Cursor Defies Microsoft Copilot as ARR Surges to $50 Million
Martin Casado, Sarah Wang, Matt Bornstein
Controlling the user interface rather than training proprietary coding models allowed Cursor to outmaneuver Microsoft by continuously cannibalizing its own features ahead of frontier model advances.
Cursor intentionally bypassed the early race to train coding-specific foundation models, betting on the 'Bitter Lesson' that general compute and frontier models would outpace domain models while the real leverage lay in the human-computer interface. (0:00)
Despite advice to build a lightweight VS Code plugin, Cursor chose to fork the IDE completely to gain total control over UI, navigation, and interaction latency, defying investor skepticism about enterprise adoption. (5:00)
The company shattered standard SaaS growth templates by expanding ARR from $4 million to $50 million in four months strictly via self-serve inbound and bottom-up developer virality. (10:00)
Cursor transformed from an IDE into an enterprise-dominant powerhouse by having technical founders allocate 40% of their time to recruiting and deploying an 'Epochs' methodology to recruit top-tier sales talent. (22:15)
Detailed notes
## Background
Published as a podcast episode on August 27, 2026, by Andreessen Horowitz General Partners Martin Casado, Sarah Wang, and Matt Bornstein, "Inside Cursor: The Anatomy of a Generational Startup" examines the meteoric ascent of AI coding platform Cursor (Anysphere). The episode unpacks how a small, product-obsessed team defied venture capital consensus to take on entrenched incumbents like Microsoft, forking VS Code, aggressively cannibalizing their own features, and mastering both bottoms-up adoption and enterprise sales.
## TL;DR
The hosts break down how Cursor rejected standard startup playbooks by forking VS Code and banking on human-model interface design rather than training domain-specific base models, propelling revenue from $4 million to $50 million in ARR over just four months. As frontier models advanced, the founders repeatedly cannibalized their own flagship features, pivoting swiftly from smart autocomplete to autonomous agents and model orchestration while remaining unfazed by rivals like Microsoft Copilot and Claude Code. To bridge into enterprise viability, the deeply technical team applied engineering-grade rigor to go-to-market recruiting, mapping historical software epochs and spending up to 40% of their time personally vetting sales talent. Supported by an uncompromising, craft-focused culture and high-leverage M&A, Cursor scaled into over half the Fortune 500 while preserving its core identity as a pure developer playground.
## Atmosphere & style
The conversation carries the energetic, admiring tone of an insider venture war room revisiting a generational triumph, blending tactical software analysis with vivid personal lore. Casado, Wang, and Bornstein display easy conversational chemistry, trading fond anecdotes about the founders' monomaniacal coding habits, their baffling indifference to conventional leisure, and their slippers-only Scandinavian office. It feels like an exhilarating, behind-the-curtain postmortem that balances hard-nosed business mechanics with genuine reverence for obsessive product craftsmanship.
## Bulletin
- `00:00` Cursor intentionally bypassed the early race to train coding-specific foundation models, betting on the 'Bitter Lesson' that general compute and frontier models would outpace domain models while the real leverage lay in the human-computer interface.
- `05:00` Despite advice to build a lightweight VS Code plugin, Cursor chose to fork the IDE completely to gain total control over UI, navigation, and interaction latency, defying investor skepticism about enterprise adoption.
- `10:00` The company shattered standard SaaS growth templates by expanding ARR from $4 million to $50 million in four months strictly via self-serve inbound and bottom-up developer virality.
- `22:15` Cursor transformed from an IDE into an enterprise-dominant powerhouse by having technical founders allocate 40% of their time to recruiting and deploying an 'Epochs' methodology to recruit top-tier sales talent.
## Key takeaways
- `00:00` The early 2024 AI coding landscape was divided between training base models from scratch, building IDE plugins, and testing pure agents, while Cursor took a contrarian middle ground focused purely on the interaction layer.
- `00:00` Adherence to the 'Bitter Lesson' convinced Cursor's founders that general models would consistently defeat domain-specific coding models over time.
- `00:00` Cursor's foundational product thesis predicted that future software code will increasingly resemble pseudocode written in natural language intent.
- `05:00` Michael Truell's GP pitch stood out for his extreme discipline in saying 'no' to 90% of investor suggestions regarding enterprise roadmaps or adjacent features.
- `05:00` Forking VS Code rather than building an extension was deemed non-negotiable by the founders because building inside someone else's product boundary limits the interface experience.
- `07:30` Word-of-mouth adoption by frontier lab researchers at OpenAI and Midjourney, alongside endorsements from Andrej Karpathy and Lex Fridman, catalyzed exponential self-serve adoption.
- `10:00` Cursor's hypergrowth from $4M to $50M ARR within four months forced growth investors to abandon traditional assumptions about self-serve growth plateaus.
- `12:35` Rapid technical iteration was exemplified when Aman Sanger fixed monorepo choke points within 24 hours of Anthropic's John Schulman reporting loading failures.
- `15:15` The team treated aggressive competitors like Microsoft Copilot and Claude Code as a rotating cast of characters, remaining unfazed because large markets inevitably attract formidable rivals.
- `17:48` Cursor embraced radical self-cannibalization in the mold of Reed Hastings, transitioning rapidly from an autocomplete IDE to an autonomous agent platform and multi-model platform.
- `20:00` Criticisms regarding low gross margins and foundation model API costs were dismissed, demonstrating the tech-historical rule that user adoption and technical transformation precede business models.
- `22:15` Recognizing that single-user tiers were subsidized while enterprise tiers held true margin, the founders executed a decisive enterprise sales pivot, devoting 40% of their time to recruitment.
- `25:03` Cursor's 'Epochs' recruiting framework analyzed past inflection points across top tech firms to target top-performing sales reps and applied rigorous three-tier backchannel referencing.
- `30:00` The company's obsessive craftsmanship extended to a Scandinavian 'hygge' office environment featuring slippers and vintage furniture, reflecting a unified aesthetic and operating standard.
- `32:38` Decisive leadership enabled the founders to make fast, unambiguous personnel cuts and maintain tight control over their public voice without compromising their focus.
- `35:21` Cursor leveraged programmatic M&A to secure elite talent density, actively placing acquired startup founders into high-leverage operational leadership roles rather than individual contributor positions.
## Detailed walkthrough
### `00:00` Setting the Stage: The AI Coding Landscape in Early 2024
To understand the rise of Cursor, Martin Casado, Sarah Wang, and Matt Bornstein revisit the market landscape of late 2023 and early 2024. At the time, the state-of-the-art foundation models were GPT-4, Claude 3, and Llama 3. The dominant paradigm for AI-assisted programming was GitHub Copilot, integrated directly through VS Code and backed by the massive distribution of Microsoft. To the broader venture and tech ecosystem, Microsoft appeared virtually unassailable: they owned the developer environment (VS Code), the office suite, the leading model partnership, and Copilot. Coding assistance worked in rudimentary ways, but autonomous agents, multi-step execution loops, and dedicated test-time reasoning models did not yet exist.
```mermaid
flowchart TD
accTitle: Early 2024 AI Coding Strategies
accDescr: Three divergent startup strategies for AI coding in late 2023 and early 2024.
subgraph Approaches["Early 2024 AI Coding Bets"]
M["Coding-Specific Base Models<br/>(Train from scratch)"]
P["IDE Plugins<br/>(Retain existing VS Code harness)"]
C["Cursor's Middle Way<br/>(Fork the IDE + Bitter Lesson pilled)"]
end
M -->|"Belief"| R1["Base models cannot code well without domain training"]
P -->|"Belief"| R2["Developers will never switch away from VS Code"]
C -->|"Belief"| R3["Interface between human and model is the real lever"]
```
The startup ecosystem divided into starkly different camps:
- **Coding-specific foundation models:** Teams that believed generic large language models were incapable of deep programming tasks and that the only winning path was training dedicated coding models from scratch.
- **Plugins and harnesses:** Teams that viewed abandoning the existing IDE as reckless; they built lightweight plugins on top of VS Code to preserve developer habits.
- **Pure-play agent architectures:** Early experimenters trying to build fully autonomous agents.
Cursor (Anysphere) deliberately took a contrarian middle position. The founders were staunch adherents of the "Bitter Lesson"—the principle that general computational scaling and general models will consistently outperform human-crafted domain specializations over time. Rather than wasting capital training a coding model to compete with OpenAI or Anthropic, Cursor focused intensely on the human-computer interaction layer.
Founders Michael Truell and Aman Sanger operated on a core thesis about the future of software development: source code will eventually resemble the pseudocode found in classical computer science papers. Matt Bornstein highlights this conviction, explaining that Cursor anticipated natural-language intent specification long before it became the industry standard:
> “Code in the future will look like pseudocode.”
`05:00` Michael Truell's GP Pitch and Saying "No" to Conventional Wisdom
When Michael Truell came in to pitch the Andreessen Horowitz general partner (GP) group, the partnership witnessed an extraordinary level of discipline. Rather than pandering to venture capitalists by promising adjacent features, service revenue, or enterprise roadmaps, Michael spent roughly 90% of the meeting politely declining suggestions. VCs would pose open-ended hypothetical questions about other potential directions or expansion areas, and Michael would listen patiently before answering with flat rejection: “Hmm, interesting. No, we're not going to do that.”
Sarah Wang notes that one of the most contentious debates among investors at the time was Cursor's architecture: Cursor was a complete fork of VS Code rather than a plugin. Conventional wisdom dictated that large enterprises would never permit developers to install an unauthorized, third-party IDE fork. When the partners pushed Michael on whether a fork would ruin enterprise adoption, he held firm with absolute conviction. Martin Casado and Matt Bornstein explain the structured logic behind Cursor's choices:
1. **Why not train a coding-specific foundation model?** Cursor reasoned that humans do not communicate with programming models purely in syntax; they speak to them in human language and intent. Therefore, a true coding model must be a frontier natural language model that understands human experience—an endeavor far better suited to frontier research labs while Cursor focused on tasks like tab-completion.
2. **Why fork the IDE instead of building a plugin?** If the interface is the primary value driver, building a plugin leaves your experience constrained inside someone else's product boundary. A standalone fork provides complete control over UI, keyboard navigation, and the entire UX loop.
3. **Why resist enterprise sales early on?** If a product is genuinely exceptional, bottoms-up developer love serves as the ultimate go-to-market engine. Rushing into enterprise sales prematurely distracts from building the maximal version of the core product.
Martin Casado summarizes how this product-first focus eliminated the typical scattered roadmap of early AI startups:
> “If you really believe in the product, you would never do a plug-in because then you're part of somebody else's product. If you really believe in the product, you wouldn't do enterprise early, you do it later because the product itself is to kind of go to market motion.”
`07:30` Monomaniacal Focus, Founding Lore, and the Early Seed-to-Growth Transition
The partners recount personal anecdotes illustrating the team's relentless intensity and young age. During the diligence process, Matt Bornstein attempted to break the ice with Michael Truell by asking what he liked to do for fun outside of work. Truell seemed completely puzzled by the premise, repeatedly asking, “What do you mean?” When Matt clarified, “When you're not in the office,” Michael remained baffled.
Later, during a celebratory lunch following a16z's investment, the partners tried reminiscing about historical parallels from the dot-com internet boom, only for Michael to deadpan: “I was five.” Despite their youth, the Cursor founders demonstrated an encyclopedic grasp of computing history and software evolution.
For early-stage investors, backing Cursor felt obvious:
- Leading engineers at frontier AI labs (OpenAI, Midjourney, Replicate) were already passionate daily users.
- AI luminary Andrej Karpathy publicly praised and adopted the tool.
- User growth was tracking asymptotically via sheer word of mouth.
While seed and Series A investors could easily buy into early developer passion, conventional growth investors typically demand predictable enterprise metrics, multi-year contracts, and established sales pipelines. However, Cursor's explosive trajectory quickly forced growth investors—including Sarah Wang, who moved aggressively to lead Cursor's subsequent growth round—to compress their evaluation timelines and reconsider their standard playbooks.
`10:00` Breaking Growth Playbooks and Defying Conventional Wisdom
The velocity of Cursor’s ascent was dizzying. Andreessen Horowitz led the Series A in early summer (May–June), and by October, they were already doing the Series B. That summer became an inflection point: Andrej Karpathy posted a high-profile tweet praising Cursor, and the founders appeared on the Lex Fridman Podcast, delivering an exceptionally thoughtful breakdown of their vision for the future of coding.
```mermaid
flowchart LR
accTitle: Cursor Hypergrowth and Fundraising Timeline
accDescr: Diagram illustrating Cursor rapid growth trajectory from early summer to the Series B round, driven by viral adoption.
subgraph S1["Early Summer"]
A["a16z Leads Series A<br/>(May/June)"]
end
subgraph S2["Summer Inflection"]
B["Andrej Karpathy Tweet"]
C["Lex Fridman Podcast"]
B --> D["Vertical Liftoff<br/>ARR $4M to $50M"]
C --> D
end
subgraph S3["Autumn"]
E["a16z Co-Leads Series B<br/>(October)"]
end
A --> S2
D --> E
```
When Sarah Wang brought the company to a16z Growth lead David George, standard venture software underwriting had to be completely rewritten. Historically, growth memos rely on well-worn SaaS assumptions: self-serve acquisition eventually peters out, so companies layer on an enterprise sales leader around $25 million to $50 million in annual recurring revenue (ARR), and growth models deliberately assume growth asymptotes toward ~25% within five years.
When Wang met Cursor co-founder Michael Truell for coffee and presented standard growth-stage advice—offering to help them recruit an enterprise sales leader—Truell looked her dead in the eye and flatly rejected the premise: “Self-serve is not petering out.” Backing Cursor required tossing standard software growth templates out the window. As Wang reflected, “I think a lot of the growth investment decision at the time was throwing out some of the assumptions that we would typically make on how a company would, you know, eventually start to slow growth.”
- **Fundraising velocity**: The firm led the Series A in May/June and followed up with the Series B by October, tracking revenue jumping from $4 million to $50 million in roughly four months.
- **Defying standard SaaS mechanics**: Rejecting the immediate need for an enterprise sales motion because self-serve inbound showed no signs of slowing down at the typical $25M–$50M ARR plateau.
- **Public validation**: Karpathy's organic endorsements alongside the founders' deep-dive interview on the Lex Fridman Podcast drove widespread developer awareness.
---
`12:35` The Microsoft Juggernaut and the Monorepo Test
Looking back, backing Cursor might seem like an obvious move, but earlier in the company's lifecycle, the competitive picture appeared nearly impossible. An early-stage startup was daring to fork VS Code—an open-source editor owned and stewarded by Microsoft—while Microsoft held an unmatched distribution advantage through GitHub (with 100 million developers), a world-class enterprise sales machine, and exclusive access to OpenAI's frontier model weights for GitHub Copilot.
