey in legal or Decagon in support—to capture downstream value by packaging model primitives with deep domain workflows.
Autonomous Agent Loops and Consumer Resurgence (14:57)
Acharya outlines an architectural shift from static prompt engineering to continuous execution loops equipped with memory and external tools. In software development, agents autonomously reproduce, patch, and deploy low-risk bug fixes. In personal productivity, applications like Town and GrokBot handle end-to-end tasks, such as filtering inbox subscriptions or researching, purchasing, and shipping apparel within a preset budget. These agents compound contextual memory over a 30-day period, shifting from generic assistants into personalized operators with strong user retention and pricing power.
While consumer AI historically faced high marginal compute costs—exemplified by an experimental app costing $250 to onboard a single user—declining open-weight inference costs are repairing these unit economics. Consumer adoption is further catalyzed by AI's novel emotional and interpersonal capabilities, alongside compliance bottlenecks that culturally prevent Big Tech incumbents from launching unconstrained consumer companion products.
Shifting Venture Economics and Founder Archetypes (29:35)
The boundary between enterprise and consumer markets is fundamentally shifting. Acharya applies a pragmatic investment threshold: products with an Annual Contract Value below $15,000 cannot sustain direct sales teams and must rely on marketing acquisition, functionally categorizing small business operators like plumbers as consumers. Simultaneously, coding agents have enabled a surge in business formation, allowing non-technical founders to launch targeted 'mom-and-pop SaaS' products generating $100,000 to $1,000,000 in annual revenue.
Venture economics are adapting accordingly. Referencing colleague David Ulevitch, Acharya notes that startups rationally sacrifice traditional software gross margins to support broader model capabilities, offset by unprecedented willingness to pay for $200 to $2,000 monthly 'luxury software' tiers. In parallel, early-stage founder demographics have shifted from MBAs to technical researchers. Seed funding dynamics have also inverted: large seed rounds of $20 million to $100 million are no longer inherently fatal, as surplus capital directly funds compute and model training rather than unmanageable human headcount.
While aggressive consumer willingness to pay and structural demand favor application builders, founders still face acute challenges navigating platform gatekeepers on social networks, steep compute costs, and the lack of a centralized distribution app store.
[0:02] Anish Acharya: resourcefulness
[2:49] Anish Acharya: insufficiently optimistic
[2:49] Anish Acharya: productization of the intelligence primitive
[25:26] Anish Acharya: one model to rule them all
[27:35] Anish Acharya: renaissance for being a consumer builder
[27:35] Jen Kha: autists in San Francisco
[29:35] Anish Acharya: What's the $200-a-month SKU of your product? In fact, what's the $2,000-a-month SKU?
[29:35] Anish Acharya: luxury software
[30:44] Anish Acharya: less MBAs, more researchers.
[30:44] Ben Horowitz: ideas were too big
[30:44] Ben Horowitz: ideas are too small
[32:20] Anish Acharya: embarrassment of riches
Original transcript
[0:02] to help me break down all things uh
[0:02] to help me break down all things uh around this incredible abundance. I'm
[0:04] around this incredible abundance. I'm
[0:04] around this incredible abundance. I'm gonna bring up Anisha Charia.
[0:07] gonna bring up Anisha Charia.
[0:07] gonna bring up Anisha Charia. >> Hi.
[0:08] >> Hi.
[0:08] >> Hi. >> Awesome. Awesome. Awesome. Hey, Anish.
[0:10] >> Awesome. Awesome. Awesome. Hey, Anish.
[0:10] >> Awesome. Awesome. Awesome. Hey, Anish. Uh Anish and I were at the at GP offsite
[0:13] Uh Anish and I were at the at GP offsite
[0:13] Uh Anish and I were at the at GP offsite uh earlier this week and and he shared
[0:15] uh earlier this week and and he shared
[0:15] uh earlier this week and and he shared with me that he's already running
[0:17] with me that he's already running
[0:17] with me that he's already running Grockbot uh and has purchased a bunch of
[0:19] Grockbot uh and has purchased a bunch of
[0:19] Grockbot uh and has purchased a bunch of jeans for him. So, Anish, do you want to
[0:21] jeans for him. So, Anish, do you want to
[0:21] jeans for him. So, Anish, do you want to do you want to drop what you purchased?
[0:23] do you want to drop what you purchased?
[0:23] do you want to drop what you purchased? >> True story. True story. Yes. So, I'm
[0:25] >> True story. True story. Yes. So, I'm
[0:25] >> True story. True story. Yes. So, I'm going to reveal an important secret um
[0:27] going to reveal an important secret um
[0:27] going to reveal an important secret um protected IP, which is that I mostly
[0:29] protected IP, which is that I mostly
[0:29] protected IP, which is that I mostly wear frame jeans. Frame is a great
[0:31] wear frame jeans. Frame is a great
[0:31] wear frame jeans. Frame is a great brand. Um,
[0:32] brand. Um,
[0:32] brand. Um, >> and Grock Bots is an awesome product.
[0:34] >> and Grock Bots is an awesome product.
[0:34] >> and Grock Bots is an awesome product. Actually, I'd say the kind of defining
[0:35] Actually, I'd say the kind of defining
[0:35] Actually, I'd say the kind of defining characteristic of Grockbots is sort of
[0:37] characteristic of Grockbots is sort of
[0:37] characteristic of Grockbots is sort of resourcefulness. You know, I went to bed
[0:40] resourcefulness. You know, I went to bed
[0:40] resourcefulness. You know, I went to bed a few nights ago and said, "Hey, uh, buy
[0:42] a few nights ago and said, "Hey, uh, buy
[0:42] a few nights ago and said, "Hey, uh, buy me a pair of jeans that are inspired by
[0:43] me a pair of jeans that are inspired by
[0:43] me a pair of jeans that are inspired by these." I took a photo of my current
[0:45] these." I took a photo of my current
[0:45] these." I took a photo of my current jeans. Um, I said, "Don't spend more
[0:47] jeans. Um, I said, "Don't spend more
[0:47] jeans. Um, I said, "Don't spend more than $500 and get it done." I woke up in
[0:49] than $500 and get it done." I woke up in
[0:49] than $500 and get it done." I woke up in the morning and it had researched, found
[0:51] the morning and it had researched, found
[0:51] the morning and it had researched, found a pair, same fit, different wash, used
[0:54] a pair, same fit, different wash, used
[0:54] a pair, same fit, different wash, used my credit card, purchased them and
[0:56] my credit card, purchased them and
[0:56] my credit card, purchased them and they're on the way. So, I think that is
[0:57] they're on the way. So, I think that is
[0:57] they're on the way. So, I think that is going to be something that we see more
[0:59] going to be something that we see more
[0:59] going to be something that we see more and more of. We already have the
[1:00] and more of. We already have the
[1:00] and more of. We already have the capabilities and now a lot of the kind
[1:01] capabilities and now a lot of the kind
[1:01] capabilities and now a lot of the kind of unlock will come from resourcefulness
[1:04] of unlock will come from resourcefulness
[1:04] of unlock will come from resourcefulness um and also the kind of product
[1:06] um and also the kind of product
[1:06] um and also the kind of product architecture delivered in a way that
[1:07] architecture delivered in a way that
[1:07] architecture delivered in a way that most consumers um can understand.
[1:10] most consumers um can understand.
[1:10] most consumers um can understand. >> Awesome. Awesome. Awesome. Yeah, I told
[1:12] >> Awesome. Awesome. Awesome. Yeah, I told
[1:12] >> Awesome. Awesome. Awesome. Yeah, I told my team that I'm going to set my bot to
[1:14] my team that I'm going to set my bot to
[1:14] my team that I'm going to set my bot to finally take care of the pile of things
[1:17] finally take care of the pile of things
[1:17] finally take care of the pile of things I've been promising my husband that I'm
[1:18] I've been promising my husband that I'm
[1:18] I've been promising my husband that I'm going to sell for the last two years.
[1:21] going to sell for the last two years.
[1:21] going to sell for the last two years. That is the project for for this
[1:23] That is the project for for this
[1:23] That is the project for for this weekend. Uh so Anish, we asked the
[1:26] weekend. Uh so Anish, we asked the
[1:26] weekend. Uh so Anish, we asked the question earlier, which one of today's
[1:27] question earlier, which one of today's
[1:27] question earlier, which one of today's AI leaders will be the clear winner in
[1:29] AI leaders will be the clear winner in
[1:29] AI leaders will be the clear winner in three years from now? What's what's your
[1:30] three years from now? What's what's your
[1:30] three years from now? What's what's your take?
[1:31] take?
[1:31] take? >> I'm a many winners guy and I see I'm in
[1:33] >> I'm a many winners guy and I see I'm in
[1:33] >> I'm a many winners guy and I see I'm in a you I'm in good company with many of
[1:34] a you I'm in good company with many of
[1:34] a you I'm in good company with many of you. I mean, if you look at what's
[1:36] you. I mean, if you look at what's
[1:36] you. I mean, if you look at what's happened in the last two weeks, um, you
[1:38] happened in the last two weeks, um, you
[1:38] happened in the last two weeks, um, you know, XAI went from not even being a
[1:41] know, XAI went from not even being a
[1:41] know, XAI went from not even being a real contender on the model side to
[1:42] real contender on the model side to
[1:42] real contender on the model side to being, you know, one of three. So, we
[1:45] being, you know, one of three. So, we
[1:45] being, you know, one of three. So, we extraordinarily went from a two- horse
[1:46] extraordinarily went from a two- horse
[1:46] extraordinarily went from a two- horse race to a three-h horsese race. And, you
[1:48] race to a three-h horsese race. And, you
[1:48] race to a three-h horsese race. And, you know, even more broadly over the course
[1:50] know, even more broadly over the course
[1:50] know, even more broadly over the course of the year, we went from anthropic
[1:52] of the year, we went from anthropic
[1:52] of the year, we went from anthropic feeling like they were so dominant, they
[1:53] feeling like they were so dominant, they
[1:53] feeling like they were so dominant, they could do no wrong to OpenAI who's just
[1:56] could do no wrong to OpenAI who's just
[1:56] could do no wrong to OpenAI who's just had an excellent three months. You know,
[1:57] had an excellent three months. You know,
[1:57] had an excellent three months. You know, the new models are exceptional. The new
[2:00] the new models are exceptional. The new
[2:00] the new models are exceptional. The new codeex harness and chatgpt desktop app
[2:02] codeex harness and chatgpt desktop app
[2:02] codeex harness and chatgpt desktop app is very well done and we're seeing the
[2:04] is very well done and we're seeing the
[2:04] is very well done and we're seeing the sort of specialization in different
[2:06] sort of specialization in different
[2:06] sort of specialization in different directions of these labs. They're both
[2:07] directions of these labs. They're both
[2:07] directions of these labs. They're both growing like crazy, you know, despite
[2:09] growing like crazy, you know, despite
[2:09] growing like crazy, you know, despite each other's continued successes. Um,
[2:11] each other's continued successes. Um,
[2:11] each other's continued successes. Um, XAI and as openweight does well. Um, so
[2:14] XAI and as openweight does well. Um, so
[2:14] XAI and as openweight does well. Um, so I'm definitely in the many winners camp.
[2:16] I'm definitely in the many winners camp.
[2:16] I'm definitely in the many winners camp. >> Yeah, it's interesting to see the
[2:17] >> Yeah, it's interesting to see the
[2:17] >> Yeah, it's interesting to see the sentiment also on X, which is not always
[2:19] sentiment also on X, which is not always
[2:19] sentiment also on X, which is not always a perfect, you know, uh, weather vein
[2:21] a perfect, you know, uh, weather vein
[2:21] a perfect, you know, uh, weather vein for the future, but but oftentimes a
[2:23] for the future, but but oftentimes a
[2:23] for the future, but but oftentimes a early indicator of at least where
[2:25] early indicator of at least where
[2:25] early indicator of at least where developer sentiment is. And there's been
[2:26] developer sentiment is. And there's been
[2:26] developer sentiment is. And there's been a lot of push back from Claude it seems
[2:28] a lot of push back from Claude it seems
[2:28] a lot of push back from Claude it seems like recently on people uh in terms of
[2:31] like recently on people uh in terms of
[2:31] like recently on people uh in terms of token usage etc. And so you know
[2:33] token usage etc. And so you know
[2:34] token usage etc. And so you know developers tend to be fair weather fans
[2:35] developers tend to be fair weather fans
[2:35] developers tend to be fair weather fans on these things they will go where the
[2:37] on these things they will go where the
[2:37] on these things they will go where the latest and greatest and very best model
[2:39] latest and greatest and very best model
[2:39] latest and greatest and very best model isn't and particularly the last six to
[2:41] isn't and particularly the last six to
[2:41] isn't and particularly the last six to eight weeks I think we're going to see
[2:42] eight weeks I think we're going to see
[2:42] eight weeks I think we're going to see some very interesting um traction in
[2:45] some very interesting um traction in
[2:45] some very interesting um traction in terms of the flow of activity but
[2:46] terms of the flow of activity but
[2:46] terms of the flow of activity but obviously anthropics going public uh
[2:48] obviously anthropics going public uh
[2:48] obviously anthropics going public uh later this year um and also you know
[2:51] later this year um and also you know
[2:51] later this year um and also you know there's a lot of lot of keen interest on
[2:52] there's a lot of lot of keen interest on
[2:52] there's a lot of lot of keen interest on this. So with that uh that actually
[2:54] this. So with that uh that actually
[2:54] this. So with that uh that actually brings us straight into the topic of
[2:56] brings us straight into the topic of
[2:56] brings us straight into the topic of discussion today. So where where and
[2:58] discussion today. So where where and
[2:58] discussion today. So where where and what is next in the next frontier of
[3:00] what is next in the next frontier of
[3:00] what is next in the next frontier of intelligence.
[3:01] intelligence.
[3:02] intelligence. >> Amazing. Thank you Jen. So let me tee
[3:03] >> Amazing. Thank you Jen. So let me tee
[3:03] >> Amazing. Thank you Jen. So let me tee this up for everybody. Um and please hop
[3:05] this up for everybody. Um and please hop
[3:05] this up for everybody. Um and please hop in if you've got questions. So let's
[3:07] in if you've got questions. So let's
[3:07] in if you've got questions. So let's first cover the kind of macro and what's
[3:08] first cover the kind of macro and what's
[3:08] first cover the kind of macro and what's happening um at a market level. Then
[3:10] happening um at a market level. Then
[3:10] happening um at a market level. Then we're going to hop into the application
[3:12] we're going to hop into the application
[3:12] we're going to hop into the application layer broadly and sort of talk through
[3:14] layer broadly and sort of talk through
[3:14] layer broadly and sort of talk through why applications are the productization
[3:16] why applications are the productization
[3:16] why applications are the productization of the intelligence primitive. And then
[3:18] of the intelligence primitive. And then
[3:18] of the intelligence primitive. And then finally let's talk about consumer. you
[3:19] finally let's talk about consumer. you
[3:19] finally let's talk about consumer. you know, with the launch of Grockbots and a
[3:20] know, with the launch of Grockbots and a
[3:20] know, with the launch of Grockbots and a few other products, it's actually been a
[3:22] few other products, it's actually been a
[3:22] few other products, it's actually been a very fun um couple of weeks in consumer.
[3:25] very fun um couple of weeks in consumer.
[3:25] very fun um couple of weeks in consumer. Okay. Uh hopefully our our dear friend
[3:27] Okay. Uh hopefully our our dear friend
[3:27] Okay. Uh hopefully our our dear friend Leopold doesn't mind me poking a little
[3:29] Leopold doesn't mind me poking a little
[3:29] Leopold doesn't mind me poking a little fun at him here with situational
[3:30] fun at him here with situational
[3:30] fun at him here with situational awareness, please. Next. Um all right.
[3:34] awareness, please. Next. Um all right.
