AI Capital Abandons Base Models for Physical Robotics, Gong Says

Where AI Money Actually Flows With Smart Venture Media — Ai4

Ai4

Venture capital is bypassing foundation models because defensible enterprise value accrues to application founders who pair deep domain judgment with proprietary data flywheels in trillion-dollar legacy industries.

AI-assisted summary of Where AI Money Actually Flows With Smart Venture Media — original source. Check the original for full context.

Source published: 2026-08-19

Detailed notes

## Background Published on YouTube by the Ai4 channel on August 19, 2026, this presentation features Grace Gong, venture capitalist and founder of Smart Venture Media. According to the video description, the talk examines where AI capital is actively moving—highlighting physical-world autonomy, agent-native software, and trillion-dollar legacy industries over foundation model bets.

## TL;DR Venture investor Grace Gong outlines the key trends directing capital flows across the artificial intelligence landscape, arguing that outsized value will accrue at the application layer rather than in foundation models. She details major market shifts toward hardware-as-a-service (physical AI), the digitization of massive legacy industries like logistics and insurance, agent-native software stacks, and enterprise voice automation. Gong contends that defensibility against upstream AI models relies on domain taste, proprietary distribution, and operational data flywheels rather than technical complexity alone. She concludes with an open Q&A addressing early customer validation, Silicon Valley pedigree bias, and the transition toward outcome-based software pricing.

## Atmosphere & style Gong delivers a fast-paced, highly engaging, and refreshingly candid talk punctuated by self-deprecating humor about her daily routine of podcast interviewing. The room dynamic is lively and interactive, highlighted by audience participation exercises that pivot attention away from speculative venture hype toward practical operator problem-solving. Her delivery combines insider Silicon Valley realism with pragmatic, no-nonsense advice, creating an empowering and down-to-earth atmosphere.

## Key takeaways - [01:08](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=68s) Capital is heavily flowing toward 'HaaS is the new SaaS' (Physical AI), integrating autonomous intelligence into heavy industrial machinery, robotics, and physical workflows. - [02:05](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=125s) Overlooked, paper-bound legacy industries—such as construction ($13T), logistics ($9.5T), and insurance ($7.4T)—present the most lucrative opportunities for AI-native infrastructure. - [03:45](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=225s) Software design and data infrastructure are pivoting away from human-centric GUIs toward agent-native architectures built explicitly for autonomous AI agents. - [05:10](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=310s) Enterprise voice AI is maturing far beyond customer support deflection into mission-critical systems that capture immediate, high-value commercial revenue. - [07:51](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=471s) The true future of AI work is multiplayer and collaborative, where agents function as integrated team members amplifying human colleagues rather than isolating solo founders. - [09:56](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=596s) Application layer defensibility against upstream foundation models relies on earned domain 'taste,' last-mile workflow specialization, and proprietary distribution channels. - [12:58](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=778s) Corporate operators who use accessible AI tools to solve their own internal bottlenecks build defensible data flywheels and validated venture architectures. - [14:51](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=891s) Customer willingness to pay is the only true litmus test for early startup demand, as unpaid usage provides virtually no signal of product necessity. - [16:14](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=974s) Software business models are transitioning from per-seat SaaS subscriptions toward outcome- and service-based pricing for autonomous 'AI employees.' - [18:35](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1115s) Massive seed funding rounds often stem from VC pedigree bias and institutional risk aversion rather than proven product-market fit. - [21:43](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1303s) Fast-following copycats are inevitable; enduring defensibility depends on operating velocity, continuous data flywheels, and an unreplicable layered stack of knowledge. - [22:24](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1344s) Macroeconomic bubble debates and multi-year forecasting matter far less to operators than daily hands-on tool usage, customer value creation, and consistent revenue generation.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=4s) Introduction: Front-Row Seat to Silicon Valley's Capital Flows

Grace Gong opens her session by introducing herself as the founder of Smart Venture Media and host of the *Smart Venture Podcast*. Her daily routine involves going live on LinkedIn to interview three core profiles across the tech landscape: top-tier venture capitalists, C-suite executives and vice presidents from major technology firms, and leading artificial intelligence founders. Her interview roster spans senior leaders from tech giants such as Google, Meta, Amazon, Snowflake, and Nvidia, alongside partners from venture firms including Sequoia Capital, Andreessen Horowitz (a16z), Lightspeed Venture Partners, Bessemer Venture Partners, and NEA. She has also interviewed the founders behind ubiquitous platforms and AI unicorns like Square, Loom, Cruise, Glean, Craigslist, Hinge Health, Transfix, Corgi, and Crusoe.

Framing her presentation with self-deprecating humor, Gong explains that her insights into capital allocation come directly from her daily routine of "yapping all day" with industry leaders, coupled with her direct experience writing checks as an active angel investor in Silicon Valley. Her talk focuses squarely on demystifying where capital is actively moving across the AI landscape, identifying the specific technological shifts and business models capturing meaningful venture backing.

[01:08](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=68s) Trend 1: HaaS Is the New SaaS (Physical AI)

The first major macro trend Gong highlights is the evolution of "HaaS is the new SaaS"—hardware as a service serving as the commercial vehicle for physical AI. Investors are pouring substantial capital into companies that integrate advanced autonomy into heavy physical machinery and industrial robotics rather than confining AI models to digital browser windows.

A prime example she features is Bedrock Robotics, a startup founded by alumni of Waymo. The team is effectively building the Waymo equivalent for heavy earth-moving equipment, turning everyday construction machinery into autonomous bulldozers and excavators. Gong also cites Dexterity, an industrial robotics startup valued at $1.6 billion that manufactures AI-powered robotic systems engineered to streamline physical warehouse and manufacturing operations. Rather than selling one-off hardware, these companies deploy physical robots embedded with sophisticated perceptual and decision-making intelligence to automate mission-critical physical workflows.

[02:05](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=125s) Trend 2: Boring Industries, Trillion-Dollar Outcomes

Gong moves to her second major trend, which she describes as "boring industries, big outcomes," noting that while she uses more colorful language at other events, the cleaned-up version for Ai4 is that "the unsexy industry is the new sexy industry." She presents massive total addressable market figures to illustrate why investors are turning away from consumer gimmicks toward legacy verticals:

- **Global Construction:** $13 trillion - **Global Logistics:** $9.5 trillion - **Global Insurance:** $7.4 trillion

Despite managing trillions of dollars in assets, massive swaths of the insurance sector still rely heavily on legacy infrastructure and physical fax machines. Gong points to Corgi, an AI-native commercial insurance startup that raised over $100 million to reconstruct the insurance operating stack from the ground up rather than selling a superficial digital point solution.

Additional momentum is building in back-office infrastructure: Doss raised over $70 million led by General Catalyst to build an AI-native ERP system, while Serval emerged from Smart Venture Media's "Smart AI 100" list to raise two consecutive venture rounds from Sequoia Capital to automate IT service ticketing. For Silicon Valley investors, the most lucrative opportunities lie in replacing manual, paper-bound friction inside these overlooked legacy domains.

[03:45](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=225s) Trend 3: Software Built for AI Agents, Not Humans

Turning to software design, Gong surveys the audience to confirm near-unanimous usage of human-centric creative tools like Canva and Figma. She argues that while these legacy platforms were intentionally designed for human interaction via graphic user interfaces, the next computing paradigm centers on software designed explicitly for autonomous software agents.

Because background AI agents are increasingly handling complex, multi-step workflows without real-time human intervention, an entirely new agent-native software stack is emerging. Gong highlights Moda, which she describes as an AI design engine engineered specifically for AI agents to interact with and generate assets for other agents. On the data infrastructure side, companies like TensorMesh are utilizing key-value (KV) caching to optimize data storage, retrieval, and contextual recall for autonomous agents. While legacy data platforms like Snowflake and Databricks were engineered to help human data engineers query and manage enterprise datasets, the next wave of infrastructure is being built to allow software agents to process and orchestrate massive streams of information natively.

[05:10](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=310s) Trend 4: Voice AI Beyond the Call Center

Voice technology is experiencing an unprecedented capital inflection point, moving well beyond basic consumer dictation tools like WhisperFlow toward full-scale enterprise automation. Gong highlights mega-rounds and nine-figure valuations going to customer service and communication automation platforms such as Sierra, Cresta, Forethought, and Pylon. These platforms deploy voice intelligence to completely modernize call centers, eliminating the dreaded multi-hour waits consumers traditionally face for routine tasks like product returns, medical appointments, and internet service troubleshooting.