Matt Bornstein and an internal a16z team (including colleagues like Marco and Rachel) spent over a year working on the space before the market caught on. Bornstein noted that venture capital is fundamentally a team sport, and what stood out about the Cursor founders was their off-the-charts intelligence, their ability to context-switch between low-level technical infrastructure and broad market dynamics seamlessly, and how quickly they acted on new data.
A pivotal technical data point surfaced during a dinner organized by Marco, where Cursor co-founder Aman Sanger sat beside OpenAI co-founder John Schulman (who was then at Anthropic). Schulman mentioned that he had tried Cursor, but Anthropic's massive monorepo immediately choked the editor, preventing him from even loading their codebase. Sanger went back to work immediately and came back the very next day having pushed fixes to handle the monorepo. As Bornstein noted, this proved that Cursor was iterating at breakneck speed, while also reinforcing a core technical conviction: treating AI as an afterthought plugin on top of VS Code was an architectural dead end. Building a native fork was the only viable path to handle real-world developer workloads.
> “When you have founders that are students of what they do, they're very consistent in their answers.” — Martin Casado
---
`15:15` Unfazed by Competitors: From Windsurf to Claude Code
Throughout Cursor’s history, investors and observers repeatedly warned of existential threats. In major markets, a16z maintains a philosophy of betting on the category-defining independent player, even when an incumbent looks terrifying. But Cursor faced a gauntlet of competitive challenges beyond Microsoft Copilot:
- **Windsurf (Codeium)**: Gained heavy, grassroots momentum inside Y Combinator cohorts and built impressive early agent capabilities.
- **Anthropic's Claude Code**: Launched in May 2025 with massive fanfare, sparking renewed fears about foundation model providers moving up the stack.
When Casado checked in with Michael Truell around three months after the Claude Code launch to ask how he felt about the threat, Truell responded with complete equanimity. He pointed out that they were going after the biggest market in the world, meaning there would always be formidable rivals. Truell viewed them as a rotating cast of characters—first Copilot, later Claude Code, and inevitably others. As Casado observed:
> “Look, we are going after the biggest market in the world. There's going to be competitors. In fact, there's been a rotating cast of characters from the very beginning. Microsoft Copilot was the first one. Claude Code is one of a rotating set of characters. And we think in big markets, you're always going to have formidable competitors. That does not scare us.” — Martin Casado (recalling Michael Truell)
Casado characterized this mentality as a rare, Jeff Bezos-like blend of humility and bravado: founders who remain paranoid yet fundamentally unfazed because deep mastery of their product gives them absolute clarity of focus.
---
`17:48` The True "Oh Fuck" Moments: Relentless Model Advances and Self-Cannibalization
While the Cursor team never panicked over competitor product launches, they experienced real shocks whenever frontier models made sudden, dramatic leaps. These moments forced them to confront how rapidly the underlying technology was advancing:
- **Opus 4.5**: Released while the team was on vacation, prompting a collective realization across the ecosystem that reasoning capabilities were progressing far faster than anticipated.
- **Mythos**: The pre-release access period produced a similar visceral reaction as raw model intelligence took another massive jump.
Instead of fighting these capability leaps or clinging to existing features, Cursor repeatedly cannibalized its own product roadmap in the manner of Reed Hastings pivoting Netflix. When Karpathy famously posted praising Cursor’s rapid multi-line autocomplete experience (“tab, tab, tab”), smart tab completions were the company’s signature feature. Yet within just two years, the team discarded reliance on tab completion and transformed their entire identity:
```mermaid
flowchart LR
accTitle: Cursor Product Evolution
accDescr: The three stages of Cursor self-cannibalization from an IDE to an agent platform to a model platform.
P1["1. Next-Gen IDE<br/>(Smart Autocomplete / Tab-Tab-Tab)"] -->|"Self-Cannibalization"| P2["2. Agent Platform<br/>(Autonomous Task Execution)"]
P2 -->|"Self-Cannibalization"| P3["3. Model Platform<br/>(Multi-Model Orchestration)"]
```
Bornstein emphasized how rare it is for startup founders to possess the brutal discipline required to abandon their core hit features before someone else renders them obsolete:
> “I think another thing that the Cursor team probably doesn't get is to the extent they cannibalized themselves so dramatically in like a Reed Hastings type way in a matter of two years, right? So they were first an IDE and then they moved to an agent platform, then they moved to a model platform.” — Matt Bornstein
`20:00` Tech Transformations Precede Business Models: Shrugging Off Margin Skepticism
The conversation begins with an examination of the loud skepticism Cursor faced online during the summer of 2025 and following major frontier model releases like Claude 3.5 Sonnet and Opus. Critics across X (formerly Twitter) fixated heavily on Cursor’s unit economics, predicting their gross margins would collapse under the weight of model API costs. Martin Casado dismisses this commentary as standard social media noise detached from reality: “I just feel like X is driven by like tail chasing, adrenaline sinking in schadenfreude.” Casado emphasizes that across every technological paradigm shift, market demand arrives long before business models mature. During the early internet era, companies operated for years without clear monetization or positive unit economics, yet investors continuously make the mistake of picking apart early-stage margins while ignoring unprecedented product adoption and unbounded user demand.
Sarah Wang agrees, underscoring that technical adoption must come first: “First comes the technical transformation, then comes the business model.” In Cursor’s case, their aggressive initial strategy—capturing developers at breakneck speed regardless of margin hand-wringing—functioned as a backdoor moat. By aggregating the developer base and capturing daily coding workflows, Cursor accumulated the telemetry, proprietary interaction data, and institutional intuition necessary to train and deploy their own specialized models later. Wang points out that a company could never execute that strategy in reverse unless they were already established as an exceptionally well-capitalized frontier lab.
- **The Internet Parallel:** Martin Casado reminds listeners that nascent tech transformations historically focus on user utility and demand first; monetization and margin optimization consistently follow later.
- **The Backdoor Strategy:** Accumulating massive developer usage provided Cursor with the dataset and distribution needed to shift from third-party model dependency to proprietary models.
---
`22:15` Cracking the Enterprise: The 40% Recruiting Rule and the GTM Pivot
Matt Bornstein asks Martin Casado how a team so thoroughly allergic to enterprise software conventions transformed itself into an elite, sales-driven organization. In the company’s earliest days, co-founder Michael Truell explicitly pushed back against hiring traditional sales reps, opting instead to set up a bare-bones `enterprise@cursor.com` inbox for inbound inquiries. Casado explains that the founders are fundamentally rigorous students of company building: once they analyzed the economics of the market, they realized the enterprise tier held the margin. While foundation model labs were willing to heavily subsidize single-user developer tiers, they captured massive margins in the enterprise and API tiers. Once Cursor recognized that enterprise viability was existential, they executed the pivot with absolute commitment, orchestrating one of the fastest sales team expansions in software history.
Sarah Wang highlights how Cursor applied its obsessive product philosophy directly to enterprise go-to-market hiring, guided by the mantra that "how you do anything is how you do everything." She singles out early GTM leaders Jordan and Roman, who constructed the initial enterprise engine from scratch. What made the technical founders unique was their sheer time allocation: the founding engineers routinely dedicated up to 40% of their working hours to recruiting. Casado marvels at this commitment, noting that it is virtually unheard of for deeply technical product founders to spend nearly half their working life sourcing and closing talent: “This is amazing for technical product founders to spend 40% of the time.” Wang recalls founders like Oskar taking overnight red-eye flights to Europe simply to close individual contributor engineers, demonstrating a level of hands-on intensity that set the standard for the rest of the company.
```mermaid
flowchart TD
accTitle: Cursor Enterprise Strategy and Expansion
accDescr: Flowchart illustrating how Cursor identified margin necessity and executed their GTM hiring engine.
subgraph Recognition ["Economic Realization"]
A["Frontier labs subsidize single-user tiers"] --> B["Enterprise tier holds the margin and bulk spend"]
B --> C["Enterprise transition identified as existential"]
end
subgraph Execution ["The GTM Pivot"]
C --> D["Founders dedicate 40% of time to recruiting"]
D --> E["Early GTM leaders (Jordan & Roman) build engine"]
E --> F["Rapid penetration of enterprise accounts"]
end
```
---
`25:03` The 'Epochs' Method: Applying Engineering Rigor to Sales Hiring
Wang and Casado dig into the tactical mechanics of how Cursor applied the rigorous screening methodologies of software engineering hiring to Account Executive (AE) recruitment. Rather than relying on traditional enterprise software rolodexes or hiring reps who merely followed standardized sales playbooks, Cursor conducted exhaustive, research-heavy talent mapping. They studied distinct historical "epochs" of enterprise sales across the top ten technology companies that pioneered specific go-to-market motions. From there, they identified the specific sub-teams within those companies, and ultimately isolated the individual number-one and number-two AEs who were personally responsible for driving those inflection points.
Once candidates were identified, Cursor relied heavily on three levels of backchannel referencing—speaking with a candidate's direct manager, their manager’s boss, and the executive above them—to determine if the AE was truly a self-sufficient superstar capable of navigating extreme early-stage ambiguity. Sarah Wang recalls sitting in Cursor's offices for eight consecutive hours helping the founders source: “I think figuring out what does the best look like at what period of time and then which individuals drove that, that's so applicable for engineering, obviously product building, but on the go to market side as well.” By maintaining this bar, Cursor successfully scaled its enterprise reach to penetrate over 50% of the Fortune 500 faster than any comparable startup, without diluting the company's internal culture.
```mermaid
flowchart LR
accTitle: Cursor's Epochs Sales Sourcing Framework
accDescr: Process of identifying, isolating, and validating top-tier account executive talent.
A["Identify Top 10 Epochal Companies"] --> B["Isolate Top 10 High-Performing Teams"]
B --> C["Target #1 and #2 Star Individual AEs"]
C --> D["Triple-Tier Backchannel Referencing<br>(Boss, Boss's Boss, Boss's Boss's Boss)"]
D --> E["Founding AE Cohort Hired"]
```
- **Targeted Sourcing:** Avoiding standard playbook reps in favor of "epoch-defining" performers who thrived amidst early-stage product fluidity.
- **Deep Referencing:** Running extensive multi-layer backchannels rather than relying on standard resume credentials or surface interviews.
- **Fortune 500 Velocity:** The resulting GTM engine propelled Cursor into more than half of the Fortune 500 at record pace.
---
`27:26` The Hacker Culture Paradox: Competing in the "Magnum Opus of Startups"
The partners reflect on the central paradox of Cursor’s culture: how a small group of purist programmers evolved into ruthless business operators. Bornstein notes that the founders' relentless work schedule did not stem from performative hustle culture or workaholism; they simply loved writing code, manipulating systems, and building tools they personally wanted to use. Yet, as Casado points out, their subsequent actions defied the stereotype of naive hacker founders. When corporate survival demanded it, the team immersed themselves entirely in enterprise contract structures, business models, and business development partnerships. High-powered founders work on whatever the business strictly requires, regardless of their personal background.
This adaptability proved vital given the sheer brutality of their competitive environment. While the tech industry has witnessed fierce battles before—such as the historical database wars—the scale of the software development market is orders of magnitude larger. Casado describes coding as the ultimate economic meta-category: "all expression of human thought in economic form is kind of like coding." Over a span of roughly two years, Cursor managed to navigate direct competition against Microsoft and GitHub Copilot on one flank, and frontier foundation labs like Anthropic on the other. Throughout that onslaught, they completely overhauled their technical product architecture, re-engineered their interface paradigms, and built a dominant enterprise sales operation. Casado summarizes the feat: “I think this is the magnum opus of startups.”
`30:00` Product Obsession, Taste, and the Hygge Office Culture
The partners reflect on the genuine obsession with programming that defines the Cursor founding team—Michael Truell, Sualeh Asif, and Aman Sanger. When asked what they did outside of work just for fun, their answer was invariably that they coded. As Martin Casado jokes, in an era where foundation models promise a post-scarcity future, coding may well be the last bastion of genuine human enjoyment. Yet beneath this pure technical passion, Michael Truell always maintained a keen strategic awareness of how to build an enduring enterprise. The founders understood that true impact requires pairing raw product obsession with sound business fundamentals. As Martin Casado puts it: “Yeah, if your goal is changing software, that's a critical... It's critical to get the business straight.”
This ethos of craftsmanship translated into every physical and operational detail under the principle that “how you do anything is how you do everything.” Rather than operating out of a standard, sterile tech office, the founders curated a distinctly warm, Scandinavian *hygge* environment. Visitors were famously asked to take off their outdoor shoes and slip into slippers, working beneath soft twinkle lights in a space filled with curated vintage furniture sourced from secondhand Scandinavian dealers. That welcoming, home-like atmosphere not only made long working hours comfortable, but reflected an aesthetic maturity and deliberate taste rarely seen in founders in their early twenties.
```mermaid
flowchart TD
accTitle: Cursor Operating and Cultural Philosophy
accDescr: Flowchart illustrating how Cursor's craft-driven mindset translates into execution and talent integration.
subgraph Craft["Craft & Environment"]
A["Obsessive Coding Culture"] --> B["Hygge Office & High Taste"]
end
subgraph Execution["Disciplined Execution"]
B --> C["Strict Marketing Voice ('Saying No')"]
C --> D["Rapid, Courageous Personnel Calls"]
end
subgraph Growth["Talent & M&A Leverage"]
D --> E["Talent-Driven M&A"]
E --> F["Empowering Acquired Founders"]
end
```
- **All-in coding mindset**: When pressed on what they did purely for leisure, the founders admitted that their hobby was simply more coding.
- **The "Hygge" workspace**: Cursor’s office featured a strict no-shoes, slippers-only policy, warm twinkle lights, and specialized Scandinavian furnishings designed to feel like a living room.
- **Holistic craftsmanship**: The founders applied the exact same meticulous taste they brought to user interfaces to hiring, event hosting, and office ambiance.
`32:38` Decisiveness, Hard Calls, and Cultural Alignment with Elon Musk
The same uncompromising taste guided Cursor's operational discipline, especially in their ability to say no. When venture partners proposed high-visibility promotional events or conventional marketing tactics, the founders routinely declined, insisting on preserving their understated, un-spammy tone on X. More importantly, this clarity extended to hard personnel choices. Unlike many first-time founding teams who procrastinate on organizational friction, Cursor's leadership moved fast and decisively on departures and organizational boundaries. Casado highlights this resolve as an underappreciated cornerstone of their success: “And so I would say it's almost this kind of untalked about Probably greatest strength of the company is that once they made a decision, it doesn't matter how hard that was, whether it's deciding not to enter a market, deciding to be very limited in what they're doing, they kind of follow through or who would stay in the company, who wouldn't stay in the company.”