[3:34] awareness, please. Next. Um all right. Look, I think that the kind of case for
[3:36] Look, I think that the kind of case for
[3:36] Look, I think that the kind of case for this being a bubble um is over sort of
[3:38] this being a bubble um is over sort of
[3:38] this being a bubble um is over sort of discussed or at least fully discussed. I
[3:40] discussed or at least fully discussed. I
[3:40] discussed or at least fully discussed. I think actually the out of distribution
[3:42] think actually the out of distribution
[3:42] think actually the out of distribution topic that's less discussed is what if
[3:44] topic that's less discussed is what if
[3:44] topic that's less discussed is what if we're insufficiently optimistic? And if
[3:46] we're insufficiently optimistic? And if
[3:46] we're insufficiently optimistic? And if you look at some of the underlying
[3:48] you look at some of the underlying
[3:48] you look at some of the underlying indicators, what they point to is
[3:50] indicators, what they point to is
[3:50] indicators, what they point to is essentially infinite demand and highly
[3:52] essentially infinite demand and highly
[3:52] essentially infinite demand and highly constrained supply. You know, things
[3:53] constrained supply. You know, things
[3:54] constrained supply. You know, things like B200, which is a non sort of
[3:56] like B200, which is a non sort of
[3:56] like B200, which is a non sort of cutting edge GPU prices going up on a
[3:59] cutting edge GPU prices going up on a
[3:59] cutting edge GPU prices going up on a per hour basis. That is very uh strange.
[4:02] per hour basis. That is very uh strange.
[4:02] per hour basis. That is very uh strange. Normally, we see these things be highly
[4:04] Normally, we see these things be highly
[4:04] Normally, we see these things be highly deflationary and it sort of points to
[4:06] deflationary and it sort of points to
[4:06] deflationary and it sort of points to very constricted supply and essentially
[4:08] very constricted supply and essentially
[4:08] very constricted supply and essentially infinite demand. So we're thinking and
[4:11] infinite demand. So we're thinking and
[4:11] infinite demand. So we're thinking and talking a lot about what's the kind of
[4:12] talking a lot about what's the kind of
[4:12] talking a lot about what's the kind of informed case for optimism here given
[4:14] informed case for optimism here given
[4:14] informed case for optimism here given some of these second order indicators.
[4:16] some of these second order indicators.
[4:16] some of these second order indicators. The SAS bubble was a very or the SAS
[4:18] The SAS bubble was a very or the SAS
[4:18] The SAS bubble was a very or the SAS sort of uh you know whipssaw was an an
[4:21] sort of uh you know whipssaw was an an
[4:21] sort of uh you know whipssaw was an an interesting peak into market psychology.
[4:24] interesting peak into market psychology.
[4:24] interesting peak into market psychology. You know back in February when we saw
[4:26] You know back in February when we saw
[4:26] You know back in February when we saw this you know 30 to 40% draw down on a
[4:28] this you know 30 to 40% draw down on a
[4:28] this you know 30 to 40% draw down on a bunch of SAS names. We said that the
[4:30] bunch of SAS names. We said that the
[4:30] bunch of SAS names. We said that the market has oversold software. Lo and
[4:32] market has oversold software. Lo and
[4:32] market has oversold software. Lo and behold here we are many of those names
[4:33] behold here we are many of those names
[4:33] behold here we are many of those names are back up 40%. So I'm not quite sure
[4:35] are back up 40%. So I'm not quite sure
[4:35] are back up 40%. So I'm not quite sure what we collectively accomplished, but
[4:37] what we collectively accomplished, but
[4:37] what we collectively accomplished, but I'll tell you what we said then, which
[4:39] I'll tell you what we said then, which
[4:39] I'll tell you what we said then, which is still true today, which is for the
[4:41] is still true today, which is for the
[4:41] is still true today, which is for the enterprise, software spend is 8 to 12%.
[4:44] enterprise, software spend is 8 to 12%.
[4:44] enterprise, software spend is 8 to 12%. It's just not a huge proportion of
[4:45] It's just not a huge proportion of
[4:46] It's just not a huge proportion of spend. So the upside to vibe code your
[4:48] spend. So the upside to vibe code your
[4:48] spend. So the upside to vibe code your own payroll or CRM is not particularly
[4:50] own payroll or CRM is not particularly
[4:50] own payroll or CRM is not particularly high. The downside is essentially
[4:52] high. The downside is essentially
[4:52] high. The downside is essentially unlimited. You know, obviously there's
[4:54] unlimited. You know, obviously there's
[4:54] unlimited. You know, obviously there's all kinds of sort of compliance um
[4:55] all kinds of sort of compliance um
[4:56] all kinds of sort of compliance um implications of not getting things like
[4:57] implications of not getting things like
[4:57] implications of not getting things like payroll right. So most enterprise
[4:59] payroll right. So most enterprise
[4:59] payroll right. So most enterprise software today demands a level of
[5:01] software today demands a level of
[5:01] software today demands a level of precision that just isn't afforded by
[5:04] precision that just isn't afforded by
[5:04] precision that just isn't afforded by coding agents. Um the the one thing that
[5:06] coding agents. Um the the one thing that
[5:06] coding agents. Um the the one thing that has happened though is the sort of tide
[5:08] has happened though is the sort of tide
[5:08] has happened though is the sort of tide has receded. So for a lot of SAS
[5:10] has receded. So for a lot of SAS
[5:10] has receded. So for a lot of SAS companies had a ton of SBC and you know
[5:13] companies had a ton of SBC and you know
[5:13] companies had a ton of SBC and you know things that distorted their economic
[5:15] things that distorted their economic
[5:15] things that distorted their economic performance. I think that's very much um
[5:17] performance. I think that's very much um
[5:17] performance. I think that's very much um visible now and they're going to have to
[5:19] visible now and they're going to have to
[5:19] visible now and they're going to have to sort of accelerate or die. So, so less
[5:21] sort of accelerate or die. So, so less
[5:21] sort of accelerate or die. So, so less bleak for the SAS uh sort of market than
[5:24] bleak for the SAS uh sort of market than
[5:24] bleak for the SAS uh sort of market than perhaps we all collectively thought for
[5:26] perhaps we all collectively thought for
[5:26] perhaps we all collectively thought for a few months there. But still some sort
[5:27] a few months there. But still some sort
[5:27] a few months there. But still some sort of existential questions to address. You
[5:30] of existential questions to address. You
[5:30] of existential questions to address. You know, there's been a huge sort of
[5:31] know, there's been a huge sort of
[5:31] know, there's been a huge sort of discussion of moes. Are there any moes?
[5:33] discussion of moes. Are there any moes?
[5:33] discussion of moes. Are there any moes? There's no more moes. And it's it's very
[5:34] There's no more moes. And it's it's very
[5:34] There's no more moes. And it's it's very funny because if you actually study u
[5:36] funny because if you actually study u
[5:36] funny because if you actually study u moes, which I think are most famously
[5:38] moes, which I think are most famously
[5:38] moes, which I think are most famously codified in the book seven powers that's
[5:40] codified in the book seven powers that's
[5:40] codified in the book seven powers that's one of my favorites. The vast majority
[5:42] one of my favorites. The vast majority
[5:42] one of my favorites. The vast majority of moes actually are not affected by
[5:45] of moes actually are not affected by
[5:45] of moes actually are not affected by abundant lowcost intelligence. You know,
[5:47] abundant lowcost intelligence. You know,
[5:47] abundant lowcost intelligence. You know, when you think about network effects,
[5:49] when you think about network effects,
[5:49] when you think about network effects, scale effects, which shows up in
[5:50] scale effects, which shows up in
[5:50] scale effects, which shows up in distribution, brand effects, which we
[5:52] distribution, brand effects, which we
[5:52] distribution, brand effects, which we tend to discount in Silicon Valley,
[5:54] tend to discount in Silicon Valley,
[5:54] tend to discount in Silicon Valley, these things are as good as they've ever
[5:55] these things are as good as they've ever
[5:55] these things are as good as they've ever been. You know, no amount of coding
[5:57] been. You know, no amount of coding
[5:57] been. You know, no amount of coding agents is going to make Nike not Nike.
[5:59] agents is going to make Nike not Nike.
[5:59] agents is going to make Nike not Nike. And Instagram, um, the power of
[6:01] And Instagram, um, the power of
[6:01] And Instagram, um, the power of Instagram was never the complexity of
[6:02] Instagram was never the complexity of
[6:02] Instagram was never the complexity of building the Instagram app. Of course,
[6:04] building the Instagram app. Of course,
[6:04] building the Instagram app. Of course, it was the kind of network behind it.
[6:06] it was the kind of network behind it.
[6:06] it was the kind of network behind it. So, you actually think the majority of
[6:07] So, you actually think the majority of
[6:07] So, you actually think the majority of modes are as good as they've ever been
[6:09] modes are as good as they've ever been
[6:09] modes are as good as they've ever been and and of course are still crit
[6:10] and and of course are still crit
[6:10] and and of course are still crit critical to building compounding value.
[6:12] critical to building compounding value.
[6:12] critical to building compounding value. There are a couple of modes that are
[6:13] There are a couple of modes that are
[6:13] There are a couple of modes that are exposed. For me, the integration mode is
[6:15] exposed. For me, the integration mode is
[6:15] exposed. For me, the integration mode is the most obvious one. You know, SAP is
[6:17] the most obvious one. You know, SAP is
[6:17] the most obvious one. You know, SAP is is so famously complex to integrate into
[6:20] is so famously complex to integrate into
[6:20] is so famously complex to integrate into and out of that it's a sort of
[6:22] and out of that it's a sort of
[6:22] and out of that it's a sort of existential risk to even migrate from
[6:24] existential risk to even migrate from
[6:24] existential risk to even migrate from one version of SAP to the next. Coding
[6:26] one version of SAP to the next. Coding
[6:26] one version of SAP to the next. Coding agents makes this dramatically better. I
[6:28] agents makes this dramatically better. I
[6:28] agents makes this dramatically better. I think there's a bit of an existential
[6:29] think there's a bit of an existential
[6:29] think there's a bit of an existential question actually for SIS and gsis as to
[6:32] question actually for SIS and gsis as to
[6:32] question actually for SIS and gsis as to what will their value be when they've
[6:34] what will their value be when they've
[6:34] what will their value be when they've historically been this sort of point of
[6:35] historically been this sort of point of
[6:35] historically been this sort of point of integration. So, I do think this moat is
[6:37] integration. So, I do think this moat is
[6:37] integration. So, I do think this moat is a little bit at risk, but for the other
[6:39] a little bit at risk, but for the other
[6:39] a little bit at risk, but for the other traditional modes, they persist and
[6:40] traditional modes, they persist and
[6:40] traditional modes, they persist and they're as important as they've ever
[6:41] they're as important as they've ever
[6:41] they're as important as they've ever been. Yeah, I think this is a really
[6:43] been. Yeah, I think this is a really
[6:43] been. Yeah, I think this is a really important concept. You know, as you
[6:44] important concept. You know, as you
[6:44] important concept. You know, as you start to think about what are the job
[6:46] start to think about what are the job
[6:46] start to think about what are the job functions in the enterprise that are
[6:47] functions in the enterprise that are
[6:48] functions in the enterprise that are alpha creating, it's typically product,
[6:50] alpha creating, it's typically product,
[6:50] alpha creating, it's typically product, sales, engineering, research, and
[6:53] sales, engineering, research, and
[6:53] sales, engineering, research, and conversely, what are the job functions
[6:55] conversely, what are the job functions
[6:55] conversely, what are the job functions in the enterprise that are sort of maybe
[6:57] in the enterprise that are sort of maybe
[6:57] in the enterprise that are sort of maybe administrative is is uh too uh too
[6:59] administrative is is uh too uh too
[6:59] administrative is is uh too uh too bleak, but they are supporting other
[7:01] bleak, but they are supporting other
[7:01] bleak, but they are supporting other functions, legal, um HR, finance, etc.
[7:06] functions, legal, um HR, finance, etc.
[7:06] functions, legal, um HR, finance, etc. We really think that the kind of
[7:07] We really think that the kind of
[7:07] We really think that the kind of rational architecture and the one that
[7:08] rational architecture and the one that
[7:08] rational architecture and the one that is emerging is that for jobs that um
[7:11] is emerging is that for jobs that um
[7:11] is emerging is that for jobs that um have unlimited upside like sales or
[7:13] have unlimited upside like sales or
[7:13] have unlimited upside like sales or product you always want to use frontier
[7:15] product you always want to use frontier
[7:15] product you always want to use frontier tokens. And the reason for that is you
[7:17] tokens. And the reason for that is you
[7:17] tokens. And the reason for that is you just don't know what the value of the
[7:19] just don't know what the value of the
[7:19] just don't know what the value of the new product feature or closing a
[7:20] new product feature or closing a
[7:20] new product feature or closing a customer account is. It's effectively
[7:22] customer account is. It's effectively
[7:22] customer account is. It's effectively unbounded and therefore it's
[7:24] unbounded and therefore it's
[7:24] unbounded and therefore it's economically rational to pay almost any
[7:26] economically rational to pay almost any
[7:26] economically rational to pay almost any price for a model that's even one IQ
[7:28] price for a model that's even one IQ
[7:28] price for a model that's even one IQ point smarter. You know, your Fable 5 or
[7:30] point smarter. You know, your Fable 5 or
[7:30] point smarter. You know, your Fable 5 or your Gro uh or your um GPT56.
[7:34] your Gro uh or your um GPT56.
[7:34] your Gro uh or your um GPT56. Conversely, when you talk about
[7:35] Conversely, when you talk about
[7:35] Conversely, when you talk about something like finance, you know, the
[7:37] something like finance, you know, the
[7:37] something like finance, you know, the best way to close the books is
[7:39] best way to close the books is
[7:39] best way to close the books is accurately. You can't close it, you
[7:40] accurately. You can't close it, you
[7:40] accurately. You can't close it, you know, 10x better than accurately. So, as
[7:43] know, 10x better than accurately. So, as
[7:43] know, 10x better than accurately. So, as a result, you kind of have this bounded
[7:44] a result, you kind of have this bounded
[7:44] a result, you kind of have this bounded upside problem where it makes sense to
[7:46] upside problem where it makes sense to
[7:46] upside problem where it makes sense to use openw weight models with
[7:48] use openw weight models with
[7:48] use openw weight models with reinforcement learning for the kind of
[7:49] reinforcement learning for the kind of
[7:49] reinforcement learning for the kind of paroefficient um cost curve. Maybe
[7:52] paroefficient um cost curve. Maybe
[7:52] paroefficient um cost curve. Maybe before we we go off this one because
[7:53] before we we go off this one because
[7:53] before we we go off this one because this is a big debate and and again when
[7:55] this is a big debate and and again when
[7:55] this is a big debate and and again when Kimmy dropped a few weeks ago there was
[7:56] Kimmy dropped a few weeks ago there was
[7:56] Kimmy dropped a few weeks ago there was a lot of consternation uh about this
[7:59] a lot of consternation uh about this
[7:59] a lot of consternation uh about this this topic just given the relative cost
[8:01] this topic just given the relative cost
[8:01] this topic just given the relative cost which was the focus of of the topic of
[8:03] which was the focus of of the topic of
[8:03] which was the focus of of the topic of discussion. But you know um our our
[8:05] discussion. But you know um our our
[8:05] discussion. But you know um our our founder Jesse Zang from Decagon dropped
[8:07] founder Jesse Zang from Decagon dropped
[8:07] founder Jesse Zang from Decagon dropped this great post around the fact that in
[8:09] this great post around the fact that in
[8:09] this great post around the fact that in some respects and and for a lot of
[8:11] some respects and and for a lot of
[8:11] some respects and and for a lot of companies like Decagadon open source is
[8:14] companies like Decagadon open source is
[8:14] companies like Decagadon open source is actually the only option. It's not just
[8:16] actually the only option. It's not just
[8:16] actually the only option. It's not just cost. It's it's that they can actually
[8:17] cost. It's it's that they can actually
[8:17] cost. It's it's that they can actually localize it, train, fine-tune it. And so
[8:20] localize it, train, fine-tune it. And so
[8:20] localize it, train, fine-tune it. And so maybe unpack a little bit of that
[8:22] maybe unpack a little bit of that
[8:22] maybe unpack a little bit of that configuration. Talk through the the
[8:23] configuration. Talk through the the
[8:23] configuration. Talk through the the nuances there and why folks shouldn't be
[8:25] nuances there and why folks shouldn't be
[8:25] nuances there and why folks shouldn't be concerned even though that is the case
[8:27] concerned even though that is the case
[8:27] concerned even though that is the case for startups that there's a lot in the
[8:30] for startups that there's a lot in the
[8:30] for startups that there's a lot in the way of abundance around this topic.