To illustrate the immediate economic impact, Gong shares a case study from Hosty, an AI restaurant voice platform in which she is an angel investor alongside Gradient Ventures (Google's AI-focused venture fund):

```mermaid flowchart TD accTitle: Hosty AI Voice Agent Flow subgraph Inbound["Inbound Call"] A["Customer Calls for 70-Person Booking"] --> B["Traditional Competitors: Busy signal or 'Need to check with boss'"] A --> C["Hosty Voice Agent Answers Immediately"] end subgraph Processing["Agentic Handling"] C --> D["Captures Full Party Details & Takes Actionable Notes"] D --> E["Pushes Instant Notification to Restaurant Owner"] end subgraph Conversion["Instant Revenue"] E --> F["Owner Confirms Booking Within Minutes"] F --> G["Secures Lucrative 70-Person Event Revenue"] end ```

During a chaotic graduation rush, an organizer attempting to reserve space for a 70-person party repeatedly called neighborhood restaurants, only to be met with unanswered lines or staff unable to confirm large bookings without their manager. Hosty answered the call instantly at a partner restaurant, gathered the reservation criteria, transcribed the booking requirements into an immediate alert for the owner, and allowed the restaurant to lock in a massive group order that would have otherwise walked away.

[07:51](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=471s) Trend 5: From Solo Unicorns to Multiplayer Collaboration

Gong addresses the widespread Silicon Valley speculation surrounding the imminent rise of the "one-person billion-dollar company." She points to a widely covered *Wall Street Journal* story detailing two brothers who leveraged automated AI telemarketing agents to build a billion-dollar distribution company for Ozempic, sparking a wave of software platforms like Pozeia designed to help solo founders scale one-person empires.

However, Gong pushes back against the premise that solo entrepreneurship represents the desirable future of work. Recounting an executive dinner with the CMO of Slack discussing agent integration frameworks and the Model Context Protocol (MCP), she argues that the real frontier is collaborative and multiplayer. Human beings thrive on shared mission, camaraderie, and team dynamics to sustain motivation through difficult operating days. In this model, autonomous agents do not replace human coworkers or isolate individuals in siloes; instead, they operate as intelligent new team members that collaborate alongside humans to amplify team productivity.

[09:56](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=596s) Winning the Application Layer: Taste, Distribution, and the Last Mile

While foundational model creators like OpenAI (ChatGPT), Anthropic (Claude), and Perplexity have established core cognitive capabilities, Gong asserts that the vast majority of enterprise value will be captured in the application layer. She notes the persistent fear among founders of being "Claude-shotted"—Silicon Valley shorthand for an application being rendered obsolete overnight by an upstream Anthropic or OpenAI model release. Despite this, Gong argues that no single foundational model will ever monopolize every industry's specialized workflows:

```mermaid flowchart LR accTitle: Application Layer Moat subgraph Foundations["Base Layer"] M["Foundation Models (OpenAI, Anthropic)"] end subgraph Differentiation["Defensible Application Layer"] M --> L["Last-Mile Specialization"] L --> T["Earned Taste & Industry Judgment"] L --> D["Proprietary Distribution Engine"] end subgraph Enterprise["Enterprise Moat"] T & D --> W["Workflow Integration & Data Flywheel"] end ```

Because modern coding assistants allow anyone to build functional software rapidly, technical execution alone is no longer a sustainable moat. True differentiation now depends on two qualities: - **Taste:** The deep, earned judgment gained from hands-on industry experience (as a physician, attorney, engineer, or specialized journalist) that dictates what products should be built and how workflows actually function. - **Distribution:** The ability to capture mindshare and audience directly. Gong highlights Cluey, a startup that generated viral engagement by writing and filming episodic, *The Office*-style workplace mockumentaries to market their B2B software.

Founders can replicate this distribution edge on a manageable scale by publicly sharing unique operational insights, industry trends, and authentic behind-the-scenes accounts of their operational journeys.

[12:58](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=778s) The Operator Flywheel and Standing Out in the Room

Gong pauses her presentation to run an interactive exercise, asking the audience to stand up before systematically instructing founders, investors, and then company operators to sit down. When a large contingent remains standing, Gong jokes with an attendee, clarifying that "if you work at a company, you're an operator!"

With operators making up the clear majority of the room, Gong tailors her advice directly to corporate employees. By pairing their deep domain knowledge with accessible AI platforms like Claude, Gamma, and Clay, operators can build bespoke internal software that directly addresses their team's bottlenecks.

By building solutions to their own daily problems, operators trigger an internal data flywheel: they become the primary user, roll the tool out to colleagues and clients, collect rapid user feedback, and iterate rapidly. This hands-on process builds a defensible operational moat—and potentially provides the validated architecture for a venture-backed startup down the road.

[14:51](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=891s) Q&A: Validating Early Demand Beyond Family and Friends

An audience member asks Gong how early-stage founders should approach initial fundraising and customer discovery after exhausting personal networks. Gong gives unambiguous advice: founders must bypass friends and family entirely to find objective, third-party validation.

When searching for authentic early design partners, Gong recommends embedding directly where target users congregate, whether attending local industry meetups, participating in specialized design workshops, or joining professional Facebook groups. Her critical litmus test for true customer validation is simple: **you must get them to pay.** From an investor's diligence perspective, unpaid usage metrics provide virtually no signal regarding product necessity or willingness to adopt. Only genuine customer capital validates whether a software solution is solving a problem urgent enough to sustain a scalable venture.

[16:14](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=974s) Q&A: Personal Distribution and the Outcome-Based Pricing Shift

An audience member asks whether building personal distribution and public visibility is truly relevant for corporate operators who are not currently fundraising. Gong enthusiastically affirms the strategy: "Hell yeah! I'm sorry, I don't know how to say this in a more calm way, but definitely."

Describing herself as an unapologetic "LinkedIn warrior," Gong explains that the entire corporate workforce is undergoing a profound structural shift driven by AI agents. Software vendors are aggressively pivoting away from standard per-seat SaaS subscription models (such as Canva's $12.99 per month tier) toward outcome- and service-based pricing models:

- Software vendors deploy autonomous agents packaged as "AI employees." - Instead of charging per user seat, vendors charge $10,000 per month based on the deliverable or business outcome (e.g., functioning as a full-time in-house graphic designer).

In this environment, operators who build a public profile as domain thought leaders preserve clear strategic leverage within their organizations. Gong emphasizes that Silicon Valley's unique competitive advantage stems from a ubiquitous "builder mindset" where non-technical employees, from sales reps to growth marketers, routinely build custom CRMs and automated Airtable databases at home. Publicly sharing insights creates a healthy forcing function: "When you have to output, you really have to input." Building in public forces operators to continually test emerging AI tooling, engage top industry practitioners, and retain an indispensable edge over peers.

[18:35](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1115s) Q&A: Pedigree Bias in Venture Capital and False Validation

A female attendee asks why certain founders manage to raise massive venture rounds with little more than an unproven concept. Gong gives a candid insider response: "I have an answer that you probably don't want to hear."

She reveals that Silicon Valley investors overwhelmingly invest in founder pedigree and institutional logos. Backing a proven, second-time founder—such as Travis Kalanick of Uber—serves as an institutional shortcut that saves venture partners weeks of intensive due diligence while insulating them from reputational risk. It is far safer internally for a partner to lose fund capital on an executive with an established track record than to gamble on an unknown, first-time entrepreneur.

However, Gong warns that securing massive upfront venture capital is frequently a vanity metric that harms the founder: - Substantial seed capital creates an illusion of commercial validation without proving product-market fit. - Legendary breakout startups rarely begin with unanimous institutional approval; their initial greatness stems from surviving skepticism and proving their thesis through gritty execution, product iteration, and proprietary distribution. - Capital volume alone has zero correlation with eventual startup success.

Noticing several audience members standing up to exit the room as she discusses large fundraises, Gong playfully quips, "Did all of you raise a lot of money? Is that why you're leaving?"

[20:38](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1238s) Q&A: Outrunning Copycats Through Speed and Data Flywheels

The session wraps with an attendee asking how an early startup can defend its intellectual property and operational methods from aggressive copycats. Gong answers frankly: if an idea demonstrates early commercial success, fast followers and clones are a 100% certainty. No legal barrier or secrecy can prevent competitors from attempting to replicate a working model feature-for-feature.

The only viable defense is operating velocity coupled with a proprietary data flywheel. A founder who has spent six months operating an application has accumulated invaluable "earned insight" through messy real-world customer friction, edge cases, and user feedback. While a copycat spends months struggling to understand why customers prefer the original product, the incumbent founder is already using proprietary usage data to refine service quality and entrench workflows, staying consistently out of reach of commoditized clones.

[21:43](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1303s) Data Flywheels and the Unreplicable Stack of Knowledge

Addressing how companies build defensibility in the current market, Grace Gong emphasizes that utilizing incoming proprietary data to continually improve services creates an essential "data flywheel"—a mechanism that stands as one of the few true moats available to founders. Drawing on insights from her interview with Square co-founder Jim McKelvey, she notes that lasting resilience stems from a cumulative "stack of knowledge." Rather than relying on a single breakthrough, an organization's defense is forged through diverse, hard-won operational experiences layered together in a manner that competitors simply cannot copy or replicate easily.