This velocity, tenacity, and immunity to external skepticism drew direct comparisons to Elon Musk’s operating playbook. On every level—technical, strategic, and cultural—the hosts see a natural alignment with Musk’s ecosystem: pairing massive compute with Cursor’s proprietary data and developer distribution, united by the conviction that software engineering is the critical lever to reshape global computing. The partners recall meeting SpaceX in 2019 when Starlink was merely an audacious concept without a single satellite in orbit, facing down legacy telecom giants like Comcast. Cursor demonstrated that same disregard for entrenched incumbents, repeatedly defying consensus predictions of their demise and crushing every internal forecast they set. Casado underscores the compatibility: “And so, I mean, I've been involved in a lot of M&A. And I would say this is probably the best fit I've ever seen as far as just the full package.”
- **Marketing restraint**: Cursor refused broad, generic promotional campaigns or conventional startup marketing, restricting public communications strictly to authentic product demonstrations and engineering updates.
- **Fast personnel calls**: The founders demonstrated an exceptional willingness to execute difficult departures and structural cuts rapidly without lingering ambiguity.
- **SpaceX parallel**: Much like SpaceX taking on legacy aerospace and telecom monopolies with Starlink, Cursor tackled incumbents like Microsoft by out-iterating them and shattering internal growth milestones.
- **Skeptic resilience**: Both Cursor and Musk’s ventures consistently endured waves of consensus pessimism proclaiming their imminent demise, answering with relentless operational focus.
`35:21` Talent-First M&A and Integrating Founder-Level Operators
Cursor approached mergers and acquisitions with an unconventional, talent-first strategy. In an intensely competitive generative AI talent market, they used programmatic acqui-hires to pull in premier engineering density. Where most early-stage startups avoid acquisitions due to fear of operational friction, Cursor’s internal culture was so well-defined that they were fearless about taking on complexity. While later transactions—such as acquiring code-review platform Graphite—took on explicit strategic and product importance, their foundational M&A muscle was built around bringing elite operators and former founders into their ranks.
Unlike the common industry pattern—seen frequently at labs like Anthropic, where incoming startup executives are relegated to individual contributor (IC) roles—Cursor actively empowered acquired founders to run high-leverage functions across the business. Key additions like Tikhon "Tito" Bernstam (founder of Koala) and Adam Ward on talent slotted directly into leadership, accelerating Cursor’s velocity rather than diluting it. The hosts conclude that while conventional wisdom views integrating acquired founders as an operational nightmare, Cursor’s execution proved it can become an extraordinary competitive advantage when backed by authentic cultural clarity.
“I remember, and they have so many great founders there, but I remember when they brought on Tito from Koala and then Adam Ward on the talent side, actually, and just seeing firsthand the leverage that they got from bringing in someone who was a founder, CEO, effectively, but slotted really well into the Cursor culture and just take that to great heights and add to the acceleration.” — Sarah Wang
- **Talent density via M&A**: Cursor treated early acquisitions as a primary pipeline for securing rare AI and infrastructure talent amid severe market competition.
- **Fearlessness toward complexity**: Unafraid of organizational messiness, the founders trusted their strong internal culture to absorb outside teams cleanly.
- **The Graphite transaction**: Represented a progression from pure talent acquisition into larger, strategic product expansion around the developer workflow.
- **Founder leverage over IC conversion**: Rather than burying former founder-CEOs into individual contributor roles, Cursor integrated them into pivotal operational positions to scale the company's trajectory.
## Notable quotes
> `05:00` If you really believe in the product, you would never do a plug-in because then you're part of somebody else's product. — Martin Casado
> `20:00` First comes the technical transformation, then comes the business model. — Sarah Wang
> `00:00` Code in the future will look like pseudocode. — Matt Bornstein
> `12:35` When you have founders that are students of what they do, they're very consistent in their answers. — Martin Casado
## Mentioned
Adam Ward, Aman Sanger, Andreessen Horowitz (a16z), Andrej Karpathy, Anthropic, Claude 3, Claude 3.5 Sonnet, Claude Code, Comcast, Cursor (Anysphere), David George, Elon Musk, GitHub Copilot, GPT-4, Graphite, Jeff Bezos, John Schulman, Jordan, Koala, Lex Fridman Podcast, Llama 3, Martin Casado, Matt Bornstein, Michael Truell, Microsoft, Midjourney, Mythos, OpenAI, Opus 4.5, Oskar, Reed Hastings, Replicate, Roman, Sarah Wang, SpaceX, Starlink, Sualeh Asif, Tikhon 'Tito' Bernstam, VS Code, Windsurf (Codeium), Y Combinator
## Video description
a16z General Partners Martin Casado, Sarah Wang, and Matt Bornstein unpack the story of Cursor: how a small, product-obsessed team entered one of the most competitive markets in technology, took on incumbents with seemingly unbeatable advantages, and repeatedly made decisions that ran against conventional startup wisdom.
They revisit the early bet that the interface between humans and AI would matter more than building a coding-specific foundation model, why Cursor built its own product rather than a VS Code plugin, and how the founders' ability to say "no" became one of the company's defining strengths. They also discuss Cursor's rapid evolution from IDE to agent and model platform, and why the team was willing to cannibalize its own products as AI capabilities improved.
The conversation gets into what founders can learn from Cursor's approach to competition, hiring, enterprise sales, M&A, and company culture, including why the team remained unfazed by competitors from Microsoft to Anthropic and how its obsessive focus on product ultimately extended into every part of building the company.
Andreessen Horowitz partners Martin Casado, Sarah Wang, and Matt Bornstein argue that Cursor succeeded by rejecting standard software startup wisdom—forking the code editor, aggressively cannibalizing its own features, and transforming from a product-obsessed developer tool into an elite enterprise sales engine.
Architectural Conviction and Defying Startup Dogma (0:00)
In late 2023 and early 2024, the prevailing consensus favored either training dedicated coding models from scratch or building lightweight plugins on top of Microsoft's VS Code. Guided by the 'Bitter Lesson,' Cursor founders Michael Truell and Aman Sanger took a contrarian path, betting that frontier foundation labs would inevitably outperform specialized coding models. They insisted that the primary value lay in the interaction layer, choosing to fork VS Code entirely despite venture warnings that enterprises would refuse third-party editor forks.
During investor pitches, Truell rejected standard SaaS expansion playbooks and enterprise feature roadmaps, betting that bottom-up developer love would power distribution. The native fork allowed Cursor to control the entire user interface and execution loop, enabling rapid iterations such as resolving repository load failures on Anthropic's monorepo in a single day when standard plugins failed.
Hypergrowth and Unsentimental Self-Cannibalization (10:00)
Cursor's trajectory challenged standard growth underwriting models. Between May and October, annual recurring revenue surged from $4 million to $50 million, propelled by organic praise from figures like Andrej Karpathy and an appearance on the Lex Fridman Podcast. While traditional growth models assume self-serve growth slows at $25 million to $50 million ARR, requiring an enterprise sales layer, Truell firmly rejected the premise that self-serve was drying up.
Rather than protecting its initial wins, Cursor repeatedly cannibalized its product roadmap in response to rapid frontier model improvements like Claude 3.5 Sonnet, Opus 4.5, and Mythos. In less than two years, the team discarded its signature tab-autocomplete feature, transitioning from an intelligent autocomplete editor into an autonomous agent platform, and ultimately into an orchestration platform for multiple foundation models.