[8:32] way of abundance around this topic.
[8:32] way of abundance around this topic. >> Yeah, I mean one of the big topics that
[8:33] >> Yeah, I mean one of the big topics that
[8:34] >> Yeah, I mean one of the big topics that we're seeing or one of the big trends is
[8:35] we're seeing or one of the big trends is
[8:35] we're seeing or one of the big trends is that there are just one there are sort
[8:36] that there are just one there are sort
[8:36] that there are just one there are sort of comparative advantages of different
[8:38] of comparative advantages of different
[8:38] of comparative advantages of different models. So and the models often have
[8:41] models. So and the models often have
[8:41] models. So and the models often have sort of areas of focus that are almost
[8:42] sort of areas of focus that are almost
[8:42] sort of areas of focus that are almost at um tension with each other. So you
[8:45] at um tension with each other. So you
[8:45] at um tension with each other. So you see a certain set of models that have a
[8:46] see a certain set of models that have a
[8:46] see a certain set of models that have a high degree of neuroticism. Like there
[8:49] high degree of neuroticism. Like there
[8:49] high degree of neuroticism. Like there sort of autistic models. GLM52 and GLM53
[8:53] sort of autistic models. GLM52 and GLM53
[8:53] sort of autistic models. GLM52 and GLM53 are great examples of this where they're
[8:54] are great examples of this where they're
[8:54] are great examples of this where they're very literal and they'll only do exactly
[8:56] very literal and they'll only do exactly
[8:56] very literal and they'll only do exactly what you told them to do and nothing
[8:57] what you told them to do and nothing
[8:57] what you told them to do and nothing more. Then we're seeing models like a K3
[9:00] more. Then we're seeing models like a K3
[9:00] more. Then we're seeing models like a K3 um that are just much more sort of open
[9:03] um that are just much more sort of open
[9:03] um that are just much more sort of open and they're very presumptuous and
[9:05] and they're very presumptuous and
[9:05] and they're very presumptuous and they're creative and there are roles for
[9:07] they're creative and there are roles for
[9:07] they're creative and there are roles for both types of models in the organization
[9:08] both types of models in the organization
[9:08] both types of models in the organization and and often the sort of shapes of
[9:10] and and often the sort of shapes of
[9:10] and and often the sort of shapes of those minds if you will are at odds with
[9:12] those minds if you will are at odds with
[9:12] those minds if you will are at odds with each other. So that is like one reason
[9:14] each other. So that is like one reason
[9:14] each other. So that is like one reason you actually want to have multiple
[9:15] you actually want to have multiple
[9:15] you actually want to have multiple models. The reinforcement learning is a
[9:17] models. The reinforcement learning is a
[9:17] models. The reinforcement learning is a really important point. Um you know if
[9:19] really important point. Um you know if
[9:19] really important point. Um you know if you actually have a problem that you can
[9:21] you actually have a problem that you can
[9:21] you actually have a problem that you can specialize the model around with your
[9:23] specialize the model around with your
[9:24] specialize the model around with your reasoning traces, you can start to
[9:25] reasoning traces, you can start to
[9:25] reasoning traces, you can start to create this compounding advantage in
[9:27] create this compounding advantage in
[9:27] create this compounding advantage in your domain for your customer base where
[9:29] your domain for your customer base where
[9:29] your domain for your customer base where you're able to kind of shape the
[9:31] you're able to kind of shape the
[9:31] you're able to kind of shape the intelligence to be better than any
[9:33] intelligence to be better than any
[9:33] intelligence to be better than any general intelligence for your problem. I
[9:35] general intelligence for your problem. I
[9:35] general intelligence for your problem. I know Harvey's had some great results
[9:37] know Harvey's had some great results
[9:37] know Harvey's had some great results with this as well. Now the trade-off of
[9:39] with this as well. Now the trade-off of
[9:39] with this as well. Now the trade-off of that kind of reinforcement learning is
[9:40] that kind of reinforcement learning is
[9:40] that kind of reinforcement learning is you lose generality. So if you have the
[9:42] you lose generality. So if you have the
[9:42] you lose generality. So if you have the best sort of model that's fine-tuned for
[9:44] best sort of model that's fine-tuned for
[9:44] best sort of model that's fine-tuned for solving legal problems, it may not be
[9:46] solving legal problems, it may not be
[9:46] solving legal problems, it may not be great at solving sort of theoretical
[9:48] great at solving sort of theoretical
[9:48] great at solving sort of theoretical math problems and that's okay for
[9:50] math problems and that's okay for
[9:50] math problems and that's okay for Harvey's uses or in the case of decagon
[9:53] Harvey's uses or in the case of decagon
[9:53] Harvey's uses or in the case of decagon customer support. So this sort of openw
[9:55] customer support. So this sort of openw
[9:55] customer support. So this sort of openw weight specialization property is
[9:57] weight specialization property is
[9:57] weight specialization property is something that's very unique and one of
[9:58] something that's very unique and one of
[9:58] something that's very unique and one of the reasons our startups are selecting
[9:59] the reasons our startups are selecting
[10:00] the reasons our startups are selecting them. This is also a big topic. We've
[10:02] them. This is also a big topic. We've
[10:02] them. This is also a big topic. We've learned so much since January. We should
[10:04] learned so much since January. We should
[10:04] learned so much since January. We should really do this monthly gener. Yeah. I
[10:05] really do this monthly gener. Yeah. I
[10:05] really do this monthly gener. Yeah. I mean honestly we there's just so much
[10:07] mean honestly we there's just so much
[10:07] mean honestly we there's just so much changing. So in January February there
[10:09] changing. So in January February there
[10:09] changing. So in January February there was a lot of discussion and it's it's
[10:11] was a lot of discussion and it's it's
[10:11] was a lot of discussion and it's it's very idiosyncratic and interesting. You
[10:12] very idiosyncratic and interesting. You
[10:12] very idiosyncratic and interesting. You know anthropic quad released what is
[10:15] know anthropic quad released what is
[10:15] know anthropic quad released what is called a legal plugin. You know plugins
[10:17] called a legal plugin. You know plugins
[10:17] called a legal plugin. You know plugins are just collections of skill files. You
[10:19] are just collections of skill files. You
[10:19] are just collections of skill files. You can think of it as a zip of skill files.
[10:21] can think of it as a zip of skill files.
[10:21] can think of it as a zip of skill files. Skill files are just prompts. They're
[10:22] Skill files are just prompts. They're
[10:22] Skill files are just prompts. They're just long prompts. And there was this
[10:24] just long prompts. And there was this
[10:24] just long prompts. And there was this huge panic and all of a sudden Thompson
[10:26] huge panic and all of a sudden Thompson
[10:26] huge panic and all of a sudden Thompson Reuters and a bunch of other sort of uh
[10:28] Reuters and a bunch of other sort of uh
[10:28] Reuters and a bunch of other sort of uh you know big legal names traded down
[10:30] you know big legal names traded down
[10:30] you know big legal names traded down dramatically. But those were really just
[10:32] dramatically. But those were really just
[10:32] dramatically. But those were really just prompts. And there's a lot of discussion
[10:33] prompts. And there's a lot of discussion
[10:33] prompts. And there's a lot of discussion about if labs were going to integrate
[10:36] about if labs were going to integrate
[10:36] about if labs were going to integrate vertically integrate up into the
[10:37] vertically integrate up into the
[10:37] vertically integrate up into the application layer. Instead, we've seen
[10:39] application layer. Instead, we've seen
[10:39] application layer. Instead, we've seen the very opposite, which is yes, they
[10:41] the very opposite, which is yes, they
[10:41] the very opposite, which is yes, they are vertically integrating, but they're
[10:42] are vertically integrating, but they're
[10:42] are vertically integrating, but they're vertically integrating down into
[10:43] vertically integrating down into
[10:43] vertically integrating down into inference and compute. It's actually
[10:45] inference and compute. It's actually
[10:45] inference and compute. It's actually logical now um in hindsight because the
[10:48] logical now um in hindsight because the
[10:48] logical now um in hindsight because the workloads for inference are very
[10:50] workloads for inference are very
[10:50] workloads for inference are very homogeneous. So, you can build enormous
[10:52] homogeneous. So, you can build enormous
[10:52] homogeneous. So, you can build enormous scale in one part of the value chain.
[10:54] scale in one part of the value chain.
[10:54] scale in one part of the value chain. Whereas when you think about the
[10:55] Whereas when you think about the
[10:56] Whereas when you think about the application layer, you know, you've got
[10:57] application layer, you know, you've got
[10:58] application layer, you know, you've got so many idiosyncrasies and unique needs
[11:00] so many idiosyncrasies and unique needs
[11:00] so many idiosyncrasies and unique needs in terms of pricing, packaging, um, sort
[11:03] in terms of pricing, packaging, um, sort
[11:03] in terms of pricing, packaging, um, sort of productization, how the market wants
[11:05] of productization, how the market wants
[11:05] of productization, how the market wants to buy. So, it's actually a much more
[11:07] to buy. So, it's actually a much more
[11:07] to buy. So, it's actually a much more challenging and opex heavy proposition
[11:09] challenging and opex heavy proposition
[11:10] challenging and opex heavy proposition to move into the application layer
[11:11] to move into the application layer
[11:12] to move into the application layer versus moving down into the inference
[11:14] versus moving down into the inference
[11:14] versus moving down into the inference layer. And this is the point I alluded
[11:16] layer. And this is the point I alluded
[11:16] layer. And this is the point I alluded to earlier, which is sort of this
[11:17] to earlier, which is sort of this
[11:17] to earlier, which is sort of this discussion of model commoditization. And
[11:19] discussion of model commoditization. And
[11:19] discussion of model commoditization. And you know if you use the models every day
[11:21] you know if you use the models every day
[11:21] you know if you use the models every day which I do I sort of hold myself to a
[11:22] which I do I sort of hold myself to a
[11:22] which I do I sort of hold myself to a standard of making something either
[11:24] standard of making something either
[11:24] standard of making something either small or big with every model that comes
[11:26] small or big with every model that comes
[11:26] small or big with every model that comes out you you start to appreciate the fact
[11:28] out you you start to appreciate the fact
[11:28] out you you start to appreciate the fact that these things are are not
[11:29] that these things are are not
[11:29] that these things are are not commodities that they have comparative
[11:32] commodities that they have comparative
[11:32] commodities that they have comparative advantage at a domain level. So a great
[11:34] advantage at a domain level. So a great
[11:34] advantage at a domain level. So a great example is open AI with their new um GPT
[11:37] example is open AI with their new um GPT
[11:37] example is open AI with their new um GPT models are just so so good at knowledge
[11:39] models are just so so good at knowledge
[11:39] models are just so so good at knowledge work. The harness is also very well set
[11:42] work. The harness is also very well set
[11:42] work. The harness is also very well set up for knowledge work. You know if
[11:43] up for knowledge work. You know if
[11:43] up for knowledge work. You know if you've used the chat GPT desktop app you
[11:45] you've used the chat GPT desktop app you
[11:45] you've used the chat GPT desktop app you know what I mean. If you haven't please
[11:46] know what I mean. If you haven't please
[11:46] know what I mean. If you haven't please install it. It's very very cool and
[11:48] install it. It's very very cool and
[11:48] install it. It's very very cool and interesting and it's the perfect sort of
[11:50] interesting and it's the perfect sort of
[11:50] interesting and it's the perfect sort of when I say harness I kind of mean kind
[11:52] when I say harness I kind of mean kind
[11:52] when I say harness I kind of mean kind of product container like a browser. Um
[11:54] of product container like a browser. Um
[11:54] of product container like a browser. Um it's the perfect product container to do
[11:57] it's the perfect product container to do
[11:57] it's the perfect product container to do spreadsheets and slide presentations and
[11:59] spreadsheets and slide presentations and
[11:59] spreadsheets and slide presentations and written documents and all of that type
[12:01] written documents and all of that type
[12:01] written documents and all of that type of work. If you look at cloud code which
[12:03] of work. If you look at cloud code which
[12:03] of work. If you look at cloud code which many of you I'm sure have used it's just
[12:05] many of you I'm sure have used it's just
[12:05] many of you I'm sure have used it's just so oriented towards software engineering
[12:07] so oriented towards software engineering
[12:07] so oriented towards software engineering you know it's in a terminal UI.
[12:09] you know it's in a terminal UI.
[12:09] you know it's in a terminal UI. Everything from the small design
[12:10] Everything from the small design
[12:10] Everything from the small design decisions to the areas in which it
[12:12] decisions to the areas in which it
[12:12] decisions to the areas in which it specializes like code planning and code
[12:15] specializes like code planning and code
[12:15] specializes like code planning and code testing is oriented towards the software
[12:17] testing is oriented towards the software
[12:17] testing is oriented towards the software engineer and there are many trade-offs
[12:19] engineer and there are many trade-offs
[12:19] engineer and there are many trade-offs both products are making for that sort
[12:21] both products are making for that sort
[12:21] both products are making for that sort of respective specialization. So one
[12:23] of respective specialization. So one
[12:23] of respective specialization. So one you've kind of got this domain level
[12:25] you've kind of got this domain level
[12:25] you've kind of got this domain level specialization that's already occurring
[12:27] specialization that's already occurring
[12:27] specialization that's already occurring and then two as I mentioned earlier
[12:29] and then two as I mentioned earlier
[12:29] and then two as I mentioned earlier you've got this sort of I think of it as
[12:30] you've got this sort of I think of it as
[12:30] you've got this sort of I think of it as the big five sort of personality traits
[12:32] the big five sort of personality traits
[12:32] the big five sort of personality traits if if folks have studied that you know
[12:34] if if folks have studied that you know
[12:34] if if folks have studied that you know you can't be both highly open and highly
[12:37] you can't be both highly open and highly
[12:37] you can't be both highly open and highly neurotic. Um, and you know, sometimes
[12:39] neurotic. Um, and you know, sometimes
[12:39] neurotic. Um, and you know, sometimes when you have an intelligence you're
[12:41] when you have an intelligence you're
[12:41] when you have an intelligence you're applying to an accounting problem, you
[12:42] applying to an accounting problem, you
[12:42] applying to an accounting problem, you want neuroticism. When you're applying
[12:44] want neuroticism. When you're applying
[12:44] want neuroticism. When you're applying it to a design problem, you want
[12:45] it to a design problem, you want
[12:45] it to a design problem, you want openness. So you actually have a need
[12:47] openness. So you actually have a need
[12:47] openness. So you actually have a need for both types of minds in the
[12:49] for both types of minds in the
[12:49] for both types of minds in the organization, which is why you would
[12:50] organization, which is why you would
[12:50] organization, which is why you would select something like a GLM53 versus a
[12:53] select something like a GLM53 versus a
[12:53] select something like a GLM53 versus a Kimmy K3. So definitely not commodities
[12:55] Kimmy K3. So definitely not commodities
[12:55] Kimmy K3. So definitely not commodities in our view. This is an important point.
[12:57] in our view. This is an important point.