- **Data Flywheels:** Leveraging fresh, domain-specific user data to perpetually refine service quality remains one of the primary sustainable moats for modern software ventures. - **The Knowledge Stack:** Referencing Jim McKelvey's philosophy from building Square, enduring defensibility comes from a tightly integrated stack of varied problem-solving experiences rather than isolated technological advantages.

```mermaid flowchart LR accTitle: Data Flywheel and Knowledge Stack Defensibility subgraph Flywheel ["The Feedback Flywheel"] A["New Operational Data"] --> B["Enhanced Service Quality"] B --> C["Continuous Customer Usage"] C --> A end subgraph Defensibility ["Moat Construction"] C --> D["Layered Operational Experience"] D --> E["Non-Replicable Knowledge Stack"] end ```

[22:24](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1344s) Unpredictable Horizons, Market Bubbles, and Operator Resilience

When asked by an audience member to predict where the landscape will sit over the next two to three years, Grace candidly deflects long-range forecasting with humor. Directing her attention to an attendee, she remarks, "Oh next 2 to 3 years. I'll tell you next year," explaining that with the pace of disruption, "everything changes all the time" and attempting to map out multiple years in advance is a fool's errand.

Turning to a question from another attendee about whether the AI sector is currently trapped in an inflated valuation bubble, Grace likens the tech cycle to an ongoing roller coaster ride with perpetual peaks and troughs. In her view, the perennial debate over whether the industry is in a bubble is largely beside the point for operators and real businesses. A bubble fundamentally exists whenever capital is poured into ideas that never make economic sense, yet businesses that generate real revenue and create demonstrable customer value have nothing to fear—"if the customer continuously buying, you're safe." For everyday employees, customers, and operators, anxiety about macroeconomic bubbles is unhelpful; the only real protection is getting hands-on. By actively using and building with AI tools on a daily basis, operators gain the firsthand experience necessary to determine what genuinely streamlines their work and what is merely empty hype.

- **The Folly of Multi-Year AI Predictions:** Given the relentless pace of technological shifts, attempting to forecast two or three years out is impractical; true market realities are best reassessed year by year. - **Fundamentals Over Market Cycles:** Speculative fundraising will always fuel bubble dynamics, but genuine commercial viability is governed by revenue generation and customer retention. - **Practical Operator Defense:** Rather than obsessing over venture valuations or macro cycles, professionals should use AI tools daily to build firsthand discernment about which tools actually drive productivity.

## Notable quotes > [00:00](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=0s) The model wars will produce a few winners. The application wars will produce hundreds. — Grace Gong

> [02:05](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=125s) the unsexy industry is the new sexy industry. — Grace Gong

> [12:58](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=778s) if you work at a company, you're an operator! — Grace Gong

> [14:51](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=891s) you must get them to pay. — Grace Gong

> [16:14](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=974s) Hell yeah! I'm sorry, I don't know how to say this in a more calm way, but definitely. — Grace Gong

> [16:14](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=974s) When you have to output, you really have to input. — Grace Gong

> [18:35](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1115s) I have an answer that you probably don't want to hear. — Grace Gong

> [18:35](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1115s) Did all of you raise a lot of money? Is that why you're leaving? — Grace Gong

> [22:24](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1344s) Oh next 2 to 3 years. I'll tell you next year, — Grace Gong

> [22:24](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1344s) if the customer continuously buying, you're safe. — Grace Gong

## Mentioned Grace Gong, Smart Venture Media, Smart Venture Podcast, Google, Meta, Amazon, Snowflake, Nvidia, Sequoia Capital, Andreessen Horowitz (a16z), Lightspeed Venture Partners, Bessemer Venture Partners, NEA, Square, Loom, Cruise, Glean, Craigslist, Hinge Health, Transfix, Corgi, Crusoe, Bedrock Robotics, Waymo, Dexterity, Doss, General Catalyst, Serval, Canva, Figma, Moda, TensorMesh, Databricks, WhisperFlow, Sierra, Cresta, Forethought, Pylon, Hosty, Gradient Ventures, Wall Street Journal, Pozeia, Slack, OpenAI, Anthropic, Perplexity, Cluey, Claude, Gamma, Clay, Airtable, Travis Kalanick, Jim McKelvey

## Video description Grace Gong - VC & Podcaster, Smart Venture Media

The model wars will produce a few winners. The application wars will produce hundreds. A look at where capital is really moving right now: physical-world autonomy, agent-native software, and the trillion-dollar industries still running on faxes. Not one of the biggest checks this year went to a foundation model.

[Source](https://www.youtube.com/watch?v=i-YZnU0zMRE)

中文简报

Gong表示AI资本正放弃基础模型转向实体机器人风险投资正在绕开基础模型，因为具备防御壁垒的企业价值正沉淀至那些在数万亿美元级传统行业中将深厚行业认知与专有数据飞轮相结合的应用层创始人手中。风险投资人 Grace Gong 概述了引导人工智能领域资本流动的关键趋势，认为超额价值将沉淀在应用层而非基础模型之中。她详细分析了市场向硬件即服务（Physical AI）、物流与保险等庞大传统行业的数字化、Agent 原生软件技术栈以及企业级语音自动化方向的重大转变。Gong 指出，对抗上游 AI 模型的防御壁垒依赖于对垂直领域的深度理解、专有分发渠道以及运营数据飞轮，而非单纯依靠技术复杂性。最后，她在开放问答环节探讨了早期客户验证、硅谷背景偏见以及向基于成效的软件定价模式的转型。资本正大举流向“HaaS即新SaaS”（物理AI），将自主智能深度集成至重型施工机械和仓储机器人等实体工业场景与工作流中。 (1:08) 仍严重依赖纸质传真等遗留系统的传统行业（如13万亿美元的建筑业、9.5万亿美元的物流业和7.4万亿美元的保险业），正在催生万亿级AI原生基础设施新机遇。 (2:05) 软件设计与数据基础设施正加速从以人类为中心的图形界面，转向专为自主AI Agent设计与调度的“Agent原生”软件架构。 (3:45) 应用层抵御上游基础大模型侵蚀的核心壁垒，在于行业专业“品味”（Taste）、最后一公里工作流深度整合以及专有的用户分发引擎。 (9:56) 客户真实的付费意愿是检验初创企业需求的唯一硬指标，而依托持续运营构建的数据飞轮与多维经验沉淀则是抵御同质化抄袭的最强防线。 (14:51)

Reading summary

Grace Gong of Smart Venture Media argues that AI venture capital is shifting away from generic consumer tools toward physical automation, legacy industrial back-offices, and agent-native software infrastructure. In her view, application-layer defensibility relies not on raw technical execution, but on earned industry judgment, distribution channels, and compounding data flywheels. Physical AI and Massive Legacy Markets (1:08) Capital is moving heavily into physical automation—framed as 'hardware as a service' (HaaS)—and traditionally unglamorous legacy sectors. Startups like Bedrock Robotics (founded by Waymo alumni) and Dexterity ($1.6 billion valuation) integrate autonomous software into heavy construction machinery and warehouse operations. Rather than treating AI as browser-bound software, these ventures embed machine intelligence directly into mission-critical physical workflows. Simultaneously, investors are pursuing enormous legacy addressable markets: global construction ($13 trillion), logistics ($9.5 trillion), and insurance ($7.4 trillion). While legacy sectors like insurance still operate on physical fax machines and antiquated infrastructure, new entrants are building full-stack replacements. Examples include Corgi raising over $100 million for commercial insurance infrastructure, Doss securing more than $70 million led by General Catalyst for AI-native ERP, and Serval raising two consecutive rounds from Sequoia Capital for IT ticketing automation. Agent-Centric Software and Voice Automation (3:45) Software architecture is pivoting from human-facing graphic user interfaces to systems built specifically for autonomous AI agents. Gong notes that while legacy platforms like Canva, Figma, Snowflake, and Databricks were engineered for human workflows and human data engineers, emerging platforms cater directly to background agents. Startups like Moda are building design engines for agent-to-agent asset generation, while TensorMesh applies key-value (KV) caching to optimize contextual data retrieval for autonomous software. In communication, enterprise