The Systematic Pivot to Enterprise Margins (20:00)
When social media critics predicted gross margins would collapse under foundation model API expenses, Casado and Wang argued that user adoption and technical disruption invariably precede unit-economic optimization. High developer volume functioned as a defensive moat, capturing the interaction telemetry required to train proprietary models. Simultaneously, recognizing that frontier labs subsidize consumer tiers while extracting profits from business accounts, Cursor identified enterprise sales as existential.
Applying engineering discipline to sales hiring, Cursor analyzed historical 'epochs' across ten pioneering enterprise software firms to identify top individual account executives, validating them through three layers of backchannel referencing. The founding engineers dedicated up to 40% of their working hours to recruiting. This approach enabled Cursor to penetrate more than half of the Fortune 500 without diluting its core operational culture.
Taste-Driven Culture, Discipline, and Acqui-Hiring (30:00)
Beneath Cursor's technical drive was an exacting operational taste. The founders maintained a distinctive Scandinavian 'hygge' office—featuring slippers and soft lighting—under the operating principle that meticulous craftsmanship applies to every domain. They restricted public communication to technical updates, declined standard startup marketing events, and acted decisively on organizational restructuring and staff departures.
In contrast to the broader industry trend of relegating acquired startup founders to individual contributor roles, Cursor actively leveraged mergers and acquisitions for operational talent. Acqui-hiring experienced founders, such as Koala founder Tikhon Bernstam, and acquiring companies like Graphite allowed the firm to integrate proven leaders into critical operational roles, accelerating technical velocity amidst fierce competition from Microsoft and frontier AI labs.
The partners view Cursor as an exceptional startup case study, having navigated direct competition against incumbent Microsoft and frontier labs like Anthropic. Questions remain regarding how long bottom-up developer affinity can withstand continuous architectural shifts as frontier foundation model capabilities advance.
精炼阅读摘要
a16z 合伙人 Martin Casado、Sarah Wang 与 Matt Bornstein 探讨了 Cursor（Anysphere）的崛起历程，指出其凭借反共识的架构抉择、以自噬核心功能应对模型迭代的敏捷度，以及将工程严谨性注入企业级销售的高效执行力，在巨头封锁下成长为代表性的 AI 原生软件企业。
反共识架构选择与苦涩教训 (0:00)
在 2024 年初，AI 编程领域普遍由背靠微软与 VS Code 生态的 GitHub Copilot 主导，多数初创团队选择从头训练专用代码模型或开发轻量级 IDE 插件。Cursor 创始人 Michael Truell 和 Aman Sanger 坚信“苦涩教训”（Bitter Lesson），即通用计算扩展和前沿大模型的能力演进终将胜过领域专用优化。他们认为人类用自然语言与意图同模型交流，代码未来会趋于论文中的伪代码形式，因此核心壁垒不在自研底层基础模型，而在人机交互界面。
面对企业客户可能拒绝第三方未授权 IDE 的普遍质疑，Cursor 坚决拒绝走轻量插件路线，全面 Fork 了 VS Code。合伙人指出，若将 AI 视为既有编辑器之上的插件，体验必然受制于他人的产品边界。独立 Fork 赋予了团队对界面、键盘导航和完整用户交互闭环的绝对掌控。此外，当 Anthropic 联合创始人 John Schulman 反馈其庞大单体代码库（monorepo）导致 Cursor 崩溃时，团队仅用一天便推送修复，这种对真实开发者复杂工作负载的原生适配能力，印证了独立架构的必要性。
颠覆传统增长模型与激进自我颠覆 (10:00)
Cursor 展现出打破传统 SaaS 规律的扩张速度。a16z 于初夏主导其 A 轮融资，到同年 10 月又联合领投 B 轮，其年化经常性收入（ARR）在短短约四个月内从 400 万美元垂直飙升至 5000 万美元。在 Andrej Karpathy 等技术领袖公开推荐及播客深度访谈的推动下，产品依靠纯粹的口碑实现爆炸性增长。当投资人按照传统软件规律建议引入企业销售高管以应对自服务增长放缓时，创始人断然回绝，证明在 ARR 达到 2500 万至 5000 万美元的典型平台期前，自服务需求仍未见顶。
面对底层基础模型的快速进化，Cursor 展现出类似 Netflix 式的自我颠覆意志。在两年时间内，团队并没有固守其最初大获成功的智能补全（“tab, tab, tab”）标志性功能，而是主动推进产品重构：先是从下一代 IDE 转型为自主任务执行的 Agent 平台，随后进一步迭代为多模型编排平台。无论面对 Opus 4.5、Mythos 等前沿推理模型的飞跃，还是 Windsurf 与 Claude Code 等竞品的入局，团队均保持高度专注，快速调整架构以吸纳底层能力。
技术采用先行与纪律严明的企业级扩张 (20:00)
针对外界对其模型 API 成本可能侵蚀毛利率的唱衰，a16z 合伙人反驳称，在重大技术范式转移初期，市场需求和用户效用必然领先于商业模式的成熟。通过以极快速度聚合庞大的开发者群体与日常代码交互数据，Cursor 实际上构建了一条“后门护城河”，为日后部署和微调专有模型积累了关键数据资产与分发渠道。
意识到前沿实验室在企业级和 API 层面捕获核心利润后，原本高度排斥传统销售的团队迅速转向，将其技术严谨性融入企业级拓展（GTM）。创始团队甚至将高达 40% 的工作时间投入招聘。在招募客户经理（AE）时，团队推行“历史纪元”（Epochs）研究法：深入复盘率先跑通特定销售模式的十家标杆科技企业及其顶尖团队，锁定推动业务拐点的核心骨干，并执行穿透三级汇报关系的深度背景调查。这套高密度招聘机制推动 Cursor 以创纪录的速度渗透了超过 50% 的财富 500 强企业。
工匠文化、决断力与以人才为核心的并购 (30:00)
Cursor 的组织文化兼具纯粹的极客热爱与成熟的商业务实精神。创始人将编程本身视为最大的业余爱好，并将严苛品味融入工作空间，营造出需要换穿拖鞋、点缀柔和灯光与北欧二手家具的 hygge 办公环境。这种对细节的把控同样体现在精简务实的对外发声，以及对人员任免和组织边界的果断决策上，展现出类似 SpaceX 挑战老牌电信垄断巨头般的执行定力与抗外部干扰能力。
在激烈的人才争夺战中，Cursor 发展出独特的人才导向型并购策略。与传统大厂收购后将初创团队降级为普通个体贡献者（IC）的做法不同，Cursor 大胆重用被并购团队的创始人与高管（例如 Koala 创始人 Tito Bernstam 以及人才负责人 Adam Ward），让他们直接负责高杠杆的关键业务线。随后对代码审查平台 Graphite 的收购，更展示了其从纯人才吸纳迈向开发者工作流战略拼图扩张的路径演进。
Martin Casado 将 Cursor 的崛起评为初创企业的“代表作”（magnum opus）。该案例表明，在基础模型剧烈波动的时代，深刻理解底层计算规律、牢牢掌控终端交互界面，并敢于随着底层能力跃升而持续自噬现有产品的团队，能够突破看似不可逾越的巨头垄断壁垒。
Detailed notes — zh
## Background
本期内容由 a16z 普通合伙人 Martin Casado、Sarah Wang 与 Matt Bornstein 于 2026 年 8 月 27 日在播客及网络平台发布。节目回顾并剖析了代际级 AI 创业公司 Cursor 的崛起历程，探讨一支专注于极致产品的年轻团队如何在面对微软等行业巨头的极端竞争下，打破传统创业智慧，完成从轻量 IDE 到智能体与模型平台的跨越。
## TL;DR
节目深入拆解了 Cursor 团队反共识的战略抉择：他们在 2024 年初顶住微软 GitHub Copilot 与 VS Code 的垄断压力，放弃开发轻量插件与昂贵的基础代码模型，而是独立 Fork VS Code 并全面押注人机交互重构。随着底层大模型能力爆发，创始团队通过多次决绝的自我吞食（cannibalization），将产品从代码补全工具迅速演进为 Agent 和模型平台，并在四个月内实现 ARR 从 400 万美元跃升至 5000 万美元。此外，投资人们还详细剖析了 Cursor 在面对社交媒体质疑时的定力、将工程思维引入企业级销售的高效转型、独特的办公空间审美文化，以及通过高密度吸纳前创始人的激进并购策略。
## Atmosphere & style
对谈洋溢着风险投资人面对罕见“天才型创始人”时的赞叹、敬畏与复盘反思，充满行业内部视角的生动细节与第一手轶事。三位合伙人在对话中情绪饱满，交织着对初创团队惊人执行力的由衷折服，以及在传统 SaaS 投资模型被彻底打破时的戏剧性震撼。整场讨论节奏紧凑、思想密度极高，既有拆解残酷商业博弈的冷静推演，又带有见证科技历史“巅峰之作（magnum opus）”诞生时的激昂与兴奋。
## Bulletin
- `00:00` Cursor 在面对微软 GitHub Copilot 与 VS Code 的既有垄断时，坚决放弃轻量插件路线与自研代码大模型，直接深度 Fork VS Code 打造独立 IDE，将赌注完全押在交互界面与意图精简上。
- `10:00` 业务呈现出极端罕见的爆发式增长曲线：a16z 领投 A 轮后仅约四个月内，Cursor 的 ARR 迅速从 400 万美元跃升至 5000 万美元，并以前所未有的速度渗透了超过 50% 的《财富》500 强企业。
- `18:10` 团队在两年内经历两次决绝的“自我吞食”（cannibalization）：从以 Tab 代码补全闻名的 AI IDE，主动颠覆核心交互，跃升为自主智能体平台（Agent Platform），随后进一步演变为模型平台（Model Platform）。
- `22:20` Cursor 将严格的工程思维注入销售与并购体系，不仅创始人投入高达 40% 的时间在招聘上，还通过精细的多层穿透背调建立起极高“创始人密度”（Founder Density）的人才护城河。
## Key takeaways
- `00:00` 2024 年初科技界普遍认为微软在 AI 编程赛道已锁定胜局，Cursor 既不做专用基础模型也不寄生于插件，走出了一条被视为非理性的独立 Fork 路线。
- `00:00` 创始团队笃信“苦涩的教训”（The Bitter Lesson），洞察到编写代码正趋向自然语言与伪代码的高阶抽象，因此交互界面与意图对齐比孤立的代码补全更重要。
- `04:30` 创始人 Michael Truell 在向 VC 融资展示中花费 90% 的时间对各种衍生业务提议坚定说“不”，展现出极度罕见的产品定力与克制。
- `07:30` 年轻一代创始人虽然在互联网泡沫时期仅有几岁，却对计算机科学史与科技周期拥有极其通透的研读与敬畏，能够准确锚定公司的帕累托最优边界。
- `10:00` Cursor 彻底打破了 SaaS 领域在 ARR 达到 25M–50M 时自助增长必然放缓的定律，在未建立传统庞大直销军团前就实现了 ARR 从 400 万到 5000 万美元的狂飙。
- `12:15` 掌控底层编辑器的必要性通过实践得到检验：面对 Anthropic 庞大的代码单体仓库（Monorepo）导致系统崩溃的问题，团队次日即完成底层修复，这是轻量插件无法做到的。
- `15:20` 面对 GitHub Copilot、Windsurf 以及 Claude Code 等重量级对手的轮番进攻，创始团队将其视为大市场中的“走马灯角色”，保持专注且无所畏惧的心态。
- `18:10` 团队最大的危机感始终来自底层前沿模型能力的跨越（如 Opus 4.5、Mythos），并主动推翻曾被 Andrej Karpathy 盛赞的“Tab”王牌功能以实现组织与产品进化。
- `20:00` 针对社交媒体对其毛利率与商业模式的唱衰，投资人指出技术革命历来是技术变革先行、商业模式随后，Cursor 的后门策略使其积累了前沿实验室无法复制的交互与工程数据。
- `22:20` 技术型创始人展现出惊人的商业进化力：将 40% 的时间花在招聘上，甚至不惜乘坐红眼航班飞往欧洲亲自签约基层核心工程师。
- `25:00` 采用极其苛刻的精准猎头与多层穿透背调法选拔 GTM 人才，促使公司以极快速度渗透并拿下超过 50% 的《财富》500 强企业客户。
- `27:00` 编程被投资人定义为“以经济形式表达人类所有思想的元类别”（meta category），Cursor 在其最惨烈的竞争环境中完成了史诗级的自我迭代与蜕变。
- `30:00` 团队将产品层面的工匠精神与审美追求无缝延伸到物理办公空间与运营文化中，展现出“做一件事的方式就是做所有事的方式”的纯粹偏执。
- `32:15` 在严控业务范围与人事调整等艰难决策上雷厉风行，展现出类似马斯克旗下企业的超高速迭代与对抗外界唱衰的韧性。
- `35:00` 通过收购 Koala、Graphite 等顶尖团队，Cursor 将连续创业者高效融入体系，开创了拥有极高“创始人密度”的独特扩张组织架构。
## Detailed walkthrough
### `00:00` 逆流而上：两年前押注独立 AI 编程工具的“非理性”格局
2023 年底至 2024 年初，如果从风险投资或技术常规认知的角度来看，AI 编程赛道的竞争格局堪称严峻甚至令人望而生畏。彼时整个行业的主流共识完全倒向现存巨头，微软凭借 GitHub Copilot、Office 系列以及事实上的开发者标准编辑器 VS Code 构筑了密不透风的壁垒，整个科技界普遍认为微软在 AI 编程领域已经锁定了胜局。当时主流的前沿模型还是 GPT-4、Claude 3 和 Llama 3 时代，虽然人们已经见证了 AI 辅助编写代码的巨大潜力，但现代意义上的自主 Agent（智能体）、复杂执行循环（loops）以及高级推理模型（reasoning）都尚未成型。
在那个阶段，进入代码生成领域的创业公司呈现出两极分化的光谱：一端的创业者认为唯一的破局之道在于从头训练针对编程优化的专用基础模型（coding-specific base model），因为当时通用大模型在写代码上还不够完美；另一端则走上了最稳妥、摩擦力最小的路线——开发基于现有 VS Code 的扩展插件（plugins），理由是绝不能冒天下之大不韪去替换开发者已经高度依赖的整个 IDE。而 Cursor 却走出了一条截然相反、在当时被许多人视为“近乎非理性”的道路：他们既不做底层模型训练，也不寄生于现有的 IDE 插件生态，而是直接 Fork 了 VS Code，将赌注完全押在人机交互界面的重构上。
a16z 的合伙人回顾，Cursor 团队从一开始就深刻笃信“苦涩的教训”（The Bitter Lesson）。他们清醒地意识到，现阶段根本没必要去底层模型上同 OpenAI 或 Anthropic 展开消耗战。创始人 Michael Truell 和 Aman Sanger 当时反复强调一个核心信念：未来的代码在形式上会越来越趋近于计算机科学论文中的“伪代码”（pseudocode），人类工程师的角色是将自身意图浓缩到最极致的最小规格说明，而繁琐的实现则交给模型。面对市场上随处可见的强敌与巨头阴影，创始团队展现出罕见的沉着，Michael Truell 对此曾掷地有声地表示：“看，我们正在进军全球最大的市场，你总会遇到旗鼓相当的竞争对手，这吓不倒我们。”
```mermaid
flowchart TD
accTitle: 2024年初AI编程市场路径对比
accDescr: 展示2024年初市场主要竞争流派与Cursor独特路径的对比
subgraph Paradigm["2024年初 AI 编程的不同技术路线"]
A["底层路线：专用代码基础模型<br/>（试图在模型层战胜通用大模型）"]
B["轻量路线：VS Code 扩展插件<br/>（不破坏现有 IDE，依赖插件交互）"]
C["Cursor 路线：独立 Fork IDE<br/>（押注人机交互界面与意图精简）"]
end
A -->|"缺陷"| F["模型训练成本极高且难敌通用前沿模型"]
B -->|"缺陷"| G["交互空间受限，沦为他人产品附属品"]
C -->|"优势"| H["掌控完整交互体验，顺应底层大模型进化"]
```
- **主流模型背景**：2024 年初前沿模型以 GPT-4、Claude 3 和早期开源主力 Llama 3 为代表，底层模型迭代瞬息万变。
- **巨头垄断阴影**：微软集 GitHub Copilot、VS Code 与分发渠道于一身，在当时被绝大多数投资者与从业者视为不可战胜的行业铁板。
- **战略底层假设**：Cursor 团队笃信编写代码的本质正由逐行手写语法转向自然语言与伪代码的高阶抽象。
---