[12:58] in our view. This is an important point. You know, there are many product
[12:59] You know, there are many product
[12:59] You know, there are many product categories in which model aggregation
[13:01] categories in which model aggregation
[13:01] categories in which model aggregation delivers a greater than sum of parts
[13:03] delivers a greater than sum of parts
[13:03] delivers a greater than sum of parts outcome. And you know, a good metaphor
[13:05] outcome. And you know, a good metaphor
[13:05] outcome. And you know, a good metaphor for this is Expedia. You know, it's so
[13:07] for this is Expedia. You know, it's so
[13:07] for this is Expedia. You know, it's so much more useful to use Expedia than it
[13:08] much more useful to use Expedia than it
[13:08] much more useful to use Expedia than it is to go to United, then to go to Delta,
[13:10] is to go to United, then to go to Delta,
[13:10] is to go to United, then to go to Delta, then to go to Southwest. You just want a
[13:13] then to go to Southwest. You just want a
[13:13] then to go to Southwest. You just want a single place where you can benefit from
[13:14] single place where you can benefit from
[13:14] single place where you can benefit from seeing every airline's inventory.
[13:17] seeing every airline's inventory.
[13:17] seeing every airline's inventory. Similarly, you know, in coding, we're
[13:19] Similarly, you know, in coding, we're
[13:19] Similarly, you know, in coding, we're actually seeing this with cursor a ton
[13:21] actually seeing this with cursor a ton
[13:21] actually seeing this with cursor a ton where you want to do a very frontier
[13:23] where you want to do a very frontier
[13:23] where you want to do a very frontier model for planning, for example, but
[13:25] model for planning, for example, but
[13:25] model for planning, for example, but then you can use a lesser model for
[13:27] then you can use a lesser model for
[13:27] then you can use a lesser model for execution and you really need to have
[13:28] execution and you really need to have
[13:28] execution and you really need to have one product harness or sort of product
[13:30] one product harness or sort of product
[13:30] one product harness or sort of product architecture that lets you use multiple
[13:32] architecture that lets you use multiple
[13:32] architecture that lets you use multiple models. Creative Tools is another great
[13:34] models. Creative Tools is another great
[13:34] models. Creative Tools is another great example where you've got, you know,
[13:36] example where you've got, you know,
[13:36] example where you've got, you know, models that specialize in different
[13:37] models that specialize in different
[13:37] models that specialize in different modalities. So you've got something like
[13:39] modalities. So you've got something like
[13:39] modalities. So you've got something like an 11 Labs which of course is incredible
[13:41] an 11 Labs which of course is incredible
[13:41] an 11 Labs which of course is incredible at voice music as well and then you've
[13:43] at voice music as well and then you've
[13:43] at voice music as well and then you've got something like Black Forest which is
[13:45] got something like Black Forest which is
[13:45] got something like Black Forest which is doing such an excellent job in kind of
[13:47] doing such an excellent job in kind of
[13:47] doing such an excellent job in kind of video and and creative direction and the
[13:49] video and and creative direction and the
[13:49] video and and creative direction and the correct product is to bring all of these
[13:51] correct product is to bring all of these
[13:51] correct product is to bring all of these together into one shell. And then
[13:53] together into one shell. And then
[13:53] together into one shell. And then finally research and decisions. We see
[13:55] finally research and decisions. We see
[13:55] finally research and decisions. We see this all the time where you know the
[13:56] this all the time where you know the
[13:56] this all the time where you know the models are trained with sort of non-over
[13:58] models are trained with sort of non-over
[13:58] models are trained with sort of non-over overlapping data sets often. So you're
[14:00] overlapping data sets often. So you're
[14:00] overlapping data sets often. So you're able to just get more information by
[14:02] able to just get more information by
[14:02] able to just get more information by running the same query through many
[14:04] running the same query through many
[14:04] running the same query through many models adversarially and then having a
[14:06] models adversarially and then having a
[14:06] models adversarially and then having a separate model sort of help you
[14:07] separate model sort of help you
[14:07] separate model sort of help you converge. This is a place where the
[14:09] converge. This is a place where the
[14:09] converge. This is a place where the application layer really shines because
[14:11] application layer really shines because
[14:11] application layer really shines because labs of course are both incentivized and
[14:13] labs of course are both incentivized and
[14:13] labs of course are both incentivized and structurally only able to provide their
[14:15] structurally only able to provide their
[14:15] structurally only able to provide their own in-house models. You as an
[14:16] own in-house models. You as an
[14:16] own in-house models. You as an application sort of aggregator can
[14:18] application sort of aggregator can
[14:18] application sort of aggregator can provide the best of breed. Okay, let's
[14:20] provide the best of breed. Okay, let's
[14:20] provide the best of breed. Okay, let's jump into the apps layer. Now the key
[14:22] jump into the apps layer. Now the key
[14:22] jump into the apps layer. Now the key point about the application layer is
[14:24] point about the application layer is
[14:24] point about the application layer is that you know intelligence is a
[14:26] that you know intelligence is a
[14:26] that you know intelligence is a primitive just like buying cloud is a
[14:28] primitive just like buying cloud is a
[14:28] primitive just like buying cloud is a primitive and what does Salesforce do?
[14:30] primitive and what does Salesforce do?
[14:30] primitive and what does Salesforce do? It sort of takes the you know AWS cloud
[14:33] It sort of takes the you know AWS cloud
[14:33] It sort of takes the you know AWS cloud primitive and turns it into CRM software
[14:36] primitive and turns it into CRM software
[14:36] primitive and turns it into CRM software that delivers an economic outcome for
[14:38] that delivers an economic outcome for
[14:38] that delivers an economic outcome for all of their customer segments. The same
[14:40] all of their customer segments. The same
[14:40] all of their customer segments. The same thing is true of the AI application
[14:42] thing is true of the AI application
[14:42] thing is true of the AI application layer. You know it's great to have the
[14:44] layer. You know it's great to have the
[14:44] layer. You know it's great to have the raw intelligence primitive but you
[14:45] raw intelligence primitive but you
[14:46] raw intelligence primitive but you really need Harvey to turn that into an
[14:47] really need Harvey to turn that into an
[14:47] really need Harvey to turn that into an economic outcome for the legal industry.
[14:50] economic outcome for the legal industry.
[14:50] economic outcome for the legal industry. Similar for somebody like credit unions
[14:52] Similar for somebody like credit unions
[14:52] Similar for somebody like credit unions is a really interesting market segment
[14:53] is a really interesting market segment
[14:54] is a really interesting market segment where they're so idiosyncratic in how
[14:55] where they're so idiosyncratic in how
[14:55] where they're so idiosyncratic in how they want to buy products, how they want
[14:57] they want to buy products, how they want
[14:57] they want to buy products, how they want the product sort of productized and and
[14:59] the product sort of productized and and
[14:59] the product sort of productized and and the shape of the ambition for their
[15:01] the shape of the ambition for their
[15:01] the shape of the ambition for their market. You know, most credit unions
[15:03] market. You know, most credit unions
[15:03] market. You know, most credit unions don't want to uh decrease their
[15:04] don't want to uh decrease their
[15:04] don't want to uh decrease their headcount by half. They want to double
[15:05] headcount by half. They want to double
[15:06] headcount by half. They want to double it, right? And they want to double it
[15:07] it, right? And they want to double it
[15:07] it, right? And they want to double it while having a an economically
[15:09] while having a an economically
[15:10] while having a an economically performant business. So, it's just a
[15:12] performant business. So, it's just a
[15:12] performant business. So, it's just a very um specific way that they see the
[15:14] very um specific way that they see the
[15:14] very um specific way that they see the intelligence primitive playing out in
[15:15] intelligence primitive playing out in
[15:16] intelligence primitive playing out in their market segment. and the
[15:17] their market segment. and the
[15:17] their market segment. and the application layer's opportunity is to be
[15:19] application layer's opportunity is to be
[15:19] application layer's opportunity is to be the one that kind of delivers that. This
[15:21] the one that kind of delivers that. This
[15:21] the one that kind of delivers that. This is a bit of an advanced concept, but I
[15:22] is a bit of an advanced concept, but I
[15:22] is a bit of an advanced concept, but I think an important one. If you look at
[15:24] think an important one. If you look at
[15:24] think an important one. If you look at the kind of way that the evolution of AI
[15:26] the kind of way that the evolution of AI
[15:26] the kind of way that the evolution of AI use um has gone, it's gone from
[15:28] use um has gone, it's gone from
[15:28] use um has gone, it's gone from prompting models to putting models in
[15:30] prompting models to putting models in
[15:30] prompting models to putting models in loops. You know, the the term agent is
[15:31] loops. You know, the the term agent is
[15:32] loops. You know, the the term agent is overused, but agent is just a model in a
[15:34] overused, but agent is just a model in a
[15:34] overused, but agent is just a model in a loop with sort of tools and memory and a
[15:35] loop with sort of tools and memory and a
[15:35] loop with sort of tools and memory and a few other things. A great example of
[15:37] few other things. A great example of
[15:37] few other things. A great example of this is coding. You know, we've all seen
[15:39] this is coding. You know, we've all seen
[15:39] this is coding. You know, we've all seen this from um software companies, which
[15:41] this from um software companies, which
[15:41] this from um software companies, which is a bug gets reported, it gets
[15:43] is a bug gets reported, it gets
[15:43] is a bug gets reported, it gets reproduced, a fix gets generated, it
[15:45] reproduced, a fix gets generated, it
[15:45] reproduced, a fix gets generated, it gets verified. If it's a low-risk fix,
[15:48] gets verified. If it's a low-risk fix,
[15:48] gets verified. If it's a low-risk fix, it gets integrated and shipped and maybe
[15:49] it gets integrated and shipped and maybe
[15:49] it gets integrated and shipped and maybe the customer gets an email saying your
[15:51] the customer gets an email saying your
[15:51] the customer gets an email saying your bug was fixed. If it's a high-risisk
[15:53] bug was fixed. If it's a high-risisk
[15:53] bug was fixed. If it's a high-risisk change, perhaps a human reviews it. But
[15:55] change, perhaps a human reviews it. But
[15:55] change, perhaps a human reviews it. But that way, every bug that actually gets
[15:56] that way, every bug that actually gets
[15:56] that way, every bug that actually gets reported to the enterprise now gets
[15:58] reported to the enterprise now gets
[15:58] reported to the enterprise now gets autonomously fixed through this coding
[16:00] autonomously fixed through this coding
[16:00] autonomously fixed through this coding loop. As you start to take that idea and
[16:02] loop. As you start to take that idea and
[16:02] loop. As you start to take that idea and apply it to other parts of the business,
[16:04] apply it to other parts of the business,
[16:04] apply it to other parts of the business, things like price optimization, things
[16:06] things like price optimization, things
[16:06] things like price optimization, things like procurement, these are very natural
[16:09] like procurement, these are very natural
[16:09] like procurement, these are very natural sort of business loops that occur that
[16:10] sort of business loops that occur that
[16:10] sort of business loops that occur that can be fully automated by these models.
[16:13] can be fully automated by these models.
[16:13] can be fully automated by these models. And then perhaps the most ambitious type
[16:15] And then perhaps the most ambitious type
[16:15] And then perhaps the most ambitious type of loop is the business loop, which is
[16:17] of loop is the business loop, which is
[16:17] of loop is the business loop, which is hey, you make a change that's very
[16:19] hey, you make a change that's very
[16:19] hey, you make a change that's very crosscutting to the business and the
[16:20] crosscutting to the business and the
[16:20] crosscutting to the business and the model comes back and says, hey, I think
[16:22] model comes back and says, hey, I think
[16:22] model comes back and says, hey, I think we need to open a branch in Tijana. Now
[16:24] we need to open a branch in Tijana. Now
[16:24] we need to open a branch in Tijana. Now the model can't do that autonomously but
[16:26] the model can't do that autonomously but
[16:26] the model can't do that autonomously but it can make a change at the sort of
[16:27] it can make a change at the sort of
[16:27] it can make a change at the sort of surface level of the entire business
[16:29] surface level of the entire business
[16:29] surface level of the entire business which is extraordinary. This is how
[16:31] which is extraordinary. This is how
[16:31] which is extraordinary. This is how enterprise automation is going to occur
[16:33] enterprise automation is going to occur
[16:33] enterprise automation is going to occur through AI and I think for me coding has
[16:36] through AI and I think for me coding has
[16:36] through AI and I think for me coding has just been over and over again an
[16:37] just been over and over again an
[16:37] just been over and over again an illustration legal is another great area
[16:39] illustration legal is another great area
[16:39] illustration legal is another great area of industries not markets. This is
[16:41] of industries not markets. This is
[16:41] of industries not markets. This is something that Mark says and he's so
[16:43] something that Mark says and he's so
[16:43] something that Mark says and he's so right which is if you look at
[16:44] right which is if you look at
[16:44] right which is if you look at intelligence as a primitive let's think
[16:46] intelligence as a primitive let's think
[16:46] intelligence as a primitive let's think now about coding intelligence as a
[16:47] now about coding intelligence as a
[16:47] now about coding intelligence as a primitive. All of these products are
[16:49] primitive. All of these products are
[16:49] primitive. All of these products are working in their sort of respective
[16:50] working in their sort of respective
[16:50] working in their sort of respective areas of the stack. You know, Quad Code
[16:52] areas of the stack. You know, Quad Code
[16:52] areas of the stack. You know, Quad Code does such an excellent job of kind of
[16:55] does such an excellent job of kind of
[16:55] does such an excellent job of kind of exposing the raw hardware, so to say, to
[16:57] exposing the raw hardware, so to say, to
[16:57] exposing the raw hardware, so to say, to the developer all the way up to replet,
[16:59] the developer all the way up to replet,
[16:59] the developer all the way up to replet, which is a great abstraction layer for
[17:01] which is a great abstraction layer for
[17:01] which is a great abstraction layer for the average small business owner that's
[17:02] the average small business owner that's
[17:02] the average small business owner that's unfamiliar with code. These are
[17:04] unfamiliar with code. These are
[17:04] unfamiliar with code. These are variations of sort of pricing,
[17:05] variations of sort of pricing,
[17:05] variations of sort of pricing, producting, pack, productization,
[17:07] producting, pack, productization,
[17:07] producting, pack, productization, packaging for the coding primitive and
[17:10] packaging for the coding primitive and
[17:10] packaging for the coding primitive and intelligence and all of them are working
[17:12] intelligence and all of them are working
[17:12] intelligence and all of them are working as a result. So, I think a big mental
[17:14] as a result. So, I think a big mental
[17:14] as a result. So, I think a big mental model shift for us is ensuring that
[17:16] model shift for us is ensuring that
[17:16] model shift for us is ensuring that we're assessing these as industries, not
[17:18] we're assessing these as industries, not
[17:18] we're assessing these as industries, not necessarily simple markets. Okay. and
[17:21] necessarily simple markets. Okay. and
[17:21] necessarily simple markets. Okay. and consumer consumers had a really cool
[17:22] consumer consumers had a really cool
[17:22] consumer consumers had a really cool couple of weeks. You know, we've been
[17:24] couple of weeks. You know, we've been
[17:24] couple of weeks. You know, we've been saying for, you know, for three years
[17:25] saying for, you know, for three years
[17:25] saying for, you know, for three years that this is going to be consumer's
[17:26] that this is going to be consumer's
[17:26] that this is going to be consumer's quarter, but I I think that this might
[17:28] quarter, but I I think that this might
[17:28] quarter, but I I think that this might be consumer's quarter. Let's go into it.
[17:30] be consumer's quarter. Let's go into it.
