Cat Wu, whose X profile reads “claude code + cowork @anthropicai”