### `04:30` 拒绝的艺术：面对合伙人提问连说“No”的极致定力
当 a16z 团队开始深入调研 AI 编程生态时，Cursor 展现出了异乎寻常的特质。在许多同行仍忙于在各大科技会议上兜售宏大故事时，Cursor 却已经在业内最敏锐的顶尖工程师群体中引发了野火般的自然传播。OpenAI、Midjourney、Replicate 等一线前沿 AI 团队的工程师都在自发使用 Cursor，并深度认同其人机交互逻辑。当 a16z 邀请 Michael Truell 前往投资人合伙人会议（GP Group）进行 Pitch 时，整场会议的走向让经验丰富的合伙人们大吃一惊。
风险投资人在面对创业者时，往往习惯抛出各种发散性问题，例如询问是否会做外延功能、是否考虑兼顾插件形态、或是何时进入企业级销售等。然而在整场 Pitch 中，高达 90% 的时间竟是 Michael Truell 在极其礼貌却无比坚定地对合伙人们的各种提议说“不”。Matt Bornstein 回忆起这一幕时向创业者们感叹道：“九成的会议时间都是他在对各种提议说‘不’，对于那些正在打造自己所热爱产品的创业者来说，VC 总是会问你们一些关于其他能做的事情或相关业务的愚蠢问题。”面对投资人关于“为什么不做一个更易被企业 IT 部门接受的 VS Code 插件”的尖锐追问，Michael 看着全场投资人的眼睛，毫不动摇地给出了清晰且高度逻辑闭环的论证。
Cursor 之所以敢于做出极其逆向的决策，源于对底层逻辑的深度推演。例如在“为何不训练代码专用模型”的问题上，Cursor 团队指出：程序员并不是用纯代码在与 AI 交互，而是用人类自然语言传达意图；如果一个模型无法理解广阔的人类经验和语言语义，它根本无法胜任高阶的代码理解，因此训练一个代码模型本质上是在训练一个前沿通用语言模型，这绝非早期创业团队的最佳着力点。而在“为何必须独立 Fork VS Code”的问题上，他们的逻辑更加纯粹：如果你坚信产品本身的交互体验是决定胜负的唯一分水岭，你就绝不能将自己的命运局限在别人的插件窗口里；如果产品体验足够好，产品本身就是最好的自传播增长引擎（Product-Led Growth），过早投入大量精力去跑传统的企业自上而下销售流程只会分散精力。
- **拒绝顺从共识**：坚定拒绝开发兼容型插件路线，承受来自企业采购合规与传统 IT 采购阻力的偏见。
- **直击语言本质**：洞察到代码交互的本质是自然语言意图对齐，而非孤立的代码补全，从而放弃耗费巨资自研代码基础模型。
- **拒绝过早企业化**：将精力全部收敛于产品核心体验，坚信强大的产品驱动增长能够倒逼企业端自下而上采纳。
---
### `07:30` 纯粹的产品主义：年轻一代创始人的历史观与估值飞跃
在会后与创始团队的接触中，a16z 的合伙人们进一步体会到了这支年轻团队身上那种近乎偏执的专注力。为了争取投资份额，合伙人们在私下交流中试图了解创始人的个人生活。Martin Casado 问 Michael Truell 工作之余有什么业余爱好，结果 Michael 完全无法理解这个问题，反问 Martin 是什么意思；当 Martin 追问“就是离开办公室后你都干些什么”时，Michael 依旧一脸困惑，因为在创办 Cursor 的过程中，除了打造极致产品，他的生活里根本不存在所谓的消遣。甚至在一次完成投资后的庆祝午餐上，当合伙人们兴致勃勃地聊起第一代互联网繁荣时期的相似历史浪潮并向 Michael 询问当时的记忆时，Michael 只是平静地回答了一句：“那时候我才 5 岁”，引得全场投资人哑然失笑。
然而，尽管核心团队非常年轻，他们却对计算机科学史与科技周期有着极其深刻的研读与敬畏。Martin Casado 与 Matt Bornstein 观察到，这种“对技术发展史的透彻研读”正是顶级创始人的共有特征。他们清楚地知道前人的成功与失败轨迹，因此在设定公司边界时能够做到既极度雄心勃勃，又绝不盲目冒进。Matt Bornstein 评价道：“作为创始人，我们都渴望胸怀大志，并希望将边界推向帕累托最优前沿，但你必须清楚那条曲线究竟在哪里。”
对早期投资人而言，投资 Cursor 的决策水到渠成：Andrej Karpathy 在公开赞誉并日常使用 Cursor，硅谷顶尖技术圈将其视为主流选择，社区自发形成的品牌声量空前，用户增长呈现出几乎垂直的渐近线上升曲线。然而，这种反传统的成长节奏对于习惯用传统财务模型、自上而下销售漏斗和企业客户留存来做评估的后期成长型（Growth）投资团队而言，却带来了前所未有的认知冲击。正是凭借在产品数据、开发者狂热追捧与团队超强演进能力上的持续印证，负责成长型投资的合伙人 Sarah Wang 很快也打破了常规的评估周期，以极其紧凑的节奏领投了 Cursor 的下一轮大额融资。
- **超高专注度轶事**：创始人 Michael Truell 没有世俗意义上的“工作之余”，将全部精力纯粹倾注于产品雕琢。
- **时代反差感**：当资深合伙人探讨早期 PC 和互联网泡沫历史时，年轻创始人直言自己当时仅 5 岁，展现出新一代 AI 原生创业者的年轻特质与历史洞察力的奇妙融合。
- **投资共识的演进**：从早期投资团队凭借敏锐技术直觉和顶尖工程师（如 Andrej Karpathy）的自然背书果断入局，到成长型投资团队在极度压缩的时间窗口内打破常规估值模型紧随领投。
### `10:00` 垂直起飞与打破常规的增长神话
回顾 Cursor 的崛起速度令人感到不可思议。a16z 在初夏的五六月份刚刚领投了 A 轮，到了十月份就紧接着完成了 B 轮，整家公司的业务呈现出垂直拉升的态势。那年夏天发生了一系列关键事件：Andrej Karpathy 在 Twitter 上公开盛赞他们，随后创始团队登上了 Lex Fridman 的播客，极其深入且富有远见地阐述了他们对未来编程范式的构想。正如 Martin Casado 所指出的，Cursor 团队并不是在修修补补，而是在对一切进行彻底的重新构想。然而，这种非典型的爆发式增长也给传统的投资评估带来了挑战——按照以往的惯例，投资人根本无法写出一份符合标准的“成长型投资备忘录”（growth memo）。
Sarah Wang 回忆起当时如何说服增长基金的合伙人 David George 时提到，评估 Cursor 必须彻底打破固有思维框架。在 Martin 介绍她与 Michael 认识后，两人当天就约了一杯咖啡。按照 SaaS 行业的经典认知，当公司的年度经常性收入（ARR）达到 2500 万至 5000 万美元时，自助服务（self-serve）的自然增长往往会逐渐见顶放缓，这时就必须引入销售负责人（Sales leader）并建立企业销售团队。Sarah 当时准备了一整套关于下一阶段增长路径的材料，试图建议他们物色销售高管。然而 Michael 的反应完全颠覆了这一常理，Sarah Wang 回忆道：“Michael 直视着我的眼睛说：‘自助增长根本没有放缓。’”（“Michael looks me dead in the eye and is like, self-serve is not petering out, right?”）这种势头迫使投资团队彻底抛弃过去模型中关于“五年后增长率必然渐进下降至 25%”的减速假设，Cursor 在短短四个月内就实现了从 400 万美元 ARR 飙升至 5000 万美元的惊人跨越。
- 融资节奏极快：5/6 月完成 A 轮，10 月即完成 B 轮，ARR 在约 4 个月内从 400 万美元跃升至 5000 万美元。
- 舆论与生态破圈：Andrej Karpathy 发推力挺，团队作客 Lex Fridman 播客阐述未来编程愿景。
- 颠覆行业经验：打破了以往 ARR 达到 25M–50M 时必须依赖直销接棒自助增长的 SaaS 定律。
---
### `12:15` 挑战微软巨兽：为什么敢在死局中重注独立玩家
从当时的竞争格局来看，投资 Cursor 在许多人眼中简直是不可理喻的冒险。他们面对的直接对手是科技巨头微软，后者不仅拥有 GitHub 以及上面的 1 亿开发者，掌控着全球主流的代码编辑器 VS Code，独占 OpenAI 的核心模型权重，还拥有全世界最强大的企业软件分销渠道。按照商业常识，去 Fork 一个微软亲手打造的开源编辑器，用更小的模型去正面硬碰坐拥 OpenAI 权重的微软与 GitHub Copilot，生存几率微乎其微。然而 Matt Bornstein 及其团队早在半年前就建立起了极深的信念，并最终促成了早期投资。
Matt 指出，风险投资绝不是单打独斗的个人项目，a16z 内部为此专门组建了一个跨周期的 Cursor 研究小团队（包括 Marco 和 Rachel 等人），跟踪长达一年以上。团队敏锐地察觉到 Cursor 创始团队拥有超群的原始智力，他们能够极度丝滑地在底层硬核技术与高阶市场动态之间无缝切换上下文，并且对历史数据和市场变量有着惊人的吸收与执行能力。关于为什么必须单独 Fork 编辑器而不是做 VS Code 插件，当时的一个生动注脚是：Marco 促成了一次晚餐，让 Cursor 联合创始人 Aman 与 John Schulman（当时在 Anthropic）坐在一起。John 当时反馈说他尝试用过 Cursor，但 Anthropic 庞大的代码单体仓库（Monorepo）直接让它崩溃卡死，根本加载不出来。结果第二天 Aman 就迅速完成了修复。这一细节让投资团队更加确信：哪怕是全世界最聪明的一批研究员，也无法仅凭一个简单的插件搞定深度集成；在大体量且关键的市场中，只要存在一家极致敏捷的独立领先者，就永远值得重注。
```mermaid
flowchart TD
accTitle: Cursor 与微软现有优势的非对称竞争分析
accDescr: 对比微软在分销、模型和生态上的既有垄断优势，与 Cursor 凭借独立 Fork 架构和极致迭代速度突围的核心逻辑。
subgraph Incumbent["微软生态壁垒"]
MS1["分销渠道: GitHub (1亿开发者) + 全球企业销售网络"]
MS2["模型优势: 优先独占 OpenAI 顶尖权重"]
MS3["生态中枢: 掌控主流编辑器 VS Code"]
end
subgraph Startup["Cursor 突围路径"]
C1["产品形态: 拒绝轻量插件，深度 Fork 打造专属 IDE"]
C2["底层处理: 攻克超大 Monorepo 承载与极速索引"]
C3["团队特质: 极速迭代，次日响应核心开发者痛点"]
end
MS3 -.->|"插件架构受限"| C1
MS2 -.->|"体验压倒模型体积"| C2
MS1 -.->|"极速自助口碑爆发"| C3
```
- 微软拥有的绝对壁垒：VS Code 所有权、OpenAI 权重访问权限、GitHub 平台及 1 亿开发者基数、顶级企业销售网络。
- a16z 的研判机制：设立专属研究团队跟踪 Cursor 超过一年，Matt Bornstein 早期建立极强信念。
- John Schulman 轶事：Anthropic 的超大 Monorepo 导致初始加载崩溃，Aman 次日即推出优化版本，验证了深度掌控 IDE 架构的必要性。
---
### `15:20` 走马灯式的竞争对手与“偏执但从容”的创始人定力
自 Cursor 诞生以来，外部环境似乎始终笼罩在生存危机的阴影下。市场上不断涌现出声势浩大的挑战者：最早是巨无霸 GitHub Copilot；随后是精准切入 Y Combinator 生态、在智能体（Agent）工作流上引发巨大风潮的 Windsurf；再到后来 2025 年 5 月横空出世的 Claude Code。通常情况下，初创公司创始人在面对这种重量级选手的轮番进攻时，往往会焦虑万分、频繁向投资人求助，但 Cursor 的创始团队却展现出一种罕见的特质：既有着严阵以待的偏执，又在心态上全然不为所动（unfazed）。
在投资人决定继续重注后续轮次时，面对 2025 年初发布的 Claude Code，Sarah 曾专门询问 Michael 如何看待这一强大的新威胁。Martin Casado 赞叹道，Michael 给出的回应充满了一种极具贝索斯风格的从容、清醒与霸气，Martin Casado 引用 Michael 的话强调：“Michael 说道：‘我们在争夺的是全世界最大的市场，竞争对手必然层出不穷。事实上，从一开始就有一群走马灯般的角色在轮番登场。微软 Copilot 是第一个，Claude Code 也不过是这一连串走马灯式角色中的一个。在大市场里，你永远都会遭遇强大的对手，但这吓不倒我们。’”（“He said something to the extent, actually, you know, exactly what you're saying here, which is, look, we are going after the biggest market in the world. there's going to be competitors. In fact, there's been a rotating cast of characters from the very beginning. Microsoft Copilot was the first one. Claude Code is one of a rotating set of characters. And we think in big markets, you're always going to have formidable competitors. That does not scare us.”）这种兼具极端谦逊与强大自信的心态，构成了团队对抗行业巨头的核心壁垒。
- a16z 连续下注：从 A 轮开始参与了 Cursor 的每一轮融资，投入资金持续放大，Sarah Wang 共同领投了 B 轮与后续轮次。
- 动态竞争版图：先后经历了 GitHub Copilot 的笼罩、Windsurf 借助 YC 开发者与 Agent 能力掀起的浪潮，以及 Claude Code 的正面冲击。
- 创始人应对哲学：将竞争对手视为大市场中的“走马灯角色”（rotating cast of characters），保持专注与无畏。
---
### `18:10` 真正的危机时刻与自我迭代：从 IDE 到模型平台的两度自我吞食
对于 Cursor 团队而言，那些真正令所有人倒吸一口凉气的“要命时刻”（oh fuck moments），其实从来不是来自竞争对手的产品发布，而是来自底层基础模型能力的超预期跃迁。投资人们回忆，当 Claude 3.5/Opus 4.5 这一代模型发布时，大家正好都在度假，而全世界的技术圈仿佛都在那一瞬间被震撼了，所有人都在惊呼大模型的进化速度远比预想的要快得多；随后的 Mythos 预发布版本也带来了同样的冲击。每一次底层模型能力的台阶式跨越，都意味着现有的工程封装和交互方式面临被直接抹平的危险。
然而，Cursor 展现出了极其罕见的战略魄力——他们像当年 Netflix 的 Reed Hastings 那样，在短短两年的时间跨度内对自己的核心业务进行了多次剧烈而决绝的自我吞食（cannibalization）。起初，他们是一家专注交互的 IDE，凭借极度丝滑的“Tab”代码补全功能一举成名，正如 Andrej Karpathy 那条著名的推文所写，开发者每天都在疯狂敲击“Tab, tab, tab”；但随着模型理解与推理能力的爆发，这种曾经的核心功能迅速退化为基础标配。团队毫不犹豫地将自身重构，先是跨越成为智能体平台（Agent platform），紧接着又演进为模型平台（Model platform）。对于绝大多数初创团队而言，主动推翻并 cannibalize 刚给自己带来巨大成功的王牌产品是极度困难且反人性的，而 Cursor 团队却凭借深厚的自律与产品定力做到了这一点。
```mermaid
flowchart LR
accTitle: Cursor 业务形态的两阶段自我吞食进化图
accDescr: 展示 Cursor 如何在两年内先后从代码自动补全 IDE，主动自我吞食并升级为 Agent 平台和 Model 平台的演进过程。
subgraph Phase1["第 1 阶段：编辑器体验创新"]
IDE["下一代 AI IDE"]
TAB["核心壁垒：Tab 自动补全<br/>(Karpathy 盛赞的交互)"]
IDE --- TAB
end
subgraph Phase2["第 2 阶段：工作流主权转移"]
AGENT["智能体平台 (Agent Platform)"]
MULTI["复杂任务自主编排<br/>多文件与项目级重构"]
AGENT --- MULTI
end
subgraph Phase3["第 3 阶段：基础设施化"]
MODEL["模型平台 (Model Platform)"]
ROUTING["异构底层模型调度与聚合<br/>深度拥抱模型能力跃迁"]
MODEL --- ROUTING
end
Phase1 -->|"主动颠覆 Tab 逻辑"| Phase2
Phase2 -->|"承接底层模型跃升"| Phase3
```
- 核心危机源头：团队的危机感主要源自底层大模型能力的断层式提升（如 Opus 4.5、Mythos 预发布）。
- 自我吞食演进：两年内完成三重形态蜕变，从“AI IDE”演进为“智能体平台”，再进一步跃升为“模型平台”。
- 交互重点迁移：Andrej Karpathy 曾推崇备至的“Tab, tab, tab”自动补全，随着自主 Agent 能力的成熟，迅速让位于更高维度的编程交互。
### `20:00` 社交媒体的噪音与“技术先于商业模式”
讨论首先聚焦于社交平台 X（原 Twitter）上对 Cursor 毛利率的批评风潮。大约在 2025 年夏天，以及随后 Claude 3.5 Sonnet（或相关大模型更新）发布后的 1 月前后，网络上充斥着关于 Cursor 商业模式不可持续、缺乏单位经济效益的唱衰声音。
Martin Casado 犀利地反驳了这种观点，直言社交媒体的讨论充斥着跟风、肾上腺素刺激和幸灾乐祸（schadenfreude），言论与企业的真实经营状况往往存在巨大脱节。他指出，许多质疑其实来自试图让自己显得聪明的投资者，但技术革命的规律历来是技术变革先行，商业模式随后确立。回顾早期的互联网时代，创业者在很长一段时间内根本不知道如何变现，甚至谈不上利润率。而在当前，Cursor 面对的是近乎无限的需求和爆发式增长。
```mermaid
flowchart LR
accTitle: 技术变革与商业模式的演进路径
accDescr: 展示 Cursor 如何通过先获取用户与数据，再反向构建模型与商业闭环。
subgraph Phase1["第一阶段：后门策略"]
A["技术变革驱动"] --> B["极速获取海量用户"]
end
subgraph Phase2["第二阶段：资产沉淀"]
B --> C["积累独家交互数据与 Know-How"]
end
subgraph Phase3["第三阶段：商业闭环"]
C --> D["反向自研模型并做大 Enterprise 利润"]
end
```
Sarah Wang 赞同这一判断，并指出 Cursor 实施了一种极其大胆的“后门策略”（backdoor strategy）：团队在极短时间内抢占了全部用户群体，借此沉淀了海量代码交互数据与工程 Know-How，进而反向构建属于自己的模型。这种路径是绝大多数前沿大模型实验室（frontier labs）无法逆向复制的。
- Martin Casado 总结历史规律时指出：“Tech transformations always proceed figuring out the business.”（技术变革总是先于弄清楚商业模式而发生。）