[17:30] be consumer's quarter. Let's go into it. The things that have actually held back
[17:32] The things that have actually held back
[17:32] The things that have actually held back consumer um so far have been a couple of
[17:34] consumer um so far have been a couple of
[17:34] consumer um so far have been a couple of things. You know, the first is consumers
[17:36] things. You know, the first is consumers
[17:36] things. You know, the first is consumers don't love paying for software. We've
[17:38] don't love paying for software. We've
[17:38] don't love paying for software. We've learned this lesson um over and over
[17:39] learned this lesson um over and over
[17:39] learned this lesson um over and over again. And unfortunately, unlike the
[17:42] again. And unfortunately, unlike the
[17:42] again. And unfortunately, unlike the sort of magic of software in the past,
[17:43] sort of magic of software in the past,
[17:43] sort of magic of software in the past, um AI software has marginal costs of
[17:46] um AI software has marginal costs of
[17:46] um AI software has marginal costs of distribution and engagement. And the
[17:47] distribution and engagement. And the
[17:47] distribution and engagement. And the marginal cost can sometimes be very
[17:49] marginal cost can sometimes be very
[17:49] marginal cost can sometimes be very high. you know, I built a an app I use
[17:51] high. you know, I built a an app I use
[17:51] high. you know, I built a an app I use to help me browse my X timeline and it
[17:53] to help me browse my X timeline and it
[17:53] to help me browse my X timeline and it costs $250 to onboard a new user. So, if
[17:56] costs $250 to onboard a new user. So, if
[17:56] costs $250 to onboard a new user. So, if I'm a startup founder looking at that,
[17:57] I'm a startup founder looking at that,
[17:57] I'm a startup founder looking at that, looking at a kind of $250, even with a
[18:00] looking at a kind of $250, even with a
[18:00] looking at a kind of $250, even with a $0 TC onboarding cost, it's very hard to
[18:02] $0 TC onboarding cost, it's very hard to
[18:02] $0 TC onboarding cost, it's very hard to make a mass market free product work.
[18:04] make a mass market free product work.
[18:04] make a mass market free product work. That is changing now because of openw
[18:06] That is changing now because of openw
[18:06] That is changing now because of openw weightight models, dramatically cheaper
[18:07] weightight models, dramatically cheaper
[18:07] weightight models, dramatically cheaper and more performant. You know, the
[18:09] and more performant. You know, the
[18:09] and more performant. You know, the second is we've never had an AI native
[18:11] second is we've never had an AI native
[18:11] second is we've never had an AI native distribution channel. There's no app
[18:12] distribution channel. There's no app
[18:12] distribution channel. There's no app store for AI. So, this actual product
[18:15] store for AI. So, this actual product
[18:15] store for AI. So, this actual product cycle for consumer looks more like web
[18:17] cycle for consumer looks more like web
[18:17] cycle for consumer looks more like web 2.0 know where you have to kind of build
[18:19] 2.0 know where you have to kind of build
[18:19] 2.0 know where you have to kind of build the channel alongside the product and
[18:21] the channel alongside the product and
[18:21] the channel alongside the product and less like mobile where you actually have
[18:22] less like mobile where you actually have
[18:22] less like mobile where you actually have the central point of distribution for
[18:24] the central point of distribution for
[18:24] the central point of distribution for the entire ecosystem. Then the final
[18:26] the entire ecosystem. Then the final
[18:26] the entire ecosystem. Then the final point I think is an important one. You
[18:28] point I think is an important one. You
[18:28] point I think is an important one. You know command line is we're sort of in
[18:29] know command line is we're sort of in
[18:29] know command line is we're sort of in the the DOS era of AI and for this
[18:32] the the DOS era of AI and for this
[18:32] the the DOS era of AI and for this technology and its capabilities to sort
[18:34] technology and its capabilities to sort
[18:34] technology and its capabilities to sort of fully be embraced by consumers. We're
[18:36] of fully be embraced by consumers. We're
[18:36] of fully be embraced by consumers. We're going to need the windows so to say. So
[18:38] going to need the windows so to say. So
[18:38] going to need the windows so to say. So I think there's just a ton of work to be
[18:39] I think there's just a ton of work to be
[18:39] I think there's just a ton of work to be done around product and design craft to
[18:41] done around product and design craft to
[18:41] done around product and design craft to ensure that consumers know how to
[18:43] ensure that consumers know how to
[18:43] ensure that consumers know how to consume um all this magical new
[18:45] consume um all this magical new
[18:45] consume um all this magical new capabilities.
[18:46] capabilities.
[18:46] capabilities. two things are working. Um, so coding
[18:48] two things are working. Um, so coding
[18:48] two things are working. Um, so coding agents is are extraordinary. I know have
[18:51] agents is are extraordinary. I know have
[18:51] agents is are extraordinary. I know have been discussed. I think it's it's
[18:52] been discussed. I think it's it's
[18:52] been discussed. I think it's it's interesting to think about how they work
[18:53] interesting to think about how they work
[18:53] interesting to think about how they work for consumers. You know, if you think of
[18:55] for consumers. You know, if you think of
[18:55] for consumers. You know, if you think of this concept of the digitally native
[18:57] this concept of the digitally native
[18:57] this concept of the digitally native entrepreneur, if you're not a
[18:59] entrepreneur, if you're not a
[18:59] entrepreneur, if you're not a programmer, the way that's historically
[19:01] programmer, the way that's historically
[19:01] programmer, the way that's historically shown up is you're a YouTube creator.
[19:02] shown up is you're a YouTube creator.
[19:02] shown up is you're a YouTube creator. And there's a whole moral panic that we
[19:04] And there's a whole moral panic that we
[19:04] And there's a whole moral panic that we had, you know, 10 years ago about the
[19:05] had, you know, 10 years ago about the
[19:06] had, you know, 10 years ago about the kids want to be YouTube creators, not
[19:07] kids want to be YouTube creators, not
[19:07] kids want to be YouTube creators, not astronauts. But I would interpret that
[19:09] astronauts. But I would interpret that
[19:09] astronauts. But I would interpret that instead as the kids actually who grew up
[19:11] instead as the kids actually who grew up
[19:11] instead as the kids actually who grew up on the internet want to build businesses
[19:13] on the internet want to build businesses
[19:13] on the internet want to build businesses on the internet and the only way to do
[19:14] on the internet and the only way to do
[19:14] on the internet and the only way to do it again is being a creator. Now with
[19:16] it again is being a creator. Now with
[19:16] it again is being a creator. Now with coding agents you can build a software
[19:18] coding agents you can build a software
[19:18] coding agents you can build a software product that generates $100,000 of
[19:20] product that generates $100,000 of
[19:20] product that generates $100,000 of revenue a year, a million dollars of
[19:21] revenue a year, a million dollars of
[19:21] revenue a year, a million dollars of revenue a year. Now these are not
[19:23] revenue a year. Now these are not
[19:23] revenue a year. Now these are not venturebackable businesses but it's a
[19:25] venturebackable businesses but it's a
[19:25] venturebackable businesses but it's a sort of mom and pop SAS opportunity
[19:27] sort of mom and pop SAS opportunity
[19:27] sort of mom and pop SAS opportunity which is emerging and I think very very
[19:29] which is emerging and I think very very
[19:29] which is emerging and I think very very cool for the country. personal agents.
[19:31] cool for the country. personal agents.
[19:31] cool for the country. personal agents. We had this collective moment of
[19:33] We had this collective moment of
[19:33] We had this collective moment of excitement around openclaw um in January
[19:36] excitement around openclaw um in January
[19:36] excitement around openclaw um in January and it was an extraordinary sort of
[19:37] and it was an extraordinary sort of
[19:38] and it was an extraordinary sort of composition of primitives but it never
[19:40] composition of primitives but it never
[19:40] composition of primitives but it never really crossed over into consumer. You
[19:41] really crossed over into consumer. You
[19:41] really crossed over into consumer. You know it was sort of a developer oriented
[19:43] know it was sort of a developer oriented
[19:43] know it was sort of a developer oriented thing more of the homebrew computing
[19:45] thing more of the homebrew computing
[19:46] thing more of the homebrew computing club kind of energy. Um we're starting
[19:47] club kind of energy. Um we're starting
[19:47] club kind of energy. Um we're starting to see with the emergence of Grockbot
[19:49] to see with the emergence of Grockbot
[19:49] to see with the emergence of Grockbot and chat GBT work personal agents being
[19:51] and chat GBT work personal agents being
[19:52] and chat GBT work personal agents being turned into software that consumers can
[19:54] turned into software that consumers can
[19:54] turned into software that consumers can use. Anisha actually um do you mind just
[19:56] use. Anisha actually um do you mind just
[19:56] use. Anisha actually um do you mind just pausing on this before we go to the town
[19:57] pausing on this before we go to the town
[19:57] pausing on this before we go to the town demo because you know you you were a
[19:59] demo because you know you you were a
[19:59] demo because you know you you were a founder um building in the last era of
[20:02] founder um building in the last era of
[20:02] founder um building in the last era of the the consumer app experience and when
[20:05] the the consumer app experience and when
[20:05] the the consumer app experience and when I even think about I was like gosh how
[20:06] I even think about I was like gosh how
[20:06] I even think about I was like gosh how do you even define consumer today
[20:08] do you even define consumer today
[20:08] do you even define consumer today because you know the plumber that
[20:10] because you know the plumber that
[20:10] because you know the plumber that utilizes now Grockbot to completely turn
[20:13] utilizes now Grockbot to completely turn
[20:13] utilizes now Grockbot to completely turn around their business end to end like is
[20:15] around their business end to end like is
[20:15] around their business end to end like is that consumer or is that enterprise
[20:17] that consumer or is that enterprise
[20:17] that consumer or is that enterprise because like it's very like it's almost
[20:19] because like it's very like it's almost
[20:19] because like it's very like it's almost like a like PLG
[20:22] like a like PLG
[20:22] like a like PLG >> movement but but it's coming from as a
[20:24] >> movement but but it's coming from as a
[20:24] >> movement but but it's coming from as a consumer consumer that then cross over
[20:25] consumer consumer that then cross over
[20:25] consumer consumer that then cross over into enterprise and and particularly
[20:27] into enterprise and and particularly
[20:27] into enterprise and and particularly like the last era of consumer
[20:29] like the last era of consumer
[20:29] like the last era of consumer application is more towards
[20:30] application is more towards
[20:30] application is more towards entertainment as a way to to monetize
[20:32] entertainment as a way to to monetize
[20:32] entertainment as a way to to monetize and so maybe unpack some of that and and
[20:34] and so maybe unpack some of that and and
[20:34] and so maybe unpack some of that and and particularly where you've been spending
[20:35] particularly where you've been spending
[20:35] particularly where you've been spending time as a part of that.
[20:37] time as a part of that.
[20:37] time as a part of that. >> I mean our simple rule is if you cannot
[20:39] >> I mean our simple rule is if you cannot
[20:39] >> I mean our simple rule is if you cannot justify acquiring the customer through
[20:41] justify acquiring the customer through
[20:41] justify acquiring the customer through sales which usually means a 15k ACV you
[20:44] sales which usually means a 15k ACV you
[20:44] sales which usually means a 15k ACV you have to acquire them through marketing
[20:46] have to acquire them through marketing
[20:46] have to acquire them through marketing we think of them as a consumer which is
[20:48] we think of them as a consumer which is
[20:48] we think of them as a consumer which is most small business owners. So I think
[20:50] most small business owners. So I think
[20:50] most small business owners. So I think that the plumber is definitely the
[20:51] that the plumber is definitely the
[20:51] that the plumber is definitely the consumer in our sort of investing mind.
[20:54] consumer in our sort of investing mind.
[20:54] consumer in our sort of investing mind. Entertainment is huge and there's going
[20:55] Entertainment is huge and there's going
[20:55] Entertainment is huge and there's going to be a bunch of AI native entertainment
[20:57] to be a bunch of AI native entertainment
[20:57] to be a bunch of AI native entertainment companies. You know I would argue
[20:58] companies. You know I would argue
[20:58] companies. You know I would argue character was kind of an entertainment
[21:00] character was kind of an entertainment
[21:00] character was kind of an entertainment company. There's been a huge trend
[21:02] company. There's been a huge trend
[21:02] company. There's been a huge trend around short form drama mostly in Asia
[21:04] around short form drama mostly in Asia
[21:04] around short form drama mostly in Asia and that's starting to come over here.
[21:05] and that's starting to come over here.
[21:05] and that's starting to come over here. Many of those are generative or sort of
[21:07] Many of those are generative or sort of
[21:07] Many of those are generative or sort of generative assisted. So I look I think
[21:09] generative assisted. So I look I think
[21:10] generative assisted. So I look I think entertainment is going to be massive.
[21:11] entertainment is going to be massive.
[21:11] entertainment is going to be massive. Most people want to spend time not save
[21:13] Most people want to spend time not save
[21:13] Most people want to spend time not save time and consumer is not that interested
[21:14] time and consumer is not that interested
[21:14] time and consumer is not that interested in productivity. So that's definitely
[21:16] in productivity. So that's definitely
[21:16] in productivity. So that's definitely going to happen. Um and probably worth a
[21:19] going to happen. Um and probably worth a
[21:19] going to happen. Um and probably worth a separate deep dive. Okay. And I think
[21:21] separate deep dive. Okay. And I think
[21:21] separate deep dive. Okay. And I think town for folks who have used it, it's
[21:22] town for folks who have used it, it's
[21:22] town for folks who have used it, it's it's just such a magical experience. And
[21:24] it's just such a magical experience. And
[21:24] it's just such a magical experience. And you know this is like the the number one
[21:26] you know this is like the the number one
[21:26] you know this is like the the number one sort of um piece of advice I give to
[21:28] sort of um piece of advice I give to
[21:28] sort of um piece of advice I give to everybody, friends, family, uh folks in
[21:30] everybody, friends, family, uh folks in
[21:30] everybody, friends, family, uh folks in the industry is like please just use the
[21:32] the industry is like please just use the
[21:32] the industry is like please just use the products because it's so easy to build
[21:34] products because it's so easy to build
[21:34] products because it's so easy to build intuition when you see how they change
[21:36] intuition when you see how they change
[21:36] intuition when you see how they change day-to-day. And town is an investment um
[21:38] day-to-day. And town is an investment um
[21:38] day-to-day. And town is an investment um our partner Alex Rampel made. It's
[21:40] our partner Alex Rampel made. It's
[21:40] our partner Alex Rampel made. It's really extraordinary uh productivity
[21:42] really extraordinary uh productivity
[21:42] really extraordinary uh productivity product and you sort of see how the
[21:44] product and you sort of see how the
[21:44] product and you sort of see how the compounding um improvement of the
[21:46] compounding um improvement of the
[21:46] compounding um improvement of the product through memory advantages it
[21:49] product through memory advantages it
[21:49] product through memory advantages it over time. So the first day you use a
[21:51] over time. So the first day you use a
[21:51] over time. So the first day you use a product it doesn't know you that well.
[21:52] product it doesn't know you that well.
[21:52] product it doesn't know you that well. It's sort of like an employee a new hire
[21:55] It's sort of like an employee a new hire
[21:55] It's sort of like an employee a new hire who's just getting up to speed. By day
[21:57] who's just getting up to speed. By day
[21:57] who's just getting up to speed. By day 30 it's able to make excellent
[21:59] 30 it's able to make excellent
[21:59] 30 it's able to make excellent assumptions on your behalf because it
[22:00] assumptions on your behalf because it
[22:00] assumptions on your behalf because it just has soaked in 30 days of sort of
[22:03] just has soaked in 30 days of sort of
[22:03] just has soaked in 30 days of sort of context, memory and skills. And this is
[22:05] context, memory and skills. And this is
[22:05] context, memory and skills. And this is a pattern that we're seeing more and
[22:06] a pattern that we're seeing more and
[22:06] a pattern that we're seeing more and more. The sort of compounding value
[22:08] more. The sort of compounding value
[22:08] more. The sort of compounding value being delivered to the end customer
[22:10] being delivered to the end customer
[22:10] being delivered to the end customer showing up as retention in the business
[22:12] showing up as retention in the business
[22:12] showing up as retention in the business and sort of showing up as pricing power
[22:13] and sort of showing up as pricing power
[22:13] and sort of showing up as pricing power on a per customer basis.
[22:15] on a per customer basis.