voice AI has progressed beyond simple transcription tools into mission-critical workflow automation. Companies such as Sierra, Cresta, Forethought, and Pylon are capturing large rounds to automate enterprise customer service and eliminate call-center wait times. In the restaurant sector, Hosty (backed by Gong and Gradient Ventures) demonstrates the direct revenue capture of voice agents by immediately answering, qualifying, and forwarding high-value group reservations that busy staff would otherwise drop. Application Layer Moats: Taste, Distribution, Flywheels (9:56) With foundational models from OpenAI and Anthropic accessible to everyone and modern coding tools lowering technical barriers, code alone provides little defensibility. Founders face constant risk of being 'Claude-shotted'—rendered obsolete by upstream model updates. Gong argues defensibility requires industry 'taste' (deep domain expertise on how real-world workflows operate) and proprietary distribution, such as Cluey marketing B2B software through episodic workplace mockumentaries. Against copycats, Gong maintains that speed and proprietary data flywheels are the only true protections. Referencing Square co-founder Jim McKelvey's concept of a 'stack of knowledge,' she contends that accumulating months of messy, real-world customer edge cases creates an operating foundation that competitors cannot easily replicate. Fresh, domain-specific user data continuously refines product quality, keeping the incumbent ahead of commodity clones. Operator Leverage and Outcome-Based Business Models (7:51) Software commercial models are shifting from per-seat SaaS subscriptions (such as $12.99 monthly seats) to outcome-based pricing, where vendors package autonomous agents as AI employees delivering specialized outputs for $10,000 per month. In response to speculation regarding 'one-person billion-dollar companies,' Gong argues the enduring future of work remains multiplayer, with AI agents operating as collaborative teammates rather than isolating solo founders. For non-founding corporate operators, internal workflow automation paired with public thought leadership provides career leverage. By testing tools like Claude, Gamma, and Clay to fix internal operational bottlenecks, operators build an internal data flywheel. When validating new independent ventures, Gong advises bypassing friends and family to secure paying customers directly, noting that unpaid pilot metrics offer no reliable signal of genuine demand. Gong dismisses long-range multi-year forecasting and valuation bubble debates as unhelpful for day-to-day operators. She argues that while speculative capital fluctuates and venture funds often favor founder pedigree over raw merit, companies that deliver measurable customer value and generate recurring revenue remain insulated from macroeconomic tech cycles.

精炼阅读摘要

Smart Venture Media 创始人 Grace Gong 指出，当前硅谷人工智能资本正从表层消费端噱头，系统性地转向重工业硬件化（HaaS）、万亿美元级传统落后行业改造以及为自主智能体原生构建的基础设施。她认为大模型之上的核心商业壁垒并非纯代码实现，而是深植于产业实际场景的判断力、分发能力以及由付费客户驱动的专有数据飞轮。硬件即服务与万亿级传统垂直领域 (1:08) Gong 强调，“硬件即服务”（HaaS）正在成为实体人工智能（Physical AI）的关键商业载体，风险投资正加速涌入深度融合感知与自主决策系统的重型设备与工业机器人，而非局限于浏览器视窗内的数字模型。例如，由 Waymo 前员工创立的 Bedrock Robotics 致力于为推土机和挖掘机等重型工程机械提供自动驾驶方案；估值达 16 亿美元的 Dexterity 则制造用于仓储物流和制造流程的自主机器人系统，将一次性硬件销售转化为持续运行的自动化工作流。与此同时，资本正流向被长期忽视但总体潜在市场庞大的传统行业，包括规模达 13 万亿美元的全球建筑业、9.5 万亿美元的物流业和 7.4 万亿美元的保险业。Gong 指出，许多管理巨额资产的保险机构仍依赖传真机等过时设施，催生了如筹资逾 1 亿美元、从底层重构商用保险栈的 Corgi。此外，由 General Catalyst 领投、筹集超 7000 万美元构建 AI 原生 ERP 的 Doss，以及获得红杉资本连续投资以实现 IT 工单自动化的 Serval，均表明替代传统繁琐纸质与人工流程是目前利润最丰厚的投资切入点。智能体原生软件与多端协同范式 (3:45) 在软件架构层面，Gong 提出计算范式正由“面向人类交互界面”转向“专为自主软件智能体设计”。Canva 和 Figma 等传统工具围绕人类图形界面构建，而新一代基础设施则直接服务于无需实时人工干预的多步骤智能体。例如，设计引擎 Moda 专为智能体资产交互与生成而设；TensorMesh 则利用键值（KV）缓存优化数据存储与上下文调取，替代了以往如 Snowflake 和 Databricks 专供人类数据工程师查询的数据架构。在交互端，语音技术已超越简单听写，转向处理高价值企业任务，如 Sierra、Cresta、Forethought 和 Pylon 等获巨额估值的客服平台。Gong 以自身参与投资的餐饮语音平台 Hosty 为例，展示了智能体在毕业季高峰期直接接管 70 人订餐需求并通知店主锁定收入的实效。针对业内关于“单人十亿美元公司”的讨论，Gong 结合与 Slack 管理层的交流反驳了孤立化创业趋势，主张结合模型上下文协议（MCP），智能体应作为“多端协作”中的虚拟团队成员，在保留人类团队凝聚力的同时放大组织整体生产力。应用层竞争优势与成果导向定价 (9:56) 面对基础模型（如 OpenAI、Anthropic）迭代可能带来的“被一枪毙命”（Claude-shotted）隐忧，Gong 认为由于没有单一基础模型能够垄断所有垂直工作流，绝大部分企业价值依然沉淀在应用层。在代码辅助工具使基础开发高度普及的背景下，单纯的技术实现已无法构成持久护城河，真正的防御壁垒转移至两项要素：一是来自医疗、法律等专业实践的“行业品味”（Taste），即对业务实际痛点的深刻判断；二是自建分发能力，例如 B2B 初创公司 Cluey 通过制作情景喜剧短片实现病毒式传播。这种应用层演进同步引发了软件商业模式的根本重塑。Gong 指出，企业软件采购正从传统按人头订阅（如每月每座 12.99 美元）迅速转向以业务成果为准的“AI 员工”交付模式，例如每月按设计成果打包收取 1 万美元。在这种转变下，业务操作人员（Operators）通过公开分享实操洞察构建个人分发渠道，能倒逼自身紧跟工具前沿，确立不可替代的职场议价权。数据飞轮、真实付费与资本偏见 (20:38) 在应对产品被快速克隆与知识产权竞争方面，Gong 援引与 Square 联合创始人 Jim McKelvey 的交流指出，持久的防御力来自高度集成的“知识栈”（Stack of Knowledge）与由真实业务构筑的“数据飞轮”。一旦软件显现市场需求，追随者必然涌现；唯有在半年乃至更长实际运营中累积的真实摩擦、边缘用例和持续迭代，才能利用专有使用数据拉开与纯仿制品的体验差距。针对早期验证与融资，Gong 直言免费使用指标在投资尽调中缺乏说服力，唯有让非亲友的真实客户实际付费，才能检验解决方案是否切中核心痛点。同时，她揭示了风投行业对创始人名校背景与既往履历的强烈偏见——投资知名重复创业者（如 Uber 的 Travis Kalanick）往往是合伙人规避机构信誉风险的捷径。Gong 告诫称，高额种子轮融资常演变为脱离产品市场匹配（PMF）的虚荣指标，资金规模与初创企业最终能否成功没有任何必然关联。针对市场是否存在 AI 估值泡沫及未来两至三年的技术走向，Gong 认为由于行业变革极为剧烈，进行长期预测并不现实。投机资本固然会周期性催生估值泡沫，但只要企业能持续获取真实付费客户并创造商业价值，就能跨越市场周期波动。对于一线从业者与企业团队而言，化解宏观焦虑的最佳途径是坚持日常上手使用最新 AI 工具，在具体业务实践中辨别技术实质与营销炒作。

Detailed notes — zh

## Background 该内容由频道 Ai4 于 2026 年 8 月 19 日发布在 YouTube 平台上。页面简介指出，本期嘉宾为 Smart Venture Media 创始人兼投资人 Grace Gong，核心探讨基础模型竞争格局赢家寥寥的背景下，风投资金正全面涌向物理世界具身智能、Agent 原生软件以及仍在使用传真机的万亿美元级传统产业，且今年最巨额的支票没有一张流向底层基础模型。 ## TL;DR Grace Gong 结合其与硅谷顶级 VC、大厂高管及明星创业者的高频交流，指出资本流动已从底层大模型全面转向应用层与物理 AI（HaaS 成为新 SaaS）。建筑、物流以及依然依赖传真机的保险等体量巨大的传统行业，正迎来 AI 原生架构的颠覆重构，同时软件终端用户正演化为 AI Agent，驱动语音商业化落地并重塑协同工作方式。面对基础模型能力更新带来的“Claude-shotted”生存危机，创业者的核心壁垒在于行业专有认知沉淀出的“品味”（Taste）、自建内容分发渠道以及难以复制的数据飞轮。在软件商业化从按席位收费转向按结果交付的浪潮中，无论是初创团队还是大厂从业者，唯有打破亲友圈验证真需求并保持日日实操的“构建者心智”，才能切实穿越技术泡沫。 ## Atmosphere & style 演讲与互动整体呈现出硅谷创投圈极具一线体感的高密度、快节奏与务实风格。Grace Gong 语言风格犀利直率又极具亲和力，不仅坦率调侃自己的天使投资工作是“高强度闲聊”，更用生动的真实商业故事将晦涩的风投逻辑讲得通俗透彻。现场互动氛围热烈且极具感染力，从邀请全场业务执行者（Operators）起立分类摸底，到面对观众就抄袭焦虑、明星背景光环与 AI 泡沫提问时给出不加粉饰的诚恳解答，整场分享褪去了宏观叙事的虚浮，充满了一线实战者的敏锐与清醒。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=4s) 硅谷资本风向发生重大转移，巨额资金不再盲目下注基础大模型，而是全面涌向能够解决垂直场景问题的应用层和硬件生态。 - [01:08](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=68s) “硬件即服务（HaaS）”与物理 AI（Physical AI）正在爆发，软件突破屏幕限制，与推土机、装配机器人等重型硬件结合以改造实体工业。 - [02:05](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=125s) 建筑、物流、保险等体量达数万亿美元但依赖传真机的传统行业，正成为 AI 原生架构底层重构的最具价值风口。 - [03:45](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=225s) 软件终端使用者正从人类转向自主运行的 AI Agent，驱动了专为代理设计的物料工具及高性能 KV 缓存等底层数据设施的诞生。 - [05:10](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=310s) 语音 AI 跨越娱乐阶段迈入高价值商业闭环，企业级语音 Agent 展现出全天候秒级并发处理和直接促成大额订单转化的变现能力。 - [07:51](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=471s) 工作模式正从“一个人估值十亿”的孤独单兵叙事，演变为“人类员工 + 专属 Agent 军团”并肩作战的多人协同协作体系。 - [09:56](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=596s) 在基础模型能力不断扩张的阴影下，垂直应用层初创企业的核心护城河取决于长期行业摸爬滚打得出的“品味”与独特“分发能力”。 - [12:58](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=778s) 企业业务执行者应当主动使用 AI 工具自建内部工作流，基于高频内部实践形成独特数据闭环，筑牢个人与团队的竞争壁垒。 - [15:06](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=906s) 产品早期验证必须尽早走出亲友圈，唯有让非亲非友的真实客户自掏腰包付费，才是检验刚需与进行投资尽调的硬标准。 - [16:14](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=974s) 软件商业模式正全面从“按人头席位收费”跃迁至高客单价的“按交付结果付费”，职场人必须具备自建自动化流程的 Builder 基因。 - [18:35](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1115s) 资本对明星创始人的偏爱降低了决策风险，但巨额概念融资不等于商业模式跑通，产品体验与实际落地才是根本。 - [20:38](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1238s) 初创团队抵御竞品与巨头像素级抄袭的唯一出路，在于以速度换取认知差，通过真实用户沉淀出不可复制的专有数据飞轮与知识壁垒。 - [23:07](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1387s) 穿越科技周期与宏观泡沫的核心在于持续创造客户价值并获得稳定现金流，从业者应通过每日亲自实操构建来评估技术的真实价值。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=4s) 硅谷资金的新风向：告别基础模型，涌向应用层 Smart Venture Media 创始人 Grace Gong 每天都在 LinkedIn 上进行直播，与三类核心人群深度交流：顶尖投资人、大厂高管以及 AI 领域的明星创业者。她的节目曾邀请过来自 Google、Meta、Amazon、Snowflake、Nvidia 的高管，红杉资本（Sequoia）、a16z、Lightspeed、Bessemer、NEA 等顶级风投的合伙人，以及 Square、Loom、Cruise、Glean、Craigslist 等知名公司的创始人。作为一名活跃在硅谷的早期天使投资人，Grace 将自己的工作戏称为“整天都在高强度闲聊（yapping all day）”，但也正是在这种高频的信息碰撞中，她捕捉到了资本真正的流动趋势：今年最巨额的支票没有一张流向底层基础模型，而是全面涌向了能够解决实际场景问题的应用与硬件生态。 - **嘉宾生态矩阵**：涵盖硅谷一线 VC 合伙人、科技巨头 C-Level/VP，以及 Hinge Health、Transfix、Crusoe 等 AI 独角兽企业创始人。 - **核心论点**：基础模型之战只会诞生极少数赢家，但应用层与物理世界的结合将孕育数百家颠覆性企业。 --- ### [01:08](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=68s) 趋势一：HaaS 是新的 SaaS，物理 AI 正在爆发风投领域显现出的第一个重磅趋势是“硬件即服务（HaaS is the new SaaS）”，也就是物理 AI（Physical AI）和具身智能的崛起。软件不再局限于屏幕之内，而是与实体机械硬件深度融合，直接改造重工业与生产一线。由前 Waymo 自动驾驶核心团队创办的 Bedrock Robotics，正在将前沿 AI 技术注入推土机、挖掘机等重型工程机械，打造建筑工程领域的“自主施工版 Waymo”。另一家代表性企业 Dexterity 则致力于研发工业级智能协作机器人，帮助工厂自动化完成复杂的装配与分拣作业，其最新估值已飙升至 16 亿美元。资本正加速下注那些能够直接在物理空间干活的硬件载体。 --- ### [02:05](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=125s) 趋势二：枯燥的传统产业，蕴藏着万亿美元级大蛋糕 Grace 提到，在许多非公开场合，投资人喜欢将这一趋势称作“不性感（unsexy）的行业正在变成最性感的风口”。放眼全球，许多体量惊人却极度缺乏技术更新的传统赛道，正等待 AI 进行底层重构：建筑行业规模高达 13 万亿美元，物流运输达 9.5 万亿美元，而规模达 7.4 万亿美元的保险行业，至今大量业务流程仍依赖传真机运行。这些看似沉闷的领域正孕育出惊人的估值空间。AI 原生保险公司 Corgi 已经完成了超过 1 亿美元的融资，它不是做宠物业务，而是利用 AI 从底层彻底重构承保、理赔与风险评估的完整技术栈；AI 驱动的新型 ERP 系统初创公司 Dos，拿下了来自 General Catalyst 的 7000 万美元巨额投资；而入选 Smart AI 100 榜单的 Servo，凭借用 AI 颠覆企业 IT 服务工单处理流程，在短时间内连续完成两轮由红杉资本领投的高额融资。 ```mermaid flowchart TD accTitle: 传统万亿美元产业的 AI 原生重塑 subgraph Traditional["万亿美元级传统基底"] A["建筑工程行业 $13 万亿规模"] B["物流运输行业 $9.5 万亿规模"] C["传统保险行业 $7.4 万亿规模（仍依赖传真机）"] end subgraph AINative["AI 原生重构层"] D["Bedrock Robotics 自主工程机械"] E["Dos AI 原生 ERP 系统"] F["Corgi 全栈 AI 保险"] G["Servo 自动化 IT 工单"] end A --> D B --> E C --> F C --> G ``` - **市场体量数据**：全球建筑市场约 13 万亿美元、物流供应链约 9.5 万亿美元、保险产业约 7.4 万亿美元。 - **重构逻辑**：不再是为旧流程打补丁，而是从第一天起以 AI 原生架构替代人工工单、纸质凭证和古老的传真通讯。 --- ### [03:45](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=225s) 趋势三：用户角色转移，从“人类用户”到“Agent 代理用户” 以往无论是 Canva 还是 Figma，产品设计的核心交互逻辑都是服务于“人类的眼球与手指”。然而，第三个关键转变正在发生：软件的终端使用者正逐渐变成自主运行的 AI Agent。在这一范式下，诞生了一批完全为 AI 代理设计的工具。例如初创公司 Moda 专门为 AI Agent 提供设计工具，“这听起来非常 Meta，但后台的 Agent 确实承担了越来越多的设计与物料产出”。在底层数据栈领域，TensorMesh 利用 KV 缓存（KV Cache）技术，专门为 Agent 快速检索与存储上下文数据提供支持。过去的 Snowflake 或 Databricks 是帮助人类数据工程师处理与存储海量数据，而未来的新型基础设施则完全是为了满足 Agent 高并发、长上下文记忆和自主检索的数据吞吐需求。 --- ### [05:10](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=310s) 趋势四：语音 AI 的黄金时代与真实商业闭环在硅谷，“语音正在迎来属于它的高光时刻（Voice is the new moment）”。从个人端像 WhisperFlow 这样只需讲话就能理顺思路并自动整理成文的思考伴侣，到企业端拿下巨额融资的 Sierra、Krista AI、Forethought 和 Pylon，语音交互已迅速越过娱乐玩具阶段，直击商业客服痛点。过去消费者为了办理退货、预约服务或排查家庭 Wi-Fi 故障，往往需要在客服热线中苦等数小时。现在的企业级语音 Agent 能够瞬间并发处理海量复杂工单。Grace 分享了她作为天使投资人参与的餐饮语音服务项目 Hosty（该项目亦获得了由 Google 旗下 AI 基金 Gradient Ventures 的投资）。在毕业季的高峰期，曾有一位客户急需预订一桌 70 人的超大聚会，连续致电多家餐厅均无人接听或被前台以“需要请示经理”为由推脱；唯独接入了 Hosty 的餐厅，AI Agent 在通话瞬间记录下全部订餐细节并即刻推送到老板手机，店方秒级确认，不仅稳稳接下了这张原本绝不可能拿下的 70 人大单，也展现了语音 AI 在真实商业场景中的变现杀伤力。 - **传统客服痛点**：多分支 IVR 菜单复杂，呼叫中心排队时长动辄 1 至 2 小时。 - **现代语音 Agent**：实现全天候秒级并发响应，打通订单管理、自动记录备忘与后台业务系统。 --- ### [07:51](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=471s) 趋势五：从超级单兵到“人机协作多人游戏” 此前，一对兄弟借助 AI Agent 搭建远程营销系统推广司美格鲁肽（Ozempic），迅速将业务做成 10 亿美元估值公司的案例引发全网轰动，《华尔街日报》等顶级媒体的大篇幅报道让“一个人估值十亿（One-person billion-dollar company）”的叙事深入人心，也催生了诸如 Pozeia 这样专门赋能超级个体的自动化工具。然而，Grace 在参加与 Slack 首席营销官（CMO）的晚宴时看到了另一层未来走向：Slack 推出了模型上下文协议（MCP）等协同功能，使团队成员不仅能彼此沟通，还能无缝接入各自的专属 Agent 以及团队公共 Agent。人类每天的时间与精力极其有限，与拥有共同愿景、充满激情的伙伴并肩作战是保持工作动力的核心。“AI Agent 不会消灭你的同事，而是成为了每名员工的新型超级副驾”，工作模式正从孤独的单兵作战演化为“人类团队 + Agent 军团”的多人协同协作。 --- ### [09:56](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=596s) 应用层淘金：Claude-shotted 阴影下的品味与分发底层大模型（ChatGPT、Claude、Perplexity）已经搭好了基础设施底座，但真正的商业利润依然沉淀在应用层。虽然硅谷近期盛行“Claude-shotted”一词——意指 Anthropic 的 Claude 每更新一次功能，就会随手消灭一大批轻量级套壳创业公司——但 Grace 坚信，没有任何单一的基础模型能够通吃所有垂直行业的全流程。决定应用层创业者能否胜出的关键是“品味（Taste）”与“分发能力（Distribution）”。品味源自创业者在特定行业长期摸爬滚打得出的“挣来的洞察（Earned Insights）”，它决定了你是否真正清楚某一垂类工序中哪种痛点存在付费意愿，哪种方案无法落地。在人人皆可借助 AI 快速写出软件的时代，产品不再稀缺，分发渠道才是真正的护城河。例如硅谷公司 Cluey 的创始人通过自编自演恶搞美剧《办公室（The Office）》的短剧来推广 SaaS，这种极具传播力的新型内容获客方式证明：哪怕每天只是在社交网络分享一条深度的行业洞见或真实的创业心路，也能在同质化竞争中脱颖而出。 ```mermaid flowchart LR accTitle: 应用层初创企业的防御壁垒构建 subgraph CoreInput["创始人核心资产"] A["行业深度经验 Earned Insights"] B["内容与分发渠道 Building in Public"] end subgraph ProductLoop["产品闭环构建"] C["垂直场景品味 (Taste) 精准锁定特定工作流"] D["自有分发通路 获取高纯度测试客户"] end subgraph Moat["竞争壁垒"] E["高粘性垂直应用 抗衡平台模型通杀风险"] end A --> C B --> D C --> E D --> E ``` --- ### [12:58](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=778s) 现场互动与操作者启示：如何用 AI 构筑企业内部护城河为了摸清现场听众画像，Grace 邀请全场起立并逐一分类坐下：创始人首先就坐，随后是投资人，当轮到“业务执行者（Operators）”时，全场绝大多数在企业内部任职的听众才恍然大悟纷纷坐下。针对庞大的企业内部员工与业务人员群体，Grace 给出了极具实操性的建议：不要等公司安排，员工完全可以利用 Claude Co-work、Gamma、Clay 等新一代工具，为自己当下的岗位工作流搭建专属工具链。当你和团队成为这些内部自建产品的首批高频用户时，业务飞轮就已经开始转动。基于真实反馈以极高的速度迭代功能，这种源自内部实践的数据飞轮与流程沉淀，不仅构成了你在企业内部无法被替代的竞争壁垒，也是未来有朝一日独立融资创业的坚实土壤。 --- ### [15:06](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=906s) 问答环节：跨越冷启动，向“非亲非友”收钱验证真需求面对现场观众关于“完成种子轮研发与亲友借款后，如何进一步向市场融资与验证产品”的提问，Grace 明确指出，真正的商业验证必须从“走出熟人圈子，面向非亲非友（not family and not friends）的陌生客群”开始。创业者应当在细分圈层中锁定早期设计伙伴（Design Partners）。例如打造一款面向设计师的 AI Figma，就应当深入本地设计沙龙或垂直社群，直接与一线从业者对话，并在收集反馈后直接提出付费要求。投资人在进行尽职调查时，核心看的不是注册量，而是“到底有没有人愿意为你的产品自掏腰包”。免费用户给出的赞美极易产生误导，唯有真金白银的付费转化，才能向资本市场证明你的产品真正解决了不可替代的痛点。 - **种子期验证法**：尽早脱离亲友圈测试，寻找垂直领域的真实从业者作为种子设计伙伴。 - **尽调核心指标**：付费用户转化率高于一切虚荣指标，付费动作是唯一的硬性需求试金石。 --- ### [16:14](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=974s) 问答环节：从席位制到结果付费，大厂员工必须具备 Builder 基因现场观众追问，强调内容分发与个人 IP 是否同样适用于大厂内部的职场人。Grace 斩钉截铁地回答：“太重要了（Hell yeah）！”当前的职场逻辑正遭遇 AI 的系统性重构，软件计费模式已从过去的单人席位费（Seat-based pricing，如每月 12.99 美元的工具订阅）全面转向基于结果交付的服务型计费（Outcome-based pricing），企业宁愿每月支付 1 万美元采购一个能够稳定交付设计方案的“AI 数字员工”。在硅谷文化中，最鲜明的特质是“人人都是构建者（Builder Mindset）”。无论是市场营销专员还是销售人员，大家都在业余时间使用 AI 工具为自己搭建自动化 CRM 和 Airtable 流程，以此保持职场敏锐度。在公开渠道打造个人影响力（Building in public）不仅能建立外部护城河，更会倒逼自己保持高强度的输入与前沿工具采纳，从而在技术替代的浪潮中始终快人一步。 - **商业计费范式转换**：从低客单价的“工具按头收费”向高溢价的“工作结果/AI 雇员收费”跃迁。 - **构建者心智（Builder Mindset）**：非技术岗利用无代码与 AI 工具自行搭建业务流，以持续输出倒逼知识迭代。 --- ### [18:35](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1115s) 问答环节：资本偏爱明星履历？巨额融资不等于产品成立针对现场女性观众关于“为何部分创始团队仅凭概念便能斩获巨额投资”的困惑，Grace 直言不讳地揭示了硅谷创投圈的潜在潜规则：投资人本质上是在为顶尖人才画像下注，所谓背景背书通常来源于此前大厂的光环或成功的创业经历。对于 VC 而言，投资像 Travis Kalanick 这样成功打造过 Uber 的连续创业者，其尽调与决策风险远低于赌一个从未操盘过规模化业务的新人。然而，Grace 警示道，闪电般拿下一大笔资金固然能带来虚荣光环，但绝不代表商业模式或产品本身获得了验证；Uber 之所以伟大，是因为最初几乎没人看好，团队最终凭借极致的产品体验、地面分发推进以及坚韧的组织能力杀出重围。盲目迷信融资额并不能保证创业成功。 --- ### [20:38](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1238s) 问答环节：如何抵御抄袭？用真实业务飞轮跑赢时间差现场最后一位观众提出了创业者普遍面临的防御焦虑：“当巨头或竞品完全照抄你的产品模式时，初创团队该如何自保？”Grace 坦言，在技术透明的今天，没有任何手段可以阻止别人追随你的模式，一旦你的业务被证明有利可图，百分之百会引来像素级抄袭者。初创企业的唯一生路在于“以速度换取认知差，建立数据飞轮”。比对手早起步六个月，意味着你已经在实战中踩平了无数个认知陷阱，深刻理解客户为什么掏钱以及为什么放弃竞品；而抄袭者即便复制了界面，也仍需耗费数月甚至更长时间去经历这些摸索。此时先行者已经从真实客户身上积累了专有业务数据，并利用这些数据持续打磨出体验更优的服务，这种由实战沉淀出的飞轮壁垒，是任何外部追赶者仅靠复制代码所无法逾越的。 ### [21:43](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1303s) 数据飞轮与难以复制的知识壁垒在探讨企业如何构建长期护城河时，核心在于能否利用新产生的数据持续反哺并优化服务，从而形成闭环的“数据飞轮”（Data Flywheel）。演讲者提到，这几乎是初创公司建立差异化优势的必由之路。此外，她还分享了之前采访 Square 联合创始人吉姆·麦克凯尔维（Jim McKelvey）时的洞察：真正的壁垒往往来自于一套基于独特经验不断累积演进的“知识堆栈”（Stack of Knowledge / Innovation Stack）。这种多层经验与认知的复合体极其复杂，外界很难对其进行简单照搬或复制。 - **数据飞轮效应**：通过持续获取并沉淀用户新数据来迭代升级服务，驱动业务自我强化。 - **复合知识壁垒**：援引 Square 创始人 Jim McKelvey 的观点，跨维度经验沉淀出的知识体系构成了难以被竞争对手复制的防御壁垒。 --- ### [22:24](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1344s) 瞬息万变的行业预测与现场问答在进入现场问答环节后，有观众询问未来两到三年的行业走向与技术发展趋势。面对这类长远预测，演讲者幽默且直率地回应道：“未来两到三年会怎样？我明年再告诉你吧。”她坦言，在当前这个技术与资本瞬息万变的环境下，任何长期预判都充满不确定性，“因为谁也说不准，所有事情随时都在发生变化”。这种坦诚不仅引来现场的笑声与共鸣，也折射出当下 AI 行业演进速度之快，以至于短周期的快速响应远比死板的长远预测更为实际。 --- ### [23:07](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1387s) 穿越 AI 泡沫：聚焦实际商业价值与日常实践针对现场关于“当前是否存在 AI 泡沫”的提问，演讲者将科技周期比作过山车，认为行情总是有起有落。在她看来，过分纠结于当下究竟是不是泡沫并没有实际意义。只要一家企业能够真正创造客户价值并持续产生实际收入，“只要客户一直在买单，你就是安全的”。从某种层面上讲，创投圈几乎始终处于某种泡沫之中——每当市场上充斥着为商业逻辑无法自洽的项目大笔融资时，泡沫便客观存在。然而，对于普通的职场从业者或日常业务操盘手（Operator）而言，宏观层面的“泡沫论”无关痛痒。能够真正保护个人竞争力的唯一方式，就是“每天亲自使用 AI 工具并在工作中不断实践构建”。通过积累第一手的实操经验，每个人都能自主判断这些技术究竟是对自己的实际工作切实有效，还是仅仅停留在概念炒作阶段。 - **泡沫与商业本质**：穿越周期波动的根本标准在于企业能否创造实际价值、获取稳定现金流与客户持续复购。 - **从业者应对策略**：与其旁观宏观叙事争论，不如每日深度使用 AI 工具并亲手构建应用，依靠第一手体感评估技术实用性。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=4s) 整天都在高强度闲聊（yapping all day） — Grace Gong > [01:08](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=68s) 硬件即服务（HaaS is the new SaaS） — Grace Gong > [02:05](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=125s) 不性感（unsexy）的行业正在变成最性感的风口 — Grace Gong > [05:10](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=310s) 语音正在迎来属于它的高光时刻（Voice is the new moment） — Grace Gong > [07:51](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=471s) 一个人估值十亿（One-person billion-dollar company） — Grace Gong > [07:51](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=471s) AI Agent 不会消灭你的同事，而是成为了每名员工的新型超级副驾 — Grace Gong > [15:06](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=906s) 走出熟人圈子，面向非亲非友（not family and not friends）的陌生客群 — Grace Gong > [16:14](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=974s) 太重要了（Hell yeah）！ — Grace Gong > [16:14](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=974s) 人人都是构建者（Builder Mindset） — Grace Gong > [22:24](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1344s) 未来两到三年会怎样？我明年再告诉你吧。 — Grace Gong > [23:07](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1387s) 只要客户一直在买单，你就是安全的 — Grace Gong > [23:07](https://www.youtube.com/watch?v=i-YZnU0zMRE&t=1387s) 每天亲自使用 AI 工具并在工作中不断实践构建 — Grace Gong ## Mentioned Grace Gong, Smart Venture Media, Google, Meta, Amazon, Snowflake, Nvidia, 红杉资本 (Sequoia), a16z, Lightspeed, Bessemer, NEA, Square, Loom, Cruise, Glean, Craigslist, Hinge Health, Transfix, Crusoe, Waymo, Bedrock Robotics, Dexterity, Corgi, Dos, General Catalyst, Servo, Smart AI 100, Canva, Figma, Moda, TensorMesh, Databricks, WhisperFlow, Sierra, Krista AI, Forethought, Pylon, Hosty, Gradient Ventures, 司美格鲁肽 (Ozempic), Pozeia, Slack, Claude, ChatGPT, Perplexity, Anthropic, Cluey, Gamma, Clay, Airtable, Travis Kalanick, Uber, Jim McKelvey ## Video description Grace Gong - VC & Podcaster, Smart Venture Media The model wars will produce a few winners. The application wars will produce hundreds. A look at where capital is really moving right now: physical-world autonomy, agent-native software, and the trillion-dollar industries still running on faxes. Not one of the biggest checks this year went to a foundation model. [Source](https://www.youtube.com/watch?v=i-YZnU0zMRE)

Source quotes

[0:04] Grace Gong: 整天都在高强度闲聊（yapping all day） [1:08] Grace Gong: 硬件即服务（HaaS is the new SaaS） [2:05] Grace Gong: 不性感（unsexy）的行业正在变成最性感的风口 [5:10] Grace Gong: 语音正在迎来属于它的高光时刻（Voice is the new moment） [7:51] Grace Gong: 一个人估值十亿（One-person billion-dollar company） [7:51] Grace Gong: AI Agent 不会消灭你的同事，而是成为了每名员工的新型超级副驾 [15:06] Grace Gong: 走出熟人圈子，面向非亲非友（not family and not friends）的陌生客群 [16:14] Grace Gong: 太重要了（Hell yeah）！ [16:14] Grace Gong: 人人都是构建者（Builder Mindset） [22:24] Grace Gong: 未来两到三年会怎样？我明年再告诉你吧。 [23:07] Grace Gong: 只要客户一直在买单，你就是安全的 [23:07] Grace Gong: 每天亲自使用 AI 工具并在工作中不断实践构建

Source quotes

[0:00] Grace Gong: The model wars will produce a few winners. The application wars will produce hundreds. [2:05] Grace Gong: the unsexy industry is the new sexy industry. [12:58] Grace Gong: if you work at a company, you're an operator! [14:51] Grace Gong: you must get them to pay. [16:14] Grace Gong: Hell yeah! I'm sorry, I don't know how to say this in a more calm way, but definitely. [16:14] Grace Gong: When you have to output, you really have to input. [18:35] Grace Gong: I have an answer that you probably don't want to hear. [18:35] Grace Gong: Did all of you raise a lot of money? Is that why you're leaving? [22:24] Grace Gong: Oh next 2 to 3 years. I'll tell you next year, [22:24] Grace Gong: if the customer continuously buying, you're safe.

Original transcript

[0:04] My name is Grace Gan. I'm the founder of Smart Venture Media. I host a popular VC podcast called [0:09] Smart Venture podcast. And every day I go on live on LinkedIn. I just literally stream live with [0:16] three type of peoples. It's the top investors, C-suite from C-suite or VPs [0:21] from top tech companies and top founders in AI. So luckily we have featured some of the [0:30] you know C-level VPs of Google, Meta, Amazon, Snowflake, Nvidia, a lot of big tech companies, [0:35] founder from Sequoia, a16z, Lightspeed, Bessemer, NEA, all the top VC firms and founder of Square, [0:41] Loom, Cruise, Glean, Craigslist, a lot many things that people use every day as well as some of the [0:46] AI unicorns like founder of Hinge Health, founder of Transfix, Corgi, Crusoe among many others. [0:53] And today my topic is really based off my personal life experience of yapping all [0:58] day and as well as an angel investor myself in Silicon Valley. It's about where the money flows. [1:08] The first trend I'm seeing in the investing space is called HaaS is the new SaaS. [1:14] HaaS stands for hardware as a service. Essentially it's my take on physical AI. [1:20] At our dinner, at our conference and at our podcast we have a company called Bedrock Robotics. [1:27] It is created by people out of Waymo and they are essentially the AI version or Waymo version [1:34] of the Buddos or an activators. People are really leveraging AI to leveraging AI to build physical [1:40] robotics and really use it to do whether it's construction or another company called [1:48] Dexterity. It is a company that makes these AI industrial robots that really helps uh different [1:56] indus- industrial productions to go smoother. Uh that company is also valued at $1.6 billion. A [2:05] A second trend that [clears throat] I'm seeing is called boring industries, big outcomes. At [2:12] other conferences, I may have a different name for them, but I don't think it's appropriate to say [2:17] these words in AI 4. I call it the unsexy industry is a new sexy industry. [2:23] Um this including I'm going to throw in some statistics for you guys to [2:28] understand. So, construction industry is $13 trillion. Logistic industry is $9.5 trillion. [2:35] And insurance is $7.4 trillion. And many of the insurance company are still running on fax. So, [2:44] at in Silicon Valley, there's a company called Corgi. They are they raised over $100 million. No, [2:49] it's not for dogs or pets, but they are AI native insurance companies. People are building [2:55] AI from the ground up from different So, from the ground covering every single stack of insurance [3:02] themselves. And we have a company called Dos raised over $70 million from General Catalyst [3:08] that's building AI version of the ERP system. And obviously today, if you're using the internet, [3:14] you saw a company called Servo. Servo was on our smart AI 100 list last year. And they [3:21] since they [snorts] uh get on our list, they raised two rounds of fundings from [3:27] prestigious funds like Sequoia to really take over the IT service ticket industry. And [3:34] you're using the internet today. So, sexy industry is actually the unsexy industry nowadays. [3:45] The third trend I'm seeing is the user as an agent. What does means is, you know, I I'm sure [3:53] how many people here have used Canva? Okay, almost everyone. Okay, amazing. Uh what about Figma? [4:01] Okay, amazing. So, um so many people show their hands. Um [4:07] Why? It's because these companies are actually made for human. [4:11] Uh our podcast, there is a company called Moda. So, essentially they're the AI design tool for [4:18] AI agents. I know it sounds very meta, but your agents are doing a lot more works in the in the [4:23] background now. So, there's a lot of AI native companies that are building AI agent solutions [4:29] instead of building for human. Our podcast, there's another company called um TensorMesh. [4:35] TensorMesh is leveraging KV cache to help agents to storage and find data. Uh in the past, [4:42] people are not in the past, but the people are still like, you know, for example, Databricks or [4:47] Snowflake. A lot of these companies are uh helping large company for like human engineers to like, [4:54] you know, process these data, storage data. But nowadays, like a lot of AI native companies are [5:00] building these agentic solution for people to for like AI agents to process this information. [5:10] The next trend that I I think uh it's really popular in Silicon Valley is voice is the new [5:17] moment. So, there's a few companies. How many people have used the app called WhisperFlow? [5:23] Okay, some people. Okay, so for those who don't know, WhisperFlow is a company that essentially [5:28] you can just record on your app and then it help you like type in the sentence that you're saying. [5:35] So, it's a voice agent that help you process like your thoughts. Um that's how I think about it, but [5:42] um we have some companies that raised like the mega rounds, the billion-dollar round [5:47] uh I I guess like the