- 谈及投资者对新生事物的质疑惯性，Martin 评价道：“Every single wave you have investors trying to say something that sounds smart, which is like picking apart, you know, the business model of a nascent tech transformation. And I just think historically, they've always been proven wrong.”（每一次浪潮中，总有投资者试图发表一些听起来很聪明的言论，比如对新生技术变革的商业模式挑刺。但从历史上看，他们总是被证明是错的。）
---
### `22:20` 进军企业级市场：为毛利而生的“闪电式”销售转型
Matt Bornstein 提出疑问：Cursor 最初是一家极度排斥销售的极客团队，甚至连联合创始人 Michael Truell 早先也对销售投资说“不”，仅仅在官网上挂了一个 `enterprise@cursor.com` 的邮箱来应付企业咨询，他们究竟是如何完成蜕变，成为一家具备极强销售能力与 Go-To-Market（GTM）体系的公司的？
Martin Casado 解释道，Cursor 团队非常善于学习市场规律，每一次融资和产品决策都经过深思熟虑。他们逐渐看清了竞争格局：顶级大模型实验室往往会在个人订阅层（single-seat tier）进行大额补贴甚至免费提供，但在第三方 API 和企业级（Enterprise）市场中却能攫取极高的毛利。团队意识到，企业级市场不仅是大部分软件支出的所在，更是商业模式生死存亡的关键价值支柱。一旦做出决策，Cursor 就展现出惊人的执行力，打造了可能是“人类历史上扩张最快”的企业销售团队。
- Sarah Wang 特别提到了早期销售引擎的奠基人 Jordan 和 Roman 及其团队，赞誉他们从零搭建了不可思议的销售机器。
- 创始团队将 40% 的时间倾注在招聘上。Martin Casado 惊叹这种投入度：“This is amazing for technical product founders to spend 40% of the time. I've never actually seen it before.”（对于技术和产品背景的创始人来说，能把 40% 的时间花在招聘上，这太不可思议了，我以前从未见过。）
- 联合创始人兼工程师的执着程度极高，例如 Amananger / Oscar 等核心成员经常通宵搭乘红眼航班飞往欧洲，只为亲自去锁定并 Close 一位普通的 IC（独立贡献者）软件工程师。
---
### `25:00` 精准猎头法：将工程思维引入 GTM 招聘
讨论转向 Cursor 如何将严格的工程招聘标准迁移到 GTM（客户执行即 AE）的选拔中。Sarah Wang 透露，Cursor 招聘创始销售团队（Founding AEs）时从不坐等简历，也不走老派软件公司的大水漫灌模式，而是采取了极其严密、层层穿透的背调机制（back-channel）。
他们的具体打法表现为对行业人才周期的深度调研：
1. **圈定标杆企业**：梳理出在特定销售模式中表现最卓越的前 10 家软件公司。
2. **细化顶级团队**：深入这些企业，找出战绩最出色的前 10 个具体销售小组。
3. **锁定超级个体**：精准定位推动变革的“第 1 名和第 2 名明星 AE”。
4. **多层穿透背调**：直接联系该候选人的直属上司、上司的上司、甚至上司的上司的上司，全方位交叉验证其是否具备在混沌不确定性中开拓业务的极客能力。
这种近乎苛刻的人才筛选机制虽然极大收窄了候选池，却确保了新招进来的销售骨干与 Cursor 团队的气质高度契合，避免了早期文化被稀释。正因如此，Cursor 以前所未见的速度迅速渗透并签下了超过 50% 的《财富》500 强企业（Fortune 500）。
- Sarah Wang 用名言概括团队的一贯作风：“How you do anything is how you do everything.”（你做任何一件事的方式，就是你做所有事的方式。）
- Matt Bornstein 评价这种高标准带来的惊人结果：“I mean, they got to over 50% of the Fortune 500 and I think faster than anyone we've ever seen.”（他们渗透了超过 50% 的财富 500 强企业，我认为这比我们见过的任何人都快。）
---
### `27:00` 文化悖论与创业史上竞争最惨烈的“大作”
三人探讨了 Cursor 独特的文化画像与内在的“巨大悖论”。Matt Bornstein 回忆道，Cursor 早期团队给外界的印象是近乎纯粹的黑客——他们无休无止地写代码，并不是因为沉迷于“工作狂”式的苦役，而是因为他们发自内心地热爱编写代码、构建底层复杂系统。
然而，Martin Casado 敏锐地指出了其中的反差：如果纯粹从行动轨迹来看，他们后来在商业模式、商务拓展（BizDev）和 GTM 组织上倾注了同样庞大的精力。这证明了真正顶级的技术型创始人并不会困在象牙塔里，而是能够敏捷地将精力转移到“公司在当时当下最需要的事情”上去。
三位合伙人一致认为，Cursor 所置身的是科技史上竞争最残酷的赛道。编程不仅是一个普通的垂直软件类别，更是一种“以经济形式表达人类所有思想的元类别（meta category）”，其规模远超历史上的“数据库之战”。
Cursor 的崛起之路如同一部史诗：他们在短短两年内，从最初直面微软（Microsoft）与 GitHub Copilot 的庞大分发壁垒，到后来应对 Anthropic 等拥有海量资本与底层大模型优势的实验室的进逼。在此期间，Cursor 既从底层彻底颠覆演进自身的产品架构，又在极短周期内重塑了商业化与企业服务能力。
- Martin Casado 总结这支创始团队解决业务痛点的能力：“I think that high-powered founding teams work on exactly what is needed by the business, and I think they did a very good job, and I think it allowed them to navigate the most competitive market we've probably ever seen.”（我认为高能的创始团队会完全扑在业务真正需要的事情上，他们做得非常出色，这也让他们能够驾驭我们所见过的竞争最激烈的市场。）
- 在评估公司这两年的演变与战局跨度时，Martin Casado 赞叹道：“I think this is the magnum opus of startups.”（我认为这是初创企业历史上的巅峰之作 / 代表作。）
### `30:00` 匠人精神与审美：如何做一件事，就如何做一切事
三人谈到 Cursor 团队对编码纯粹的热爱——即使私下询问他们在业余时间做什么来消遣，答案依然是“写代码”。在 AI 模型让内容与生产力进入“后稀缺时代”的背景下，写代码几乎成了他们唯一的乐趣堡垒。然而，创始人 Michael Truell 和他的团队绝非闭门造车的极客，他们始终对商业格局有着极高的清醒认知。“如果你想彻底改变软件行业，理顺商业逻辑是至关重要的，”Martin Casado 这样评价 Michael 的长远眼界。
Sarah Wang 分享了一个关于 Cursor 团队办公空间的趣闻，深刻体现了“你如何做一件事，就会如何做一切事”（how you do anything is how you do everything）的信条。Michael Truell、Sualeh Asif 和 Aman Sanger 将产品层面的工匠精神和审美直觉（craftsmanship and taste）完全延伸到了物理空间与文化构建中：
- **无鞋文化与拖鞋**：办公室以不穿鞋闻名，进入后换上拖鞋，给人一种强烈的归属感和居家感，这也促使团队几乎时刻都自愿待在办公室高强度工作。
- **斯堪的纳维亚 Hygge 风格**：室内挂着微光小彩灯（twinkle lights），氛围温馨舒适；当他们搬入更大的第二处办公室时，Sarah 称赞其中的一张沙发，团队表示那是特意从二手 Hygge 家具商那里淘来的，精准追求某种特定的美学质感。
这群二十出头的年轻人懂得办公环境如何反哺思考空间与产品形态。此外，这种审美与克制也直接体现在他们的市场决策上：他们对自己的声调极其挑剔，拒绝垃圾营销（spammy marketing），坚持只用 X（原 Twitter）发声，甚至屡次直截了当地拒绝投资人提出的各种办活动建议。
> “再一次验证了：你如何做一件事，就如何做一切事。” —— Sarah Wang
### `32:15` 果断决策的勇气与类似马斯克的特质
Cursor 团队最不为人道、却最具威力的核心优势，在于他们做出艰难决策的速度与执行力。无论是决定不进入某个市场、严控产品范围，还是在人事上决定“谁去谁留”，他们都毫不拖泥带水，这在极早期的初创团队中极为罕见。
这种特质让投资人联想到了埃隆·马斯克（Elon Musk）的做事风格，并将 Cursor 与马斯克旗下企业的文化契合度进行了拆解：
- **技术与资源互补**：马斯克方拥有海量算力（compute），Cursor 拥有分发渠道（distribution）与高质量数据。
- **愿景高度一致**：双方都坚信“代码是重塑整个人类计算乃至整个人类未来的核心路径”。
- **文化与执行契合**：极度偏重产品与工程研发、超高速迭代，面对艰难决策雷厉风行。
Matt Bornstein 回顾了 2019 年与 SpaceX 会面的经历：当时 SpaceX 纯粹是一家火箭发射公司，却自信宣称要为全球提供卫星通信网络，那时太空中甚至连一颗星链（Starlink）卫星都还没有，面对 Comcast 等庞然大物，他们最终颠覆了市场。Cursor 同样如此，在面对不可战胜的对手时，他们实际交出的业绩一次次远超给出的所有预测。两家公司还有一个共同点：在发展过程中，外界舆论都曾无数次唱衰它们的消亡，但团队都展现出对反对者极具韧性的坚持。
> “无论决策有多艰难，无论是决定不进入某个市场，决定对所做的事情进行严格限制，还是决定谁留在公司、谁离开公司，他们都会贯彻到底。” —— Martin Casado
### `35:00` 创纪录的并购策略：如何将创业者融入公司引擎
在当下的 AI 人才争夺战中，能够通过并购（Acquihire / Talent M&A）获取顶尖团队的公司占据了巨大优势。Cursor 展现出惊人的并购整合能力，且从未被操作层面的复杂性所困扰。其并购发展路径经历了清晰的演进：
```mermaid
flowchart LR
accTitle: Cursor并购演进图
accDescr: 展示Cursor早期以人才为核心的收购，到后期Graphite战略性并购的发展阶段。
subgraph EarlyStage["早期：人才获取并购"]
direction TB
A["顶尖团队Acquihire"] --> B["强文化磁场同化"]
B --> C["注入创始人杠杆效应<br/>如引入Koala的Tito"]
end
subgraph LateStage["后期：战略扩展并购"]
direction TB
D["战略性大体量并购<br/>如收购Graphite"] --> E["极高密度的战略协同"]
end
EarlyStage --> LateStage
```
- **早期以获取人才为导向**：迅速出手拿下极其优秀的初创团队。凭借自身极其强大的团队文化，Cursor 完全不担心文化稀释或运营混乱。
- **引入资深创始人带来巨大杠杆**：例如吸收了 Koala 创始人兼 CEO Tito，以及在人才招聘端引入的 Adam Ward。这些拥有创始人经历的复合型人才无缝嵌入 Cursor 体系，带来了惊人的业务加速度。
- **后期走向战略性并购**：在收购代码审查工具 Graphite 时，并购考量升维至整体战略方向，规模也更为庞大。
业界普遍认为将前创始人直接安插到团队中极难管理，且运营极其复杂，但 Cursor 却极其罕见地把这件事变成了公司的常规建制。如今外界常讨论 Anthropic 吸引大量前高管去做独立贡献者（IC），但 Cursor 实际上更早开创了这种“以创始人密度构建公司”的打法，并在极短时间内将其执行得极其严密扎实。
> “这可能是历史上做过并购及其他所有扩张的企业中，创始人密度最集中（concentrated）的一家公司。” —— Sarah Wang
## Notable quotes
> `07:30` 作为创始人，我们都渴望胸怀大志，并希望将边界推向帕累托最优前沿，但你必须清楚那条曲线究竟在哪里。 — Matt Bornstein
> `00:00` 我们正在进军全球最大的市场，你总会遇到旗鼓相当的竞争对手，这吓不倒我们。 — Michael Truell
> `20:00` Every single wave you have investors trying to say something that sounds smart, which is like picking apart, you know, the business model of a nascent tech transformation. And I just think historically, they've always been proven wrong. — Martin Casado
> `04:30` 对于那些正在打造自己所热爱产品的创业者来说，VC 总是会问你们一些关于其他能做的事情或相关业务的愚蠢问题。 — Matt Bornstein
## Mentioned
Cursor, Martin Casado, Sarah Wang, Matt Bornstein, Michael Truell, Aman Sanger, Sualeh Asif, David George, Andrej Karpathy, Lex Fridman, John Schulman, Elon Musk, Reed Hastings, a16z, Microsoft, GitHub, GitHub Copilot, VS Code, OpenAI, GPT-4, Anthropic, Claude 3, Claude 3.5 Sonnet, Claude Code, Opus 4.5, Mythos, Llama 3, Windsurf, Midjourney, Replicate, SpaceX, Starlink, Comcast, Koala, Graphite, Y Combinator
## Video description
a16z General Partners Martin Casado, Sarah Wang, and Matt Bornstein unpack the story of Cursor: how a small, product-obsessed team entered one of the most competitive markets in technology, took on incumbents with seemingly unbeatable advantages, and repeatedly made decisions that ran against conventional startup wisdom.
They revisit the early bet that the interface between humans and AI would matter more than building a coding-specific foundation model, why Cursor built its own product rather than a VS Code plugin, and how the founders' ability to say "no" became one of the company's defining strengths. They also discuss Cursor's rapid evolution from IDE to agent and model platform, and why the team was willing to cannibalize its own products as AI capabilities improved.
The conversation gets into what founders can learn from Cursor's approach to competition, hiring, enterprise sales, M&A, and company culture, including why the team remained unfazed by competitors from Microsoft to Anthropic and how its obsessive focus on product ultimately extended into every part of building the company.
[Source](https://mgln.ai/e/1344/afp-848985-injected.calisto.simplecastaudio.com/3f86df7b-51c6-4101-88a2-550dba782de8/episodes/06262bc9-7183-4be2-a4b5-6ed18cc6a418/audio/128/default.mp3?aid=rss_feed&awCollectionId=3f86df7b-51c6-4101-88a2-550dba782de8&awEpisodeId=06262bc9-7183-4be2-a4b5-6ed18cc6a418&feed=JGE3yC0V)
Source quotes
[7:30] Matt Bornstein: 作为创始人，我们都渴望胸怀大志，并希望将边界推向帕累托最优前沿，但你必须清楚那条曲线究竟在哪里。
[0:00] Michael Truell: 我们正在进军全球最大的市场，你总会遇到旗鼓相当的竞争对手，这吓不倒我们。
[20:00] Martin Casado: Every single wave you have investors trying to say something that sounds smart, which is like picking apart, you know, the business model of a nascent tech transformation. And I just think historically, they've always been proven wrong.
[4:30] Matt Bornstein: 对于那些正在打造自己所热爱产品的创业者来说，VC 总是会问你们一些关于其他能做的事情或相关业务的愚蠢问题。
Source quotes
[5:00] Martin Casado: If you really believe in the product, you would never do a plug-in because then you're part of somebody else's product.
[20:00] Sarah Wang: First comes the technical transformation, then comes the business model.
[0:00] Matt Bornstein: Code in the future will look like pseudocode.
[12:35] Martin Casado: When you have founders that are students of what they do, they're very consistent in their answers.