[22:16] on a per customer basis. >> Yeah, this is a great one because uh
[22:17] >> Yeah, this is a great one because uh
[22:17] >> Yeah, this is a great one because uh folks can utilize town for their
[22:19] folks can utilize town for their
[22:19] folks can utilize town for their personal use case. Uh and it's a free,
[22:21] personal use case. Uh and it's a free,
[22:21] personal use case. Uh and it's a free, you know, trial. They give you I think
[22:23] you know, trial. They give you I think
[22:23] you know, trial. They give you I think something like 40 uh credits to start or
[22:26] something like 40 uh credits to start or
[22:26] something like 40 uh credits to start or something around there. Um and so you
[22:29] something around there. Um and so you
[22:29] something around there. Um and so you can kind of see it once you plug into
[22:30] can kind of see it once you plug into
[22:30] can kind of see it once you plug into your personal email how productive it
[22:32] your personal email how productive it
[22:32] your personal email how productive it actually is. Um, on the professional
[22:34] actually is. Um, on the professional
[22:34] actually is. Um, on the professional front, I'm always inbox zero. On the
[22:36] front, I'm always inbox zero. On the
[22:36] front, I'm always inbox zero. On the personal front,
[22:37] personal front,
[22:37] personal front, >> my inbox is like 20,000. Uh, David
[22:40] >> my inbox is like 20,000. Uh, David
[22:40] >> my inbox is like 20,000. Uh, David George is is probably cringing on the
[22:41] George is is probably cringing on the
[22:41] George is is probably cringing on the inside here just because it's
[22:43] inside here just because it's
[22:43] inside here just because it's unacceptable. However, uh, you know,
[22:45] unacceptable. However, uh, you know,
[22:45] unacceptable. However, uh, you know, personal life things are common. So, if
[22:46] personal life things are common. So, if
[22:46] personal life things are common. So, if you email me in my personal, I will
[22:48] you email me in my personal, I will
[22:48] you email me in my personal, I will never respond to you. However, I plug
[22:49] never respond to you. However, I plug
[22:50] never respond to you. However, I plug into it and like I don't even check
[22:51] into it and like I don't even check
[22:51] into it and like I don't even check anymore. If there's something important,
[22:52] anymore. If there's something important,
[22:52] anymore. If there's something important, town will surface it to me. And also, it
[22:54] town will surface it to me. And also, it
[22:54] town will surface it to me. And also, it does all the scrubbing of like
[22:56] does all the scrubbing of like
[22:56] does all the scrubbing of like subscriptions and all the things that it
[22:57] subscriptions and all the things that it
[22:57] subscriptions and all the things that it can optimize and it's starting to now
[22:59] can optimize and it's starting to now
[22:59] can optimize and it's starting to now self-improve upon itself. So like it'll
[23:02] self-improve upon itself. So like it'll
[23:02] self-improve upon itself. So like it'll send you emails where it says like hey
[23:03] send you emails where it says like hey
[23:03] send you emails where it says like hey this routine is costing this much like
[23:05] this routine is costing this much like
[23:05] this routine is costing this much like here's how you could actually save your
[23:07] here's how you could actually save your
[23:07] here's how you could actually save your credit swe. So it's sort of this this
[23:09] credit swe. So it's sort of this this
[23:09] credit swe. So it's sort of this this unlock into what starts on the
[23:11] unlock into what starts on the
[23:11] unlock into what starts on the productivity side and to your point
[23:12] productivity side and to your point
[23:12] productivity side and to your point maybe people won't pay for that
[23:14] maybe people won't pay for that
[23:14] maybe people won't pay for that personally but once it starts to get
[23:16] personally but once it starts to get
[23:16] personally but once it starts to get locked in and then expand in terms of
[23:17] locked in and then expand in terms of
[23:17] locked in and then expand in terms of the remit you're like okay I'll pay the
[23:19] the remit you're like okay I'll pay the
[23:19] the remit you're like okay I'll pay the whatever x bucks you know just because
[23:21] whatever x bucks you know just because
[23:21] whatever x bucks you know just because it helps to manage my life and I can put
[23:23] it helps to manage my life and I can put
[23:23] it helps to manage my life and I can put it on autopilot.
[23:25] it on autopilot.
[23:25] it on autopilot. Yeah, it's such a great point, Shannon.
[23:27] Yeah, it's such a great point, Shannon.
[23:27] Yeah, it's such a great point, Shannon. Like my mental model for this is just an
[23:29] Like my mental model for this is just an
[23:29] Like my mental model for this is just an experienced employee, a tenure employee
[23:31] experienced employee, a tenure employee
[23:31] experienced employee, a tenure employee versus a new hire. You know, the new
[23:32] versus a new hire. You know, the new
[23:32] versus a new hire. You know, the new hireer may be brilliant and may even
[23:34] hireer may be brilliant and may even
[23:34] hireer may be brilliant and may even cost less than tenure employee, but we
[23:36] cost less than tenure employee, but we
[23:36] cost less than tenure employee, but we all know the value of a tenure employee.
[23:37] all know the value of a tenure employee.
[23:37] all know the value of a tenure employee. They're just able to make great
[23:38] They're just able to make great
[23:38] They're just able to make great assumptions on behalf of the
[23:40] assumptions on behalf of the
[23:40] assumptions on behalf of the organization and you. And you know, may
[23:42] organization and you. And you know, may
[23:42] organization and you. And you know, may this is a little philosophical, but I
[23:44] this is a little philosophical, but I
[23:44] this is a little philosophical, but I think this is where it all goes. Just as
[23:45] think this is where it all goes. Just as
[23:45] think this is where it all goes. Just as we talked about kind of coding loops and
[23:47] we talked about kind of coding loops and
[23:47] we talked about kind of coding loops and business loops for the enterprise, we
[23:49] business loops for the enterprise, we
[23:49] business loops for the enterprise, we think there's a set of loops that are
[23:50] think there's a set of loops that are
[23:50] think there's a set of loops that are informally defined that really um sort
[23:53] informally defined that really um sort
[23:53] informally defined that really um sort of lay out a consumer's life. think of
[23:56] of lay out a consumer's life. think of
[23:56] of lay out a consumer's life. think of um family, friendships, money, health.
[23:59] um family, friendships, money, health.
[23:59] um family, friendships, money, health. These are all areas where you have sort
[24:01] These are all areas where you have sort
[24:02] These are all areas where you have sort of changing information, decisions,
[24:04] of changing information, decisions,
[24:04] of changing information, decisions, agency, execution, and then the loop
[24:07] agency, execution, and then the loop
[24:07] agency, execution, and then the loop continues. So, we're starting to see
[24:09] continues. So, we're starting to see
[24:09] continues. So, we're starting to see some of these sort of loops emerge
[24:10] some of these sort of loops emerge
[24:10] some of these sort of loops emerge around self-improvement, kind of health
[24:13] around self-improvement, kind of health
[24:13] around self-improvement, kind of health and finance are the two areas that
[24:14] and finance are the two areas that
[24:14] and finance are the two areas that OpenAI is focused on. We've seen a bunch
[24:16] OpenAI is focused on. We've seen a bunch
[24:16] OpenAI is focused on. We've seen a bunch of startups working on shopping, but we
[24:18] of startups working on shopping, but we
[24:18] of startups working on shopping, but we think that like the kind of way that
[24:19] think that like the kind of way that
[24:19] think that like the kind of way that this ends up playing out is a dramatic
[24:22] this ends up playing out is a dramatic
[24:22] this ends up playing out is a dramatic quality of life improvement for the
[24:23] quality of life improvement for the
[24:23] quality of life improvement for the consumer and um and that really follows
[24:25] consumer and um and that really follows
[24:25] consumer and um and that really follows the shape of past product cycles where
[24:27] the shape of past product cycles where
[24:28] the shape of past product cycles where 80% of the surplus is delivered to the
[24:30] 80% of the surplus is delivered to the
[24:30] 80% of the surplus is delivered to the mass market.
[24:31] mass market.
[24:31] mass market. >> Do you think each that all of these uh
[24:33] >> Do you think each that all of these uh
[24:33] >> Do you think each that all of these uh sorry maybe just going back to the to
[24:35] sorry maybe just going back to the to
[24:35] sorry maybe just going back to the to the last slide there's a question here.
[24:36] the last slide there's a question here.
[24:36] the last slide there's a question here. you know, when you think about these
[24:37] you know, when you think about these
[24:37] you know, when you think about these personal agent examples, whether it be
[24:38] personal agent examples, whether it be
[24:38] personal agent examples, whether it be town or ethos, etc., all point to, you
[24:42] town or ethos, etc., all point to, you
[24:42] town or ethos, etc., all point to, you know, kind of one assistant having
[24:44] know, kind of one assistant having
[24:44] know, kind of one assistant having context, but it seems like there's many
[24:45] context, but it seems like there's many
[24:45] context, but it seems like there's many different options. Do you think it'll
[24:47] different options. Do you think it'll
[24:47] different options. Do you think it'll end up being, you know, sort of one
[24:49] end up being, you know, sort of one
[24:49] end up being, you know, sort of one dominant platform for this personal
[24:51] dominant platform for this personal
[24:51] dominant platform for this personal aspect of your life as as time
[24:53] aspect of your life as as time
[24:53] aspect of your life as as time management? Um, or will it be like an
[24:55] management? Um, or will it be like an
[24:55] management? Um, or will it be like an operating system where you have many
[24:56] operating system where you have many
[24:56] operating system where you have many kind of talking to each other and kind
[24:58] kind of talking to each other and kind
[24:58] kind of talking to each other and kind of configuring on the back end?
[25:00] of configuring on the back end?
[25:00] of configuring on the back end? >> I the comparative vantage point kind of
[25:02] >> I the comparative vantage point kind of
[25:02] >> I the comparative vantage point kind of comes to mind. you know, I think the the
[25:03] comes to mind. you know, I think the the
[25:03] comes to mind. you know, I think the the characteristics you want from your CFA
[25:05] characteristics you want from your CFA
[25:05] characteristics you want from your CFA are different from the one that you want
[25:07] are different from the one that you want
[25:07] are different from the one that you want from your sort of party planner. Um, and
[25:10] from your sort of party planner. Um, and
[25:10] from your sort of party planner. Um, and that just the surface area is so broad
[25:12] that just the surface area is so broad
[25:12] that just the surface area is so broad that I think that yes, there's
[25:13] that I think that yes, there's
[25:13] that I think that yes, there's overlapping bits of context. And I think
[25:14] overlapping bits of context. And I think
[25:14] overlapping bits of context. And I think Grock Bots has done a nice job of kind
[25:16] Grock Bots has done a nice job of kind
[25:16] Grock Bots has done a nice job of kind of illustrating this in product or you
[25:19] of illustrating this in product or you
[25:19] of illustrating this in product or you have many bots that are pointed in
[25:21] have many bots that are pointed in
[25:21] have many bots that are pointed in slightly different directions that all
[25:22] slightly different directions that all
[25:22] slightly different directions that all coordinate to deliver a globally optimum
[25:24] coordinate to deliver a globally optimum
[25:24] coordinate to deliver a globally optimum uh optimal outcome.
[25:25] uh optimal outcome.
[25:26] uh optimal outcome. >> There's a few questions. I'm going to go
[25:27] >> There's a few questions. I'm going to go
[25:27] >> There's a few questions. I'm going to go back to topics you've covered earlier.
[25:28] back to topics you've covered earlier.
[25:28] back to topics you've covered earlier. So if the application layer captures
[25:31] So if the application layer captures
[25:31] So if the application layer captures economic outcomes, how do you think
[25:34] economic outcomes, how do you think
[25:34] economic outcomes, how do you think about the competition from the model
[25:36] about the competition from the model
[25:36] about the competition from the model companies um and what will they allow
[25:39] companies um and what will they allow
[25:39] companies um and what will they allow value accretion to happen downstream and
[25:40] value accretion to happen downstream and
[25:40] value accretion to happen downstream and and you know are companies at the app
[25:43] and you know are companies at the app
[25:43] and you know are companies at the app layer able to compete with the frontier
[25:45] layer able to compete with the frontier
[25:45] layer able to compete with the frontier labs going after that particular market?
[25:48] labs going after that particular market?
[25:48] labs going after that particular market? >> I mean I I think so again I think that
[25:50] >> I mean I I think so again I think that
[25:50] >> I mean I I think so again I think that we're we're underestimating the kind of
[25:51] we're we're underestimating the kind of
[25:51] we're we're underestimating the kind of complexity of product pricing packaging
[25:54] complexity of product pricing packaging
[25:54] complexity of product pricing packaging and how the end customer wants to buy.
[25:56] and how the end customer wants to buy.
[25:56] and how the end customer wants to buy. you know, the way that um you know, a
[25:58] you know, the way that um you know, a
[25:58] you know, the way that um you know, a teenager wants to consume the
[25:59] teenager wants to consume the
[25:59] teenager wants to consume the intelligence primitive is different than
[26:01] intelligence primitive is different than
[26:01] intelligence primitive is different than the way an marketing executive at credit
[26:03] the way an marketing executive at credit
[26:03] the way an marketing executive at credit union wants to actually consume it. Um
[26:05] union wants to actually consume it. Um
[26:05] union wants to actually consume it. Um and it's very heterogeneous. So to me,
[26:07] and it's very heterogeneous. So to me,
[26:07] and it's very heterogeneous. So to me, it just makes less sense for um the labs
[26:10] it just makes less sense for um the labs
[26:10] it just makes less sense for um the labs to move up to the apps layer than to
[26:12] to move up to the apps layer than to
[26:12] to move up to the apps layer than to move down to inference. So, you know,
[26:15] move down to inference. So, you know,
[26:15] move down to inference. So, you know, and that kind of permission point's an
[26:16] and that kind of permission point's an
[26:16] and that kind of permission point's an interesting one. I think if we lived in
[26:18] interesting one. I think if we lived in
[26:18] interesting one. I think if we lived in a world of 2023 when it was one model to
[26:20] a world of 2023 when it was one model to
[26:20] a world of 2023 when it was one model to rule them all, it wouldn't even matter
[26:22] rule them all, it wouldn't even matter
[26:22] rule them all, it wouldn't even matter if you had permission because the labs
[26:23] if you had permission because the labs
[26:23] if you had permission because the labs would just take 100% of your gross
[26:25] would just take 100% of your gross
[26:25] would just take 100% of your gross margin over time. But now because you've
[26:27] margin over time. But now because you've
[26:27] margin over time. But now because you've got many options at all points in the
[26:29] got many options at all points in the
[26:29] got many options at all points in the paro frontier, you know, the labs have a
[26:31] paro frontier, you know, the labs have a
[26:31] paro frontier, you know, the labs have a harder time actually doing things like
[26:33] harder time actually doing things like
[26:33] harder time actually doing things like that.
[26:34] that.