multi-million-dollar rounds and billion-dollar rounds including [5:51] companies like Sierra, company like Krista AI, Forethought, and among many others or Pylon. So, [5:58] many of them their founders and like leaders came out to our podcast. And what they're doing is [6:05] like they're leveraging voice AI to automate these call centers. So, in in the past and maybe you're [6:12] calling to return something or you're calling to book appointment and may or like maybe I'm sure [6:18] everybody have tried to, you know, try to find like, you know, calling a particular company for, [6:26] you know, Wi-Fi support. So, those may take hours, but now AI agent can process everybody's orders. [6:32] So, you don't have to wait in line for like 2 hours. You can be processed like immediately. [6:38] And in this segment I also want to highlight a company that I'm an I'm an investor of a company [6:43] called Hosty. So, Hosty was also invested by Great Adventure which was a spin-off from [6:49] Google's AI investment arm. And what they're doing is they are helping restaurants to process [6:58] bookings and orders. So, last year so there's one huge like it was like the graduation [7:07] kind of day and then there is like a person trying to book appointment for or like book [7:12] a restaurant for 70 people and they tried to call different restaurants, no one pick up or like when [7:17] people pick up they said they have to check with their boss. But this restaurant that's using Hosty [7:23] got the note immediately. The restaurant owner got a note from the the agent that's [7:29] note-taking when when the person was calling and immediately responds to the group. So, [7:36] it helps them getting a 70 people order immediately that they may never got. So, [7:43] those are the use cases for the voice AI agents that that is really trending in Silicon Valley. [7:51] The The next one is called uh I I'm observing uh that's happening in the AI investing industry is [7:59] like single player to multiplayer. Um how many people have heard of the two brothers who are uh [8:08] doing a AI Ozempic marketing company that become a billion dollar? Okay, five people. Okay, now you [8:15] know. So, there's people There's two brothers that they like leveraging AI agents to build like a [8:21] telemarketing company that's marketing Ozempic and they became a billion dollar company and then Wall [8:26] Street Journal and a lot of like these top press really covered them and that basically lead to uh [8:33] a saying that people think there will be a lot of one person billion dollar company in the future. [8:39] Uh so, uh on our show, there's a company called Pozeia. So, they The founder is [8:46] essentially creating these um softwares for or like AI agents for people to build these one [8:53] person billion dollar companies. Um but uh a few weeks ago, I was invited to uh [8:59] a dinner with a Slack CMO. Slack launches uh MCP and which helps you to collaborate with your team, [9:10] your team's agent, your agent, and everyone else. Which make me think the next generation [9:16] of work shouldn't be you work by yourself. Um essentially, like I feel like every day we only [9:23] have like limited amount of time. Um spending the time with the people who have the same passion, [9:28] the same dream, sharing the same goal is extremely important. It helps you and motivates you to go [9:34] through the day-to-day work. Agent doesn't really replace your co-worker, but it creates a new kind [9:41] of like assistant or co-worker that for you, so you are able to leverage these things with [9:47] uh your teammates to create more value for your company and for yourself. [9:56] What does this mean for a regular founder or investor is [10:01] many of the foundational model already did the groundwork for people to leverage AI day-to-day, [10:08] whether you're using ChatGPT or Anthropic or Perplexity, but the money is still going to be [10:15] made in the application layers. Uh in Silicon Valley, there is a really popular term called [10:21] cloud-shotted, which means like Anthropic is creating a solution to disrupt different [10:27] industries. They're obviously really, really good and uh but I personally don't think in the future [10:33] every single industry will only rely on one single model. There is still going to be application to [10:40] be made that really solves the last mile problem. People here are probably having [10:46] different expertise, whether you're doctor, you're lawyer, you are engineer, you are you probably or [10:53] like maybe we have I was just chatting with my new friend, he is a sports journalist, right? So, you [10:59] have your own expertise and insights that nobody else have. Uh we're in a era that anyone can build [11:05] anything, which means like taste and distribution more than matters more than ever. Um taste means [11:13] like you have a judgment, it's based off your personal experience and you can tell like what [11:19] can work, what cannot work only based off your earned insight from your past like work experience [11:24] or life experience. And that really helps you to like home in on what exactly to build because um [11:33] we are never short of any apps or any like AI solutions, but there is always there's always like [11:39] a different process or like, you know, people are locking in onto like a very different workflow. So [11:46] your solution kind of matters to people in your industry. And second of all, I want to highlight [11:52] for the founders out there, if you want to found a company nowadays, it's more important than ever [11:57] to build a distribution channel. Whether is How many people have heard of Cluey? [12:05] Okay, two people. Wow, okay. So it's Yeah, sorry. It is a Silicon Valley company that has like [12:11] essentially like the founders are continuously like creating a lot of like really interesting [12:15] media content. They remade like The Office. Though like I think like maybe that's a really extreme [12:20] example of like every single day you're creating a mini TV show to promote your SaaS product, but [12:26] many of the I feel like a lot of but we can all learn from this new way of building distribution. [12:33] As every day you just share a piece of information that you find interesting [12:38] in your industry or like an insight that you have earned from working the industry [12:43] or just share your life journey as a founder to to people that like who follows you and those [12:50] will significantly help you stand out among every other person. So I want to Can we show [12:58] Well, maybe we'll do it like a quick exercise before we we wrap. So maybe everybody stand up. [13:08] Okay, so founders, please sit down. [13:14] Okay, investors, please sit down. [13:20] Operators, please sit down. Who are the rest of you? [13:27] Alfred Oh my god, the operators means like if you work at a company, you're operator. [13:32] Okay, you can sit down. Yeah. Um so okay, so we have a lot of operators in here. So [13:40] I guess like the way the fastest way what you can learn from like this talk hopefully is [13:47] in the end of day, I think like you can start building things off your own experience. Nowadays [13:52] like leveraging cloud co-work or leveraging different tools like Gamma, like you know, Clay, [13:58] you can build a lot of like different solutions for your own company. And I think throughout [14:04] building you actually create the data flywheel as you are the user of your own product. Maybe [14:10] your team member is using your product or like your clients are using your product. And you [14:15] after that, I think speed really just matters and you immediately take on feedback from [14:21] the people who are leveraging like are using your product and are really improving on [14:26] the solution. And those are what's going to give you a moat and whether it's like your own workflow [14:32] or to like maybe one day you want to raise money to start a business. So those are my takes on [14:39] AI investing trends. Anyone have any questions since we have a very intimate audience group here. [14:46] It's not very intimate but it you know, if