Original transcript
[0:00] We don't need to compete with Anthropic and OpenAI on models right now. The interface between the human and the model is the key thing. If you looked at the competitive landscape, it was almost silly. I asked Michael, I was like, what do you think about Claude Coates? And he said something to the extent, look, we are going after the biggest market in the world. You're always going to have comparable competitors. That does not scare us. As founders, we all want to be ambitious and we want to push to the maximum point, the Pareto frontier. But you have to kind of know what the curve is. For us, the decision to invest in Curse Free was actually pretty obvious. Andre Karpathy was using it. It was clearly a phenomenon. It was clearly like a known brand. They went and got all the users in record time. Now they have the asset, the data, the know-how to build their own models. You really couldn't do the flip of that unless you started as a frontier lab. They were definitely a couple of old moments. Like, remember when? Two years ago, betting on an independent AI coding company looked almost irrational. and access to some of the world's best AI models. And yet, Cursor broke through. In this episode, Martin Casado, Sarah Wang, and Matt Bornstein look back at the decisions that shaped Cursor from its earliest days, building a standalone product instead of a plugin, resisting the push into enterprise sales too early, and staying relentlessly focused even as the competitive landscape changed around them. They also unpacked how the company repeatedly reinvented itself as AI advanced, built a culture around speed and taste, and eventually intensity it brought to product to hiring, sales, and M&A. It's a conversation about Cursor, but also a broader look at what it takes to build a defining company in a market where the technology, competition, and conventional wisdom can change almost overnight. So to kick off, I actually wanted to take us back to early 2024. And obviously, you guys led the that year. But can you walk us through what was going on in the Q1, Q2 of that year and maybe how you guys started working with Cursor well before we ever partnered officially with them? So you have to sort of transport yourself back to late 2023, early 2024. I went back and looked it up. The leading models at the time were GPT-40, Cloud3, and Llama3. That feels like forever ago. Llama3? Yeah. That was actually pretty good. That was great. Yeah, it was a great model. Change in strategy, obviously. Now, the leading coding harness was Copilot. by like a mile, right? This was like the consensus thing. Microsoft was doing their Microsoft thing, right? They had Copilot, they had Office, they had VS Code, all of these things combined, they were going to somehow win AI. And so it was clear that AI coding was sort of working, but it was a different universe compared to today, right? Like you had the sort of early signs that you could do meaningful work with AI, but agents didn't really exist, loops didn't really exist, reasoning didn't really exist at the time. So it's sort of a crazy time. was a continuum of like companies going after coding, right? So on one hand, you had those that are like, the only way to win coding is to build a base model. And it has to be a coding specific one. And on the other hand, you had very presciently, by the way, companies just focus on agents, right? And like Cursor was somewhere in between. And not only... there were a million companies doing plugins to it. And both of these were very reasonable things to do, done by very smart people, right? That's what's so interesting about these new spaces. It's like smart entrepreneurs show up and do really interesting things. And in this case, model training was a very reasonable thing to do because the models weren't that good at coding yet. And there was sort of a theory that if we focus on this into a coding-specific model, that's the breakthrough. People doing plugins on the other side were sort of saying, hey, let's not toss out the whole IDE. That would be a crazy thing to do. the interface between human and model is really important. So Plugin is a great place to go. You know, it just kind of turned out that the thing that Cursor was doing was the right thing to do, right? Like us, they're a very bitter lesson pilled. At the time, they would have said, oh, there's no need for us to go train our own model. Of course, that changed later for different reasons, which we're going to get into later in this podcast. But it's like, we don't need to compete with Anthropic and OpenAI on models right now. The interface between the human and the model and Amon saying a lot and just saying very clearly from the earliest days of Cursor saying, Code in the future will look like pseudocode. I remember Michael making this point a lot. Like when you look at old like computer science papers, there's a section. Here's the algorithm, not in code, but in like pseudocode, which I'm sure many people are used to. And we don't exactly write in pseudocode now, but probably you could. Probably you could just take a pseudocode block from an old paper and like the models now would implement that. This is like really what the cursor guys thought from the very beginning. And they've proven to be exactly right about this, right? It's like this sort of natural language-ish,
[5:00] specification, like pairing the programmer's intent down to the minimum possible spec, That's what really sort of matters now. So anyway, we were in this world, as Martine said, we were talking to a ton of very smart, very interesting people working on coding in various ways. The cursor team kind of stood out because not only did they sort of align with the things that we thought about the world, more importantly, smarter people than us and engineers working. At the time, the leading AI companies like OpenAI, MidJourney, Replicate, a bunch of others all kind of use this and subscribe to this model. And the team, frankly, just seemed kind of special, right? They had an unusual degree of focus. We got Michael to come pitch our GP group. And 90% of the meeting was him saying no to things, which is just like for founders out there working on products that you love. VCs will always ask you kind of dumb questions about other things you might do or things you're related to or whatever. And Michael literally just sat there and very politely listened to all the questions and said, hmm, interesting. No, we're not going to do that. Wait, wait. Actually, on that point, can I ask you guys a question? And one of the things that he said no to was actually, and to your point on there are a maybe more consensus thought that a, VS Code fork versus the plugin into VS Code would, one would be better for enterprise and it wasn't the former, it would be the latter. And I recall, I think it was you, Martine, or maybe it was both of you asking him, well, are you going to get enterprise customers with this medium, right? And he looked at... all of us in the eye with high conviction. And well, I'm going to turn it back to you. And what did you think of his answer? And what gave you guys the conviction to take a pretty contrarian bet at the time? They actually had a very reasoned articulation for why, where they run the design spade mates the most sense. So for example, why wouldn't you build a coding specific foundation model? So what they said at the time was, it turns out that you don't speak to these models in code, you speak to these as humans and natural languages. Therefore, the model has to understand human experience and language, right? And so that means in order to build a coding model, you're actually building a frontier language model, and that's just a very hard thing to do. And then, of course, there are other companies doing it, and there are much more focused problems like TAM Complete to do. So I think every one of their decisions was very reasoned, in our opinion. They were also very product-focused, which you actually don't see a lot in AI. So they really believe that you can build a product, the product, if it's really good, will be adopted, which, by the That dictated a lot of their decisions. If you really believe in the product, you would never do a plug-in because then you're part of somebody else's product. If you really believe in the product, you wouldn't do enterprise early, you do it later because the product itself is to kind of go to market motion. And so there was, I would say, and Matt, let me know if you would agree, tremendous clarity on exactly what type of company they were and the decisions that they said all fell from that consistently. Where many of the companies we saw were kind of these pastiches of, we'll try all of these different things. We're not quite sure what's going to work. plugins and their services and their whatever, whatever, whatever. This was a product company going after it, but they considered it to be a product problem. Yeah, no, I think that's exactly right. And I think they had the ambition and the courage to kind of do the maximal form of what the product would look like, right? Yeah. you know, as founders, we all want to be ambitious, right? And we want to push to the maximum point, the Pareto frontier. But like, you have to kind of know what the curve is. And they just kind of knew what the curve was, right? And then get distracted by the sort of non-product things. I remember when we were trying to convince the Cursor team to let us invest in their company, I asked Michael once what he did for fun outside of work. And I was like, he didn't understand the question. He's like, what do you mean? I'm like, when you're not in the office, he's like, what do you mean? I also remember, by the way, when we had the lunch after we did the investment, about internet days that were similar? He was like, I was five. And we were like, oh. But I mean, what's amazing about not just Michael, but all of the founders is even though they were pretty young when a lot of these things happened, they actually were kind of students of all this history, which is a trait that we see, Al Trampel talks about this a lot, we see this across all of our top founders that they're like able to study and really learn from what's kind of happened before. I'm gonna turn the question back on you. I think Matt would agree. is actually pretty obvious. The team is phenomenal. It's incredibly focused. All of us were users. Half the team actually used Cursor. Remember Andre Karpathy was using it. It was clearly a phenomenon. It was clearly a known brand. The growth was asymptotic. So from an early stage investor, it was a pretty straightforward decision. It was a much less obvious decision for a growth investor. And so I think probably what you heard in that first meeting maybe was a little bit dissonant with how you'd growth invest. But then you very, very quickly led the next round. So maybe talk through how you kind of got past the typical growth evaluation and got conviction. I mean... I think the timelines are so compressed here.
[10:00] It's crazy to think back on because I remember you guys led the A in May, June, right? Early summer. And by October, we did the B. And so it was clearly just vertical liftoff. And there were a couple of things that summer, right? You guys talked about Andre Carpathy tweeting about them, but they went on Lex. Incredibly thoughtful discussion of how they thought about their vision for coding going forward. And so I think it comes back to, Martine, something you said, which is they were just reimagining everything. Right. And you just talked about this as well, Matt. But you can't you can't write growth memos that way. Never mind you. So, you know, you got the bug, you know, because you were calling us and we were so excited. How did you convince David George? Like, don't you have to show like, you know. Yeah. You know, it's so funny because and I think you had to think out of the box for this one. I mean, of course, the momentum was vertical, but there were so many things that they defied conventional thinking where I remember I the first time, you know, Martin, you introduced me to Michael. We got a coffee that day, you know, as Michael, you know, they move quickly. So I pulled some stuff together for them Right Because we you know we trying to show them hey this is the next stage of what growth looks like And there so many conventional pieces of conventional wisdom that they just completely broke. Like, we would say, hey, we'll help you find a sales leader. Because, you know, previously, it made sense to start layering a sales leader at 25 to 50 million of ARR when self-serve starts to peter out. Michael looks me dead in the eye and is like, self-serve is not petering out, right? And so then you start to pull these drivers that you're like, oh, these assumptions that would go into any growth model historically, you can't have that asymptoting at 25% growth in five years, right? And so I think a lot of the growth investment decision at the time was throwing out some of the assumptions that we would typically make on how a company would, you know, eventually start to slow growth. I'm very glad that we did that. And then the second piece is we should talk about the competitive landscape, right? Because you mentioned Copilot. I was actually going to say, and actually, I think Matt deserves a ton of credit here, which is we talk about now in retrospect, how it was a pretty easy decision on the Series A. But you know Matt led the early deal and he saw something a lot of people didn which was this which is if you looked at the competitive landscape it was almost silly which is you had a large incumbent which is Microsoft that had a company at scale that had the opening I wait and they owned VS Code So everything, you know, everything would say that there's just no way that you could survive relative to this one single competitor. We'll talk about the other ones later. And I think that also was the thing you said about Michael was what I witnessed. It would be great to hear Matt, which I even remember when Matt said this early on, he was like, you know, they have a good answer for everything. When you have founders that are students of what they do, they're very consistent in their answers. And, you know, and this is not, I think when we did the deal, it was pretty obvious, but it wasn't so obvious six months before. You know, and Matt and the team had that deep conviction way back then. And I think for every one of these races, they're mirrors of that. Anyway, is this guy? Yeah, no, totally. I mean, people think venture capital is like an individual sport. Like it's not at all, at all. I mean, our whole team was working on this at the time. I think it was over a year. We literally had a cursor team on our team. I remember that, yeah. Marco was great, Rachel was great. Yeah, yeah, yeah. You led the team I mean it was great But yeah no you exactly right I mean these guys are crazy smart They off the charts just raw intelligence They're able to do this thing we talk about a lot, like context switch between tech and markets really, really seamlessly. And they ingest data, whether it's historical data or things happening in the market and act very, very quickly, which you said before, Sarah. I agree with that, but specifically, it was just kind of crazy to think you're going to fork something from Microsoft who owns the thing, and then you're going to compete with them with smaller models, even though they own the opening eye weights, and they have enterprise, Salesforce, and they have… The greatest enterprise distribution. Yeah, ever. You know, it's so interesting, right? It's like… They have GitHub. Yeah, they have GitHub. Just 100 million developers. They literally own every piece of that. You know, it's so interesting, right? Because that's definitely the conventional wisdom, and it's usually correct, the conventional wisdom, in many of these cases. One of the data points at the time was actually Marco and our team set up a dinner where Amon sat right next to John Schulman, and they talked a lot about Cursor. And I think John's thing was like, oh yeah, I tried
[15:00] cursor, but our code base, I think he was at Anthropic at the time, our monorepo just immediately killed it. I couldn't even load the code base. I think the next day, Amon came back and was like, oh, we made some changes. And not only is it speed of iteration, but like, We just look for these little data points where it's like, actually doing a plugin or just writing on top of VS Code is clearly not going to work because the smartest people in the world have tried it and it doesn't work. So these guys are sort of listening to it. have always... Bet on the independent. in a large market, right? If there's a leading independent, we'll always bet even if the incumbent seems very scary. I think it's worth talking about competition because my experience with all of our experience with Cursor is that there was always a looming existential threat. And we always forget about it, but of course, Copilot was the big one. And then remember the whole Windsurf thing? So Windsurf actually did this very smart thing where they went to YC and there was a lot of YC users using it. So you kind of had this kind of whole movement on Windsurf. company actually was making real waves with agents and they've done a great job of that. And then you had the first version of Cloud Code that came out and like that was kind of like, you know, a competitive thing. And so like you kind of go through like there was always some competition. And one thing I would say, and I would love again to actually hear your perspective on is like it is such a pleasure to work with founders that are entirely unfazed. You know, like normally, like, you know, founders of this thing's happening are calling and they're kind of freaking out and whatever. They were 100% unfazed by the competition. Yeah, paranoid, but unfazed. I'll never forget, you know, and we've been lucky to do... I think all of their rounds, starting from the A. Is that true? Did we invest in every cursor round? We did, yeah, with increasing amounts of dollars. Sarah co-led the B and the C. That's amazing. That's real conviction. That's amazing. That was a harder conversation to convince the team for the C. You know, at the B, you could argue... God, just bet on vertical momentum. That was actually misogamous. Who grows from 4 to 50 in like 4 months or whatever. But the... or actually no, it was the D. So the Cloud Code, not to take us on a time travel trip again, but Cloud Code came out May 2025 with a splash. You know, it wasn't, the pickup wasn't right away, but obviously formidable. And I remember there was around, I think a couple months after, three months after that. And I asked Michael, I was like, what do you think about Claude Cote. And he said something to the extent, actually, you know, exactly what you're saying here, which is, look, we are going after the biggest market in the world. there's going to be competitors. In fact, there's been a rotating cast of characters from the very beginning. Microsoft Copilot was the first one. Cloud Code is one of a rotating set of characters. And we think in big markets, you're always going to have formidable competitors. That does not scare us. And I just remember being... Yeah, somewhat blown away by the clarity of that comment, the lack of fear, right? There's humility as well. It's humility mixed with bravado. And I just think that is kind of the winning... combo when you're going after these juggernauts like that. And I know it sounds vague, but like things like this, it just it's almost a Bezos-ish thing. It's like if you're smart and right, It actually gives you, I think, the confidence and the focus to have that kind of attitude when you have this mega competitor sort of coming after you. Yeah, that I'll say I will say they were definitely a couple of oh fuck moments, but they were always around models. Remember when Opus 4.5 came out? We're all on vacation, and I think everybody in the world collapsed. We were like, oh, fuck. That's true. These things are moving much, much faster than any of us thought. And then I think the same thing when the pre-release of Mythos came out and people had access to it. These models were getting so good so fast. And I think another thing that the cursor team probably doesn't get is... to the extent they cannibalized themselves so dramatically in like a Reed Hastings type way in a matter of two years, right? So they were first an IDE and then they moved to an agent platform, then they moved to a model platform. And it's very, very tough for, you know, founding teams to actually have the discipline. Yeah, tab was the big thing. You know, Karpathy had that famous post, I guess he has a lot of famous posts where he's like, tab, tab, tab. Yeah, exactly. Like that kind of doesn't matter that much anymore. Yeah. segue to maybe some of the other strategic decisions they made along the way around business model. And to your point on, Oh fuck moments. I don't know if they would consider this that because I don't think the vagaries of Twitter... impact.
[20:00] them and as much. As the investors. Yeah, as the rest of us just scrolling all day. But it was probably summer of 25 where there was a lot of shit written about their gross margins, et cetera. And then, of course, post the Opus 4-5, maybe the January period. I mean, we should all talk about that because those are both different but interesting times. Yeah. I mean, one thing I'll say is I just feel like X is driven by like tail chasing, adrenaline sinking in schadenfreude. And, and, there tends to be a massive disconnect by what people say, on X and like how the business is actually doing. And so even though there was a lot of this, the business tended to do very well. And it's so interesting is so much of this often comes from investors, especially things talking about like business quality. But tech transformations always proceed figuring out the business. Like you guys remember the internet, like we didn't even figure out how to monetize it for a long time. Forget margins, making any money at all, right? And here you've got like unlimited demand, unlimited growth and margins. So I think this you know, a exit its own kind of echo chamber, as people say, but also be every single wave you have investors trying to say something that sounds smart, which is like picking apart, you know, the business model of a nascent tech transformation. And I just think historically, they've always been proven wrong. Yeah, for sure. I think I'm going to borrow this quote from something you said, Martine, but First comes the technical transformation, then comes the business model. And in this case, I think the strategy where they went and got all the users... in record time. Now they have the assets... the data, the know-how, etc., to build their own models, right? You really couldn't, do the flip of that unless you started as a frontier lab. And it It's clearly paying off in this combo that we'll talk about later, but I think seeing that backdoor strategy was really inspiring, and they took a lot of heat for it. So, Martine, how do you think Cursor really became a sales driven company or a company that's good at sales? Because, you know, early on, I think I told Michael a few times, too, it's like, oh, you better invest in sales. He's like, nope, I'm not going to do that. They literally set up enterprise at cursor.com. It's like, if you're an enterprise, send your inquiries to this. But like, but obviously now they're very good at sales, right? And, and, and, or, you know, I'm not going to do that. go to market in generally? How do you think they made that jump? It's a great question. So here'd be my best guess, which is they are very much a student of what they do. They're very thoughtful. Every raise, every product decision. And I think as they learn more about the market, they realize that the margins were in the enterprise. So very classically, if you view your competitive set, say the large labs, the large labs will basically subsidize or give away the single-cert tier. But then they actually have, they command great margins in third party and they command great margins in the enterprise. And, you know, of course, there's really a scale. And as the company has always gone, they decided, okay, like, so, you know, for future viability, we need to go into the enterprise. That's where, A, the bulk of the spend is, but also a lot of the margin. The margin tends to be the value here. So, you know, this is existential and this is important. And then when they decided to do it, they really went in. And that may be the fastest crowd on the sales team in the history of the planet, right? I mean, you know, you were very involved, Sarah, for example, on recruiting. So maybe talk about that actual expansion because it was... Yeah, I mean, it's so neat to see, you know, that expression, how you do anything is how you do everything. and how they thought about product, et cetera, was reflected in how they did hiring. And go-to-market hiring in particular. And we got to shout out Jordan and Roman and their team there, who in the early days built that. from that engine from scratch. They're just the most incredible, you know, some of the most incredible people I've ever worked with. But we've all ever worked with. And well, they also spend 40 percent of the time recruiting. Yeah, exactly. This is amazing for technical product founders to spend 40% of the time. I've never actually seen it before. Absolutely. And it is so effective. And like a company really is the team that you build. And they took this to heart. Like everybody talks about it. Everybody talks about like doing A-class founders, whatever. But it's entirely different. It's like how often was Oscar, you know, hey, I'm on a red eye to Europe to close. So I'm like literally like IC engineer. It was constant. Totally. about their hiring practices that everyone should read, a lot of that focuses on their edge hiring, like the work study, all this stuff, right? But even on the go-to-market hiring, when the thought that they would put into hiring a single AE, right? And of course, like this is nothing new, but just the...
[25:00] back channel, back channel, back channel ethos, they had that to a T. They knew exactly what they were looking for in founding AEs. I remember we would go there and sit in their office for eight hours helping them source stuff like that. And they weren't like... "Oh, we want to do it the old way." We're like, that's not going to, we're building a company in a new way, right? So we want very specific AEs that fit this profile. We're going to go get them. We're not just going to wait for people to walk in the door. All of those things that have been written a bit on the end side, they did that on the go-to-market side and clearly to phenomenal results. I mean, they got to over 50% of the Fortune 500 and I think faster than anyone we've ever seen. Can you talk about specifically how they... applied the lessons in end hiring to like AE hiring? Because this is something I think our founders deal with a lot. Like, you know, I, you know, I've dealt with this a lot personally. Like, how do you actually make that jump? And like, what lessons really do carry over? Yeah. Or that you observe in person? No, yeah. And, you know, hopefully, I think this has been written about. So hopefully it's not sharing alpha. But the way that they think about sort of epochs of a person's career or of a company and who was the key dial in making that happen. I love this intense research period that goes on where it's like, hey, these are the top 10 companies who have seen great in this type of selling. And then let's go a level deeper. Here are the top 10 teams in those companies. And then here are the one at number one and number two single AEs who drove that change and accounted, you know, Again, back channel, back channel, back channel, right? Like who their boss, their boss's boss and their boss's boss's boss would say, this is a superstar AE. And of course, in the early days, it was like, can they deal with some uncertainty? Because you don't want someone who can just run a playbook. At this point, maybe they can get those people. But in the early days, right, the founding AE set was super strong. And I think figuring out what does the best look like at what period of time and then which individuals drove that, that's so applicable for engineering, obviously product building, but on the go to market side as well. And I think that nailing. The right star who did it at the right time is really hard. I mean, it narrows the base that you can hire from completely. But because, you know, I think Jordan Roman and the early team got that right. They just were able to scale it quickly without losing the cursor culture. Which, you know, a lot's been written about the cursor culture. I'm curious, you know, maybe you guys can share your own impression of it. You know, I mentioned earlier that like, you know, kind of, coding was all that they did. My impression at the time, and the founders could correct us on this, my impression at the time is they weren't doing that because they were obsessed with work. And I think they're still not doing it because they're obsessed with work. They literally just love writing code. and love systems. And like, I think the three of us sort of, you know, can feel that way at times too, right? Like that's like a very, very understandable thing. And that seems to be the driving cultural thing. you know, kind of impetus at the time. It was a very small team. Everybody had that attitude and like we're orienting towards the future of like, what is this actually going to be? And that kind of like, you know, from an edge standpoint or product standpoint, like very narrowly focused on what they were trying to accomplish and like how to run the team. But there's a huge paradox because I totally agree. I 100% agree. However, if you actually detail their actions later on, that's not the conclusion you get to. Like, what do they spend their time on? It was like business models, biz dev. And so in a way, again, I just totally agree. And again, on one hand, you have a very product-focused team that comes from the tech that literally wanted to build products. that they wanted to use. So listen, I think that high-powered founding teams work on exactly what is needed by the business, and I think they did a very good job, and I think it allowed them to navigate the most competitive market we've probably ever seen. Has there been a more competitive market than coding? I mean, it's just hard to imagine. There's so many companies. Database wars? Yeah, I mean, you have very intense competition in relatively narrow markets. I mean, databases are huge, but compared to coding, I actually think it's much smaller, right? I mean, coding is kind of the meta category, right? It's like all expression of human thought in economic form is kind of like coding. I mean, imagine starting competing with Microsoft, ending up competing with, say, Anthropic, both of these kind of mega companies with tremendous amount of capital and power and distribution. In the meantime, you evolve fundamentally your product approach, and you change your go-to-market, a matter of two years. I think this is the magnum opus of startups. No, totally. There's definitely something to what you're saying. In the beginning, these guys were incredibly focused on the product and literally enriching their own lives.
[30:00] Literally, they would be coding for fun. Like when I finally got them to admit they did something for fun, it was like, oh, that's also coding. Yeah, which by the way is the only fun left to any of us, by the way. Now that the models have brought the post-scarcity future to us, it's like coding is the last bastion. But there was always awareness, or at least I always thought, talking to Michael, he was always aware of the game, if that makes sense. Truly was, and they were all truly in love with coding and creating great products. But like they kind of knew... that building an enduring company or a great company isn't just about product obsession. I just got the sense he always kind of understood that and was thinking a couple of steps. Yeah, if your goal is changing software, that's a critical... It's critical to get the business straight. Yeah. Yeah. I was going to add an anecdote that's not related to, um, them coding. But the other thing that, you know, we've sort of seen across a number of founders, but Michael and Soale and Amman, they sort of embodied this, is this Again, how you do anything is how you do everything, right? And so, of course, this detailed attention to the product, very artisan. There's craftsmanship, right? Overused taste. They have taste in spades. But it extended to how... Not only they did hiring on the end, go to market side, but even the space they created, the events they throw. Like I remember walking into their office and... thinking, wow, in my early 20s, I would not have paid attention to putting twinkle lights here. Famously, no shoes, right? But you'd go in, you'd put on your slippers, and it would feel... you feel at home. Hence why everyone was there working all the time because you're like, this feels great like home. And it was very Hege, the Scandinavian style. And I remember going to their second office when they finally expanded and complimenting them on a couch or something because I think I was redoing my own house at the time. And they were like, oh yeah, we got that from a... secondhand like very Hege furniture dealer. And so they were going after a specific look and feel. And I just was struck by that because I was like, wow, that is not how I remember when guys in their early 20s, like thinking about the space that you create or are in, how that leads to the product and mind space you're in. Yeah, there's actually a nice adjacency here to a lot of the decisions they made. So marketing was a very obvious one. They were very, very particular. We use X, that's not our voice, we're not spammy. We made so many suggestions. I remember we're like, oh, let's do this event. No, that's not our style. So they're very good at saying no for stuff. I will say that they were also very good about this when it came to personnel. make the hard decision pretty quickly. You very, very rarely see this with early phonic teams. And they're very consistent about that. And so I would say it's almost this kind of untalked about Probably greatest strength of the company is that once they made a decision, it doesn't matter how hard that was, whether it's deciding not to enter a market, deciding to be very limited in what they're doing, they kind of follow through or who would stay in the company, who wouldn't stay in the company. Yeah. Kudos to them for actually having the courage and the tenacity for these tough decisions. Which is sort of an Elon-ish trait, too. It's also very Elon-ish, isn't it? Yeah. I actually think that these two companies, I mean, we just talked about the M&A a little bit. are very good fits. They're very good fits on... On like a technical level, right? Like Elon has a compute, they have the distribution and the data. On a vision level, they both think that, you know, code is the path to changing all of compute and maybe all of humanity. But also on a culture level, right? They're very product engineering focused. Very, very, very. Very product engineering heavy, very fast iterating. You know, they make hard decisions very quickly. And so, I mean, I've been involved in a lot of M&A. And I would say this is probably the best fit I've ever seen as far as just the full package. Yeah, I couldn't agree more. I mean, there's a... Funny analogy with our experience on the SpaceX side, I remember we met them in 2019. Back then, it was fully a launch business, but the SpaceX team was like, no, we're going to launch global communications. We're going to power the internet for everyone in the world. They had nothing at the time, no Starlink satellites in the sky. And of course, they did it. And if you think about the cursor team, telling us, and, you know, they actually beat every forecast they ever gave us. Oh, by a lot. There's just so many Elon-ish characteristics of moving fast, going after large markets, winning them in the feet of impossible competitors, you know, Comcast, right, in the Starlink case, that there's so many parallels, totally different venues, obviously, but fascinating to see that. There's another thing I would say also, which I think between kind of Elon and Cursor, Thank you. In both cases, the consensus
[35:00] was decrying the death of both companies so many times. I think there's also this kind of like, you know, cultural perseverance to the naysayers, which I actually really respect on both sides. Actually, I do have one other real question I wanted to ask you guys. Cursor is also really good at doing M&A, like doing small talk. I was actually going to mention that. Yeah, I would love to know a little bit more from your perspectives, like why they were and what the philosophy was and all that. So I will say, listen, so there is such a war for talent. um, right now that the companies that have figured out how to do, uh, talent via acquisition can really get ahead. I mean, there's just a lot of advantage there. And so the early motions that they did were of, of that nature, which is like really good team, you know, they'll just go ahead and do the acquisition. And what was really nice about cursor, they were just never worried about like operational complexity. They're just like, it doesn't matter how messy it is. We'll too much about their ability to set culture because they were just so good at it. Like, again, like that was just part of it. It wasn't until later, like the graphite acquisition, where they started thinking about acquisitions in terms of strategic direction. And of course, that was a much larger acquisition. And, you know, listen, I think... Um, The company was so fast. Like... Who knows to the extent that they would have figured out how to integrate all of these things and execute on them in the limit? Just because now they're part of a very large acquisition. But I will say that they approached these in the same way that they approached everything, which is incredibly thoughtful. They had a plan. They executed quite well. And so this may be the most concentrated company in the history of companies having done M&A and everything else. CEO of a company says, I have X founders in my company. And, I think it should be a badge of honor, right? I don't know how many, what that stat is for Cursor. Do you guys know? But I remember, and they have so many great founders there, but I remember when they brought on Tito from Koala and then Adam Ward on the talent side, actually, and just seeing firsthand the leverage that they got from bringing in someone who was a founder, CEO, effectively, but slotted really well into the Cursor culture and just take that to great heights and add to the acceleration. I think that lay book, you know, I don't know. I'm sure they're not the ones who started it per se in this new Gen.A.I. era, but they certainly are... I think one of the best examples of how to integrate founders really, really well into your company. It's funny, right? Because Anthropic kind of gets that tag a lot now. It's like, oh, like, exec at company A just went to take an IC role at Anthropic. But, like, Cursor kind of did this first, and they weren't just, like, slotting people into IC roles. Like, they're actually building the company this way, which I thought. It's very unconventional, but, like, really hard. It's actually operation incredibly complex. Yeah. That's a conventional wisdom. Yeah, yeah, and I think it's actually true. did something that people think is hard, but is it actually easy? I think it's actually hard, and they just managed to pull it off. Thanks again for listening, and I'll see you in the next episode. and is not directed at any investors or potential investors in any A16Z fund. Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see a16z.com forward slash disclosures.