[26:34] that. >> Awesome. There was a question just on
[26:35] >> Awesome. There was a question just on
[26:35] >> Awesome. There was a question just on traction. So, do you fund anything where
[26:37] traction. So, do you fund anything where
[26:37] traction. So, do you fund anything where there's um there there's no revenue at
[26:40] there's um there there's no revenue at
[26:40] there's um there there's no revenue at this point just given how quickly people
[26:42] this point just given how quickly people
[26:42] this point just given how quickly people have been making progress or is it
[26:44] have been making progress or is it
[26:44] have been making progress or is it extremely difficult? um we try not to I
[26:47] extremely difficult? um we try not to I
[26:47] extremely difficult? um we try not to I I certainly have spent less time um on
[26:49] I certainly have spent less time um on
[26:49] I certainly have spent less time um on that strategy. Look, I I think that the
[26:51] that strategy. Look, I I think that the
[26:51] that strategy. Look, I I think that the basket is majority investments that are
[26:53] basket is majority investments that are
[26:53] basket is majority investments that are showing some signs of working. Certainly
[26:55] showing some signs of working. Certainly
[26:55] showing some signs of working. Certainly from a product velocity perspective,
[26:56] from a product velocity perspective,
[26:56] from a product velocity perspective, that used to be something we measure
[26:58] that used to be something we measure
[26:58] that used to be something we measure pretty carefully. Like it's
[26:59] pretty carefully. Like it's
[26:59] pretty carefully. Like it's disqualifying to not be showing a live
[27:01] disqualifying to not be showing a live
[27:01] disqualifying to not be showing a live product in a pitch at any stage these
[27:03] product in a pitch at any stage these
[27:03] product in a pitch at any stage these days because it's so trivial to build
[27:04] days because it's so trivial to build
[27:04] days because it's so trivial to build stuff. So almost everything we we're
[27:07] stuff. So almost everything we we're
[27:07] stuff. So almost everything we we're seeing are showing signs of you know
[27:09] seeing are showing signs of you know
[27:09] seeing are showing signs of you know sort of some sort of breakout. I mean my
[27:11] sort of some sort of breakout. I mean my
[27:11] sort of some sort of breakout. I mean my model is somewhat simplistic where I
[27:13] model is somewhat simplistic where I
[27:13] model is somewhat simplistic where I just sort of look at once you have stats
[27:15] just sort of look at once you have stats
[27:15] just sort of look at once you have stats sig sales and product if we extrapolate
[27:17] sig sales and product if we extrapolate
[27:17] sig sales and product if we extrapolate from there um do we kind of like the
[27:20] from there um do we kind of like the
[27:20] from there um do we kind of like the price that we have to pay to be a part
[27:21] price that we have to pay to be a part
[27:21] price that we have to pay to be a part of it and the risks that we're taking
[27:23] of it and the risks that we're taking
[27:23] of it and the risks that we're taking implicitly and that's I'd say the
[27:25] implicitly and that's I'd say the
[27:25] implicitly and that's I'd say the majority of the work that we do look for
[27:26] majority of the work that we do look for
[27:26] majority of the work that we do look for very talented experienced folks we do
[27:29] very talented experienced folks we do
[27:29] very talented experienced folks we do kind of take a small call option which
[27:31] kind of take a small call option which
[27:31] kind of take a small call option which looks like a pre- everything round um
[27:33] looks like a pre- everything round um
[27:33] looks like a pre- everything round um but that's not the majority of what we
[27:34] but that's not the majority of what we
[27:34] but that's not the majority of what we do.
[27:35] do.
[27:35] do. >> Yeah. Yeah. Well, when you think about
[27:37] >> Yeah. Yeah. Well, when you think about
[27:37] >> Yeah. Yeah. Well, when you think about the the um kind of competitive landscape
[27:40] the the um kind of competitive landscape
[27:40] the the um kind of competitive landscape on this uh consumer has been unloved for
[27:43] on this uh consumer has been unloved for
[27:43] on this uh consumer has been unloved for so long um are you seeing now this
[27:46] so long um are you seeing now this
[27:46] so long um are you seeing now this reversion just given it's clear that
[27:48] reversion just given it's clear that
[27:48] reversion just given it's clear that apps is sort of this next layer of value
[27:49] apps is sort of this next layer of value
[27:50] apps is sort of this next layer of value creation like the model sort of layer
[27:51] creation like the model sort of layer
[27:51] creation like the model sort of layer has been somewhat set and I say that
[27:53] has been somewhat set and I say that
[27:53] has been somewhat set and I say that with a huge aster because there might be
[27:55] with a huge aster because there might be
[27:55] with a huge aster because there might be new algorithmic breakthroughs you know
[27:57] new algorithmic breakthroughs you know
[27:57] new algorithmic breakthroughs you know kind of kind of folks coming out from
[27:58] kind of kind of folks coming out from
[27:58] kind of kind of folks coming out from left field as as we have in the
[27:59] left field as as we have in the
[27:59] left field as as we have in the portfolio as well um but do you feel
[28:01] portfolio as well um but do you feel
[28:01] portfolio as well um but do you feel like the shift from the competitive
[28:03] like the shift from the competitive
[28:03] like the shift from the competitive dynamic shifting more towards
[28:04] dynamic shifting more towards
[28:04] dynamic shifting more towards application
[28:05] application
[28:05] application >> 100% I It's sort of a renaissance for
[28:07] >> 100% I It's sort of a renaissance for
[28:07] >> 100% I It's sort of a renaissance for being a consumer builder because you've
[28:09] being a consumer builder because you've
[28:09] being a consumer builder because you've got this extraordinary primitive that
[28:11] got this extraordinary primitive that
[28:11] got this extraordinary primitive that you can work with. By the way, we now
[28:13] you can work with. By the way, we now
[28:13] you can work with. By the way, we now have a primitive that can kind of
[28:15] have a primitive that can kind of
[28:15] have a primitive that can kind of operate in the, you know, emotional
[28:17] operate in the, you know, emotional
[28:17] operate in the, you know, emotional interpersonal domain. You know, you can
[28:19] interpersonal domain. You know, you can
[28:19] interpersonal domain. You know, you can like have a conversation with claude or
[28:21] like have a conversation with claude or
[28:21] like have a conversation with claude or openai or K3 and and feel feelings. And
[28:23] openai or K3 and and feel feelings. And
[28:23] openai or K3 and and feel feelings. And we've had 40 years of technology that
[28:25] we've had 40 years of technology that
[28:25] we've had 40 years of technology that really boosted our intellect and
[28:26] really boosted our intellect and
[28:26] really boosted our intellect and productivity, but nothing that kind of
[28:28] productivity, but nothing that kind of
[28:28] productivity, but nothing that kind of spoke to our humanity. So, it's a whole
[28:30] spoke to our humanity. So, it's a whole
[28:30] spoke to our humanity. So, it's a whole different technology surface. It's very
[28:32] different technology surface. It's very
[28:32] different technology surface. It's very wide. I think there are a set of
[28:34] wide. I think there are a set of
[28:34] wide. I think there are a set of products that labs are just culturally
[28:35] products that labs are just culturally
[28:36] products that labs are just culturally not set up and big tech not set up to go
[28:38] not set up and big tech not set up to go
[28:38] not set up and big tech not set up to go after. You think about launching, you
[28:39] after. You think about launching, you
[28:40] after. You think about launching, you know, a companion product at Google that
[28:42] know, a companion product at Google that
[28:42] know, a companion product at Google that may disagree with you that may um have
[28:44] may disagree with you that may um have
[28:44] may disagree with you that may um have sexual innuendo in it. Like these are
[28:46] sexual innuendo in it. Like these are
[28:46] sexual innuendo in it. Like these are things that there's a thousand
[28:47] things that there's a thousand
[28:47] things that there's a thousand committees at Google um are designed to
[28:50] committees at Google um are designed to
[28:50] committees at Google um are designed to prevent. So startups have areas where
[28:52] prevent. So startups have areas where
[28:52] prevent. So startups have areas where they're kind of uniquely capable and
[28:53] they're kind of uniquely capable and
[28:53] they're kind of uniquely capable and then look finally the consumer sort of
[28:55] then look finally the consumer sort of
[28:55] then look finally the consumer sort of excited to download new software,
[28:57] excited to download new software,
[28:57] excited to download new software, excited to pay for it. It's like
[28:58] excited to pay for it. It's like
[28:58] excited to pay for it. It's like Christmas 2009 with the iPhone. People
[29:00] Christmas 2009 with the iPhone. People
[29:00] Christmas 2009 with the iPhone. People want to try try new apps, but unlike the
[29:02] want to try try new apps, but unlike the
[29:02] want to try try new apps, but unlike the 99 cents days, they're willing to pay
[29:04] 99 cents days, they're willing to pay
[29:04] 99 cents days, they're willing to pay 200 a month. So, it's sort of a
[29:06] 200 a month. So, it's sort of a
[29:06] 200 a month. So, it's sort of a renaissance for consumer builders and
[29:07] renaissance for consumer builders and
[29:07] renaissance for consumer builders and and yeah, I think that things have
[29:09] and yeah, I think that things have
[29:09] and yeah, I think that things have changed.
[29:09] changed.
[29:09] changed. >> Trying to come up with a joke that the
[29:11] >> Trying to come up with a joke that the
[29:11] >> Trying to come up with a joke that the autist in San Francisco are keenly
[29:13] autist in San Francisco are keenly
[29:13] autist in San Francisco are keenly keenly waiting for this moment for a
[29:15] keenly waiting for this moment for a
[29:15] keenly waiting for this moment for a very a very long time. Uh
[29:18] very a very long time. Uh
[29:18] very a very long time. Uh there's a there's a good um a question
[29:20] there's a there's a good um a question
[29:20] there's a there's a good um a question from Michelle here. You know, how should
[29:21] from Michelle here. You know, how should
[29:21] from Michelle here. You know, how should we think about the new economics of AI
[29:23] we think about the new economics of AI
[29:23] we think about the new economics of AI apps companies because there's a great
[29:25] apps companies because there's a great
[29:25] apps companies because there's a great there's a debate around unit economics
[29:27] there's a debate around unit economics
[29:27] there's a debate around unit economics of of um apps companies, right? like for
[29:29] of of um apps companies, right? like for
[29:29] of of um apps companies, right? like for example, they may may have lower gross
[29:30] example, they may may have lower gross
[29:30] example, they may may have lower gross margins. They're just getting more
[29:31] margins. They're just getting more
[29:31] margins. They're just getting more pressure just because they don't have as
[29:33] pressure just because they don't have as
[29:33] pressure just because they don't have as much compute access. Um capital is such
[29:35] much compute access. Um capital is such
[29:35] much compute access. Um capital is such a moat in this environment. Um it's it's
[29:37] a moat in this environment. Um it's it's
[29:37] a moat in this environment. Um it's it's hard to be competitive. So how do you
[29:39] hard to be competitive. So how do you
[29:39] hard to be competitive. So how do you think about the economics of of
[29:40] think about the economics of of
[29:40] think about the economics of of underwriting returns in in um in
[29:42] underwriting returns in in um in
[29:42] underwriting returns in in um in companies today?
[29:44] companies today?
[29:44] companies today? >> I mean David wrote a great post on this.
[29:45] >> I mean David wrote a great post on this.
[29:46] >> I mean David wrote a great post on this. I think that the kind of the margin
[29:47] I think that the kind of the margin
[29:47] I think that the kind of the margin topic is a lot more nuanced than it once
[29:49] topic is a lot more nuanced than it once
[29:49] topic is a lot more nuanced than it once was. I think it's actually rational in
[29:51] was. I think it's actually rational in
[29:51] was. I think it's actually rational in many cases to trade away margin to have
[29:53] many cases to trade away margin to have
[29:53] many cases to trade away margin to have wider product surface. I think the very
[29:55] wider product surface. I think the very
[29:55] wider product surface. I think the very positive part of what's happening in
[29:57] positive part of what's happening in
[29:57] positive part of what's happening in this product cycle is the willingness to
[29:58] this product cycle is the willingness to
[29:58] this product cycle is the willingness to pay is extraordinary. And that's why the
[30:00] pay is extraordinary. And that's why the
[30:00] pay is extraordinary. And that's why the exercise that we often do with founders
[30:02] exercise that we often do with founders
[30:02] exercise that we often do with founders is like on the consumer side, for
[30:03] is like on the consumer side, for
[30:03] is like on the consumer side, for example, is if $20 was the historic
[30:06] example, is if $20 was the historic
[30:06] example, is if $20 was the historic ceiling, what's the $200 a month skew of
[30:08] ceiling, what's the $200 a month skew of
[30:08] ceiling, what's the $200 a month skew of your product? And in fact, what's the
[30:10] your product? And in fact, what's the
[30:10] your product? And in fact, what's the $2,000 a month skew? Like what's the
[30:11] $2,000 a month skew? Like what's the
[30:12] $2,000 a month skew? Like what's the Birkin bag of software? I think we're
[30:13] Birkin bag of software? I think we're
[30:13] Birkin bag of software? I think we're going to have this luxury software.
[30:15] going to have this luxury software.
[30:15] going to have this luxury software. We're already seeing willingness to pay
[30:16] We're already seeing willingness to pay
[30:16] We're already seeing willingness to pay for it. So the margin topic is more
[30:19] for it. So the margin topic is more
[30:19] for it. So the margin topic is more nuanced, but the willingness to pay and
[30:20] nuanced, but the willingness to pay and
[30:20] nuanced, but the willingness to pay and buy is higher than ever. So, you know,
[30:22] buy is higher than ever. So, you know,
[30:22] buy is higher than ever. So, you know, it's a little bit of fog of war, but
[30:24] it's a little bit of fog of war, but
[30:24] it's a little bit of fog of war, but we're we're thinking about all those
[30:25] we're we're thinking about all those
[30:25] we're we're thinking about all those topics.
[30:26] topics.
[30:26] topics. >> Anish dropping Birkin bag framed jeans.
[30:29] >> Anish dropping Birkin bag framed jeans.
[30:29] >> Anish dropping Birkin bag framed jeans. Like, I had no idea you were such a
[30:30] Like, I had no idea you were such a
[30:30] Like, I had no idea you were such a fashion. This is like your your butt is
[30:32] fashion. This is like your your butt is
[30:32] fashion. This is like your your butt is helping you get up to to seat here, my
[30:34] helping you get up to to seat here, my
[30:34] helping you get up to to seat here, my friend. Uh for a guy secret
[30:38] friend. Uh for a guy secret
[30:38] friend. Uh for a guy secret is a good steward of capital. Okay,
[30:40] is a good steward of capital. Okay,
[30:40] is a good steward of capital. Okay, that's all that I am.
[30:41] that's all that I am.
[30:41] that's all that I am. >> For for a guy I only see in quarter zip
[30:43] >> For for a guy I only see in quarter zip
[30:43] >> For for a guy I only see in quarter zip ups. I'm just saying. Uh
[30:46] ups. I'm just saying. Uh
[30:46] ups. I'm just saying. Uh >> um Okay, maybe one question for you on
[30:48] >> um Okay, maybe one question for you on
[30:48] >> um Okay, maybe one question for you on just on the on the founders because I I
[30:49] just on the on the founders because I I
[30:49] just on the on the founders because I I don't know if you remember this
[30:50] don't know if you remember this
[30:50] don't know if you remember this conversation. This is probably 5 years
[30:52] conversation. This is probably 5 years
[30:52] conversation. This is probably 5 years ago or so. Um where most of the founders
[30:55] ago or so. Um where most of the founders
[30:55] ago or so. Um where most of the founders you saw saw more uh diversity in their
[30:59] you saw saw more uh diversity in their
[30:59] you saw saw more uh diversity in their background in part because the software
[31:02] background in part because the software
[31:02] background in part because the software and technology was way more
[31:03] and technology was way more
[31:03] and technology was way more sophisticated. So you had a lot of
[31:04] sophisticated. So you had a lot of
[31:04] sophisticated. So you had a lot of program managers spinning out of Google
[31:06] program managers spinning out of Google
[31:06] program managers spinning out of Google for example and starting a company etc.
[31:07] for example and starting a company etc.
[31:08] for example and starting a company etc. What are the type of founders you see
[31:09] What are the type of founders you see
[31:09] What are the type of founders you see building an apps today? Are they do they
[31:11] building an apps today? Are they do they
[31:11] building an apps today? Are they do they tend to lean you know more technical
[31:13] tend to lean you know more technical
[31:13] tend to lean you know more technical more researcher derivatives? Are they
[31:15] more researcher derivatives? Are they
[31:16] more researcher derivatives? Are they product managers? like what what kind of
[31:17] product managers? like what what kind of
[31:17] product managers? like what what kind of archetype are you seeing at least the
[31:20] archetype are you seeing at least the
[31:20] archetype are you seeing at least the early innings of apps come out from the
[31:22] early innings of apps come out from the
[31:22] early innings of apps come out from the woodwork on?
[31:23] woodwork on?
[31:23] woodwork on? >> Yeah. Yeah. Less MBAs, more researchers.
[31:25] >> Yeah. Yeah. Less MBAs, more researchers.
[31:26] >> Yeah. Yeah. Less MBAs, more researchers. Um and they both have their kind of
[31:27] Um and they both have their kind of
[31:27] Um and they both have their kind of strengths and weaknesses. I think the
[31:29] strengths and weaknesses. I think the
[31:29] strengths and weaknesses. I think the business sophistication of the founders
[31:31] business sophistication of the founders
[31:31] business sophistication of the founders are seeing today is lower, but the kind
[31:33] are seeing today is lower, but the kind
[31:33] are seeing today is lower, but the kind of technical sophistication is
[31:34] of technical sophistication is
[31:34] of technical sophistication is dramatically higher and the technical
[31:36] dramatically higher and the technical
[31:36] dramatically higher and the technical sophistication is kind of upstream of
[31:38] sophistication is kind of upstream of
[31:38] sophistication is kind of upstream of all the good things that happened. You
[31:40] all the good things that happened. You
[31:40] all the good things that happened. You know, business sophistication can be
[31:41] know, business sophistication can be
[31:41] know, business sophistication can be kind of taught and observed, but
[31:43] kind of taught and observed, but
[31:43] kind of taught and observed, but technical sophistication typically not.
[31:45] technical sophistication typically not.
[31:45] technical sophistication typically not. So, we're definitely seeing a much more
[31:46] So, we're definitely seeing a much more
[31:46] So, we're definitely seeing a much more technical kind of earlier career
[31:48] technical kind of earlier career
[31:48] technical kind of earlier career founder, but the things they're doing
[31:50] founder, but the things they're doing
[31:50] founder, but the things they're doing are extraordinary because they don't
[31:52] are extraordinary because they don't
[31:52] are extraordinary because they don't have any sort of preconceived notions
[31:53] have any sort of preconceived notions
[31:54] have any sort of preconceived notions about what's possible. And so much of
[31:55] about what's possible. And so much of
[31:56] about what's possible. And so much of what holds back senior founders that
[31:57] what holds back senior founders that
[31:57] what holds back senior founders that don't quite get to the other side of
[31:58] don't quite get to the other side of
[31:58] don't quite get to the other side of this product cycle is, you know, they're
[32:00] this product cycle is, you know, they're
[32:00] this product cycle is, you know, they're not close enough to the technology and
[32:02] not close enough to the technology and
[32:02] not close enough to the technology and they've got an idea that's rooted in the
[32:04] they've got an idea that's rooted in the
[32:04] they've got an idea that's rooted in the past of what the ceiling is. And I think
[32:06] past of what the ceiling is. And I think
[32:06] past of what the ceiling is. And I think the best thing about these young
[32:07] the best thing about these young
[32:07] the best thing about these young founders is they assume everything is
[32:08] founders is they assume everything is
[32:08] founders is they assume everything is possible. You know, we are at an offsite
[32:10] possible. You know, we are at an offsite
[32:10] possible. You know, we are at an offsite where Ben was saying that the the
[32:12] where Ben was saying that the the
[32:12] where Ben was saying that the the biggest risk in the past with the ideas
[32:13] biggest risk in the past with the ideas
[32:13] biggest risk in the past with the ideas were too big and now the biggest risk is
[32:15] were too big and now the biggest risk is
[32:15] were too big and now the biggest risk is that the ideas are too small. But I
[32:17] that the ideas are too small. But I
[32:17] that the ideas are too small. But I think that's sort of illustrative of the
[32:18] think that's sort of illustrative of the
[32:18] think that's sort of illustrative of the different founder archetypes.
[32:20] different founder archetypes.
[32:20] different founder archetypes. >> Yep. Yep. And maybe on that similar
[32:22] >> Yep. Yep. And maybe on that similar
[32:22] >> Yep. Yep. And maybe on that similar thread, it used to be that that if you
[32:24] thread, it used to be that that if you
[32:24] thread, it used to be that that if you gave a founder too much money, it would
[32:26] gave a founder too much money, it would
[32:26] gave a founder too much money, it would wreck the company because the founder
[32:28] wreck the company because the founder
[32:28] wreck the company because the founder almost always has way too many ideas and
[32:31] almost always has way too many ideas and
[32:31] almost always has way too many ideas and is a visionary and doesn't have the
[32:33] is a visionary and doesn't have the
[32:33] is a visionary and doesn't have the talent to actually commensurately
[32:36] talent to actually commensurately
[32:36] talent to actually commensurately land with all those ideas. Um and we're
[32:39] land with all those ideas. Um and we're
[32:39] land with all those ideas. Um and we're seeing a whole new paradigm on that.
[32:40] seeing a whole new paradigm on that.
[32:40] seeing a whole new paradigm on that. Maybe unpack that idea a little bit more
[32:42] Maybe unpack that idea a little bit more
[32:42] Maybe unpack that idea a little bit more just uh because it was such a huge theme
[32:43] just uh because it was such a huge theme
[32:43] just uh because it was such a huge theme of the offsite.
[32:44] of the offsite.
[32:44] of the offsite. >> Yeah, I mean for sure this was a
[32:46] >> Yeah, I mean for sure this was a
[32:46] >> Yeah, I mean for sure this was a historic wisdom. You know, why didn't we
[32:47] historic wisdom. You know, why didn't we
[32:47] historic wisdom. You know, why didn't we give every seed company 20 or 50 or
[32:49] give every seed company 20 or 50 or
[32:49] give every seed company 20 or 50 or hundred million dollars? You know, it
[32:50] hundred million dollars? You know, it
[32:50] hundred million dollars? You know, it wasn't just the kind of riskreward, but
[32:52] wasn't just the kind of riskreward, but
[32:52] wasn't just the kind of riskreward, but rather typically the constraining factor
[32:54] rather typically the constraining factor
[32:54] rather typically the constraining factor was they just didn't have enough
[32:55] was they just didn't have enough
[32:55] was they just didn't have enough talented people to work across $20
[32:58] talented people to work across $20
[32:58] talented people to work across $20 million of product surface at the same
[32:59] million of product surface at the same
[32:59] million of product surface at the same time. They really had to focus on one
[33:01] time. They really had to focus on one
[33:01] time. They really had to focus on one idea at a time and the capital was a
[33:03] idea at a time and the capital was a
[33:03] idea at a time and the capital was a great way to enforce that focus. What
[33:05] great way to enforce that focus. What
[33:05] great way to enforce that focus. What we're now seeing is you could make
[33:07] we're now seeing is you could make
[33:07] we're now seeing is you could make different sort of product and model
[33:09] different sort of product and model
[33:09] different sort of product and model trade-offs through more or less capital.
[33:11] trade-offs through more or less capital.
[33:11] trade-offs through more or less capital. And there is a case for a company that
[33:13] And there is a case for a company that
[33:13] And there is a case for a company that raises a hund00 million, uses it
[33:15] raises a hund00 million, uses it
[33:15] raises a hund00 million, uses it productively and in a focused way and is
[33:17] productively and in a focused way and is
[33:17] productively and in a focused way and is able to deliver a different value
[33:19] able to deliver a different value
[33:19] able to deliver a different value proposition um than the very same team
[33:21] proposition um than the very same team
[33:21] proposition um than the very same team would be able to do with 20. So I think
[33:23] would be able to do with 20. So I think
[33:23] would be able to do with 20. So I think that that again like the sort of just as
[33:24] that that again like the sort of just as
[33:24] that that again like the sort of just as we talked about sort of fog of war
[33:27] we talked about sort of fog of war
[33:27] we talked about sort of fog of war around margins, I think this question of
[33:29] around margins, I think this question of
[33:29] around margins, I think this question of what is the optimal seated around size
[33:30] what is the optimal seated around size
[33:30] what is the optimal seated around size and how much capital can you put to work
[33:32] and how much capital can you put to work
[33:32] and how much capital can you put to work effectively is a much more nuanced
[33:34] effectively is a much more nuanced
[33:34] effectively is a much more nuanced topic. I mean, it's sort of this
[33:35] topic. I mean, it's sort of this
[33:35] topic. I mean, it's sort of this embarrassment of riches, but I'd rather
[33:37] embarrassment of riches, but I'd rather
[33:37] embarrassment of riches, but I'd rather have this problem than the problem we
[33:39] have this problem than the problem we
[33:39] have this problem than the problem we had five years ago, which is, hey, my
[33:41] had five years ago, which is, hey, my
[33:41] had five years ago, which is, hey, my fintech company is indirectly
[33:43] fintech company is indirectly
[33:43] fintech company is indirectly subsidizing their customers through weak
[33:44] subsidizing their customers through weak
[33:44] subsidizing their customers through weak underwriting, and we don't know the path
[33:46] underwriting, and we don't know the path
[33:46] underwriting, and we don't know the path home.
[33:46] home.
[33:46] home. >> Yeah. Yep. Yeah. The Chris Chris Dixon
[33:49] >> Yeah. Yep. Yeah. The Chris Chris Dixon
[33:49] >> Yeah. Yep. Yeah. The Chris Chris Dixon model, which is you always want the
[33:50] model, which is you always want the
[33:50] model, which is you always want the problem of supply, not of demand. Right.
[33:52] problem of supply, not of demand. Right.
[33:52] problem of supply, not of demand. Right. Right now, we have to fix the supply
[33:54] Right now, we have to fix the supply
[33:54] Right now, we have to fix the supply part, right? Uh but the demand is like
[33:56] part, right? Uh but the demand is like
[33:56] part, right? Uh but the demand is like uh so so abundantly there that that uh
[33:58] uh so so abundantly there that that uh
[33:58] uh so so abundantly there that that uh undoubtedly that will um the supply part
[34:01] undoubtedly that will um the supply part
[34:01] undoubtedly that will um the supply part will will get fixed. Um maybe I'll close
[34:03] will will get fixed. Um maybe I'll close
[34:03] will will get fixed. Um maybe I'll close on this one last uh question for Mosfa.
[34:05] on this one last uh question for Mosfa.
[34:05] on this one last uh question for Mosfa. So um double clicking on theme sector
[34:08] So um double clicking on theme sector
[34:08] So um double clicking on theme sector adoption of AI. So unlike large
[34:09] adoption of AI. So unlike large
[34:09] adoption of AI. So unlike large enterprise the friction of adoption is
[34:11] enterprise the friction of adoption is
[34:11] enterprise the friction of adoption is much less because they require less
[34:13] much less because they require less
[34:13] much less because they require less change management. I agree with many of
[34:15] change management. I agree with many of
[34:15] change management. I agree with many of that but uh not not all uh small small
[34:20] that but uh not not all uh small small
[34:20] that but uh not not all uh small small uh medium businesses sometimes have uh
[34:22] uh medium businesses sometimes have uh
[34:22] uh medium businesses sometimes have uh more habit change that you got to work
[34:23] more habit change that you got to work
[34:23] more habit change that you got to work through. But the question is how do you
[34:24] through. But the question is how do you
[34:24] through. But the question is how do you see the gotom market playbook for
[34:26] see the gotom market playbook for
[34:26] see the gotom market playbook for startups targetingmemes and has that
[34:29] startups targetingmemes and has that
[34:29] startups targetingmemes and has that changed in the age of AI? I mean a lot
[34:32] changed in the age of AI? I mean a lot
[34:32] changed in the age of AI? I mean a lot of it for existing thememes I think it's
[34:33] of it for existing thememes I think it's
[34:33] of it for existing thememes I think it's the same channels with which you
[34:35] the same channels with which you
[34:35] the same channels with which you historically reach them. I actually
[34:36] historically reach them. I actually
[34:36] historically reach them. I actually think that one of the interesting things
[34:37] think that one of the interesting things
[34:38] think that one of the interesting things about marketing in the age of AI is that
[34:40] about marketing in the age of AI is that
[34:40] about marketing in the age of AI is that all of the sort of existing networks
[34:42] all of the sort of existing networks
[34:42] all of the sort of existing networks have been so trained on the methodology
[34:44] have been so trained on the methodology
[34:44] have been so trained on the methodology of building new networks that they're
[34:46] of building new networks that they're
[34:46] of building new networks that they're very careful to ensure no one does it on
[34:48] very careful to ensure no one does it on
[34:48] very careful to ensure no one does it on their network. So Instagram, Tik Tok X,
[34:51] their network. So Instagram, Tik Tok X,
[34:52] their network. So Instagram, Tik Tok X, it's very hard to build a new sort of
[34:53] it's very hard to build a new sort of
[34:53] it's very hard to build a new sort of distribution channel off the backs of an
[34:55] distribution channel off the backs of an
[34:55] distribution channel off the backs of an existing one. So what founders have to
[34:57] existing one. So what founders have to
[34:57] existing one. So what founders have to do is actually build a product that has
[34:58] do is actually build a product that has
[34:58] do is actually build a product that has the original network effect which is
[35:00] the original network effect which is
[35:00] the original network effect which is word of mouth. So we're definitely
[35:01] word of mouth. So we're definitely
[35:01] word of mouth. So we're definitely seeing more of a focus on word of mouth.
[35:03] seeing more of a focus on word of mouth.
[35:03] seeing more of a focus on word of mouth. Yes, the kind of old channels for
[35:04] Yes, the kind of old channels for
[35:04] Yes, the kind of old channels for reaching are still there. Actually think
[35:06] reaching are still there. Actually think
[35:06] reaching are still there. Actually think the most interesting segment of the
[35:08] the most interesting segment of the
[35:08] the most interesting segment of the market though is sort of new business
[35:10] market though is sort of new business
[35:10] market though is sort of new business formation which is by the way at an
[35:11] formation which is by the way at an
[35:11] formation which is by the way at an all-time high. I think it's the highest
[35:12] all-time high. I think it's the highest
[35:12] all-time high. I think it's the highest it's been outside of a peak sort of
[35:14] it's been outside of a peak sort of
[35:14] it's been outside of a peak sort of moment during COVID. These are people
[35:16] moment during COVID. These are people
[35:16] moment during COVID. These are people who would have never otherwise been.
[35:19] who would have never otherwise been.
[35:19] who would have never otherwise been. It's not the sort of 55year-old plumber.
[35:21] It's not the sort of 55year-old plumber.
[35:21] It's not the sort of 55year-old plumber. It's a 25year-old who previously would
[35:23] It's a 25year-old who previously would
[35:23] It's a 25year-old who previously would have been a YouTube creator and now is
[35:24] have been a YouTube creator and now is
[35:24] have been a YouTube creator and now is building SAS for their, you know,
[35:26] building SAS for their, you know,
[35:26] building SAS for their, you know, neighborhood or their city or their high
[35:28] neighborhood or their city or their high
[35:28] neighborhood or their city or their high school or whatever else it is.
[35:29] school or whatever else it is.
[35:30] school or whatever else it is. >> Yep. Yep. Awesome.
[35:32] >> Yep. Yep. Awesome.
[35:32] >> Yep. Yep. Awesome. Well, thank you so much for listen. It's
[35:33] Well, thank you so much for listen. It's
[35:33] Well, thank you so much for listen. It's always great to have you on. Uh now, I
[35:35] always great to have you on. Uh now, I
[35:35] always great to have you on. Uh now, I know you're a fashionista and we're
[35:36] know you're a fashionista and we're
[35:36] know you're a fashionista and we're going to be clipping that endlessly uh
[35:37] going to be clipping that endlessly uh
[35:37] going to be clipping that endlessly uh on the socials. Um but uh but thank you
[35:40] on the socials. Um but uh but thank you
[35:40] on the socials. Um but uh but thank you for that. And if folks have any
[35:41] for that. And if folks have any
[35:41] for that. And if folks have any questions, you know where to find Anish.
[35:43] questions, you know where to find Anish.
[35:43] questions, you know where to find Anish. Um and uh we'll follow up here for some
[35:45] Um and uh we'll follow up here for some
[35:45] Um and uh we'll follow up here for some of the questions we weren't able to get
[35:46] of the questions we weren't able to get
[35:46] of the questions we weren't able to get to as well.