you have any questions, please shout. [14:51] [clears throat] [15:06] What is the company you have? Uh do you mean like how do you raise money after that or like [15:16] Amazing question. I think personally I would not Okay, so after family and friends, [15:23] I would go after the not family and not friends because I think that's how you actually find [15:28] the real customer. I may think of like a few design partners I have in mind. So for example, [15:33] if you're building, let's say the AI Figma, right? So I may talk to you like people that's [15:39] in the design industry. You may go to like a local design workshop or like find the designer [15:43] in like a Facebook group and then you started chatting with the people there to gather feedback [15:47] and then ask them to pay. I think that's a really important step. Um so I feel like as an investor, [15:54] the evaluation for us like or the diligence is more of like is there any paid user? Because if [16:00] they're not paid, they won't be able to You won't be able to to like validate if this idea actually [16:07] matters cuz customer cons. That's like all that matters in the future. Yeah, how uh [16:14] So this isn't necessarily capital related, but you mentioned how the distribution being more of the [16:20] main focus. Do you think that also applies and is important for operators within companies just to [16:25] Hell Yeah, I'm sorry. Like I don't know how to say this in a more calm way, but definitely. [16:30] Because I personally, you know, I mean, I'm basically I'm a LinkedIn warrior. So [16:35] I'm I'm Grace Going on LinkedIn if you want to find me. So I think in the in the current era, [16:41] people's job are under a really huge shift, right? So with AI, with AI agent, with everything else. [16:47] Um there is a lot of AI company are doing service and outcome-based pricing, which means they're in [16:56] essentially creating the AI employee and they can charge a higher markups than a seat-based [17:02] pricing means like no one's charging $12.99 for, you know, Canva, but they are charging, you know, [17:09] 10K per month as a like, you know, as a solution as like your design your as your designer, right? [17:15] So even it is designed by AI, but people are charging this price to got the same outcome. [17:21] Um so, for those who are more of like a thought leader in the industry, [17:27] you have a better advantage on staying in your current position as well as like your I think uh [17:33] I think one thing is building distribution, the other thing is actually building with the AI [17:37] tools. Like in Silicon Valley, I think that's that our culture is extremely different than the East [17:43] Coast or like and just literally anywhere else is because everybody's a builder. The builder mindset [17:47] is extremely like like set in among the people. So, whether you're like a marketer or you're a [17:55] salesperson, people are building their own CRM, their own air table at home because of like they [18:01] they they see the value of like leveraging these AI tools to make yourself more efficient to stay [18:07] quote-unquote in shape at work, right? So, I feel like those are my my thoughts. And like I gave So, [18:14] being a thought leader or building in public really incentivize you to consume more knowledge [18:19] in the space. When you're When you have to output, you really have to input. So, [18:23] you will have to like force yourself to adopt these new AI tools or like meet industry experts [18:28] or like just essentially giving yourself a edge than like many of your colleagues. [18:35] Okay. Uh lady. You, yeah. [18:53] I have an answer that you probably don't want to hear. Okay. So, I think investors many of [18:59] the Silicon Valley Valley investors are really investing on the talent. And the talent meaning [19:04] like a lot of times is like where does a logo actually come from? And many of the case is like, [19:10] you know, maybe people build a name for themselves in a particular industry. Uh So, they have they're [19:16] able to raise a lot of money. So, yesterday I was thinking why that happens is because of let's say [19:21] you're a second time founder, right? You have built something extremely successful. And now [19:26] you're out there fundraising because it saves investor a lot of diligence to bet on, you know, [19:31] Travis Kalanick of Uber than, you know, a random person who never started a scalable business [19:37] before. So, it's a lot safer for investors and it's easier for them. But, however, for the [19:42] founder themselves it may not be the right case. It's like, yes, you got a lot of validation from [19:47] raising a lot of money immediately, but it doesn't really validate your idea at all. It's not what [19:54] make you great for the first time. What make you great for the first time is like nobody believed [19:58] in Uber, but Uber is like break out success, right? And then you have to prove yourself [20:02] through your product, through your distribution, through your team. But, essentially, like even [20:07] people sure, people can raise a lot of money, but it doesn't mean like they will be successful. [20:14] Did all of you raise a lot of money? Is that why you're leaving? [20:18] Okay. Yeah, sir. This sir. [20:38] Sorry, can you say it again? [20:50] [snorts] [20:51] Yeah, so that basically comes with a data flywheel. So, essentially, the short answer is [20:57] there's nothing can predict someone just go after you. Like, especially if you're successful, [21:02] there is 100% sure chance that someone is going going after you. And then use exactly your method, [21:07] ball by But the thing is like I think speed matters, right? So the thing is you have earned [21:12] a lot of insight throughout your implementation of your idea. Let's say you started this app like 6 [21:18] months ago compared to me copying you starting this app like now. And I still have to spend [21:24] a few months to really going to have of like oh why is people buying my stuff and why is people [21:29] like only buying your stuff and those are a lot of a lot of time that's basically wasted [21:34] to like you know figure things out. And now you already have some data from your customer [21:40] and you're leveraging that data to create better service. And then you're leveraging [21:43] the new data to create better service. So you have a data flywheel. That's the only thing that [21:50] I guess you can do. And then there is also like I mean I interviewed Jim McKelvey who is the founder [21:54] of like Square and essentially he said there is a um you know it's a stack of knowledge that's [22:00] like based off like different experiences that you have that it's like a hard to be replicated. [22:08] Okay, anyone else? [22:16] Okay. [22:24] Oh next 2 to 3 years. I'll tell you next year. [22:29] Yes, sir. Are you do you Yeah, you. Oh okay, good. [22:37] Yeah, I'm sorry. I like literally my answer is like I will tell you next year cuz who knows. [22:42] Everything changes all the time. [22:46] Okay, everyone oh okay, yes lady. [23:07] I think it's kind of like you're on a roller coaster. Sometimes it's like up, [23:12] sometimes it's down. So like I personally feel like it doesn't really matter if we're at a [23:16] bubble or not. Like I think the company that will actually generate revenue and creating [23:22] value for the customer. If the customer continuously buying, [23:26] you're safe. But like I think essentially the bubble or not bubble thing is like we're kind of [23:32] always in a bubble. As like if you're building something that's like or like if people are [23:36] raising money for things that like that may never really make sense. I feel like it is a bubble, but [23:42] I I personally my take is like it doesn't really matter to like a general like you know customer or [23:47] like a general like operator at at a job. I think the only thing you can protect yourself is like by [23:54] like using AI tool every day and then building every day. So you have first hand experience. So [24:00] So you can decide for yourself on like if this thing is actually helpful for your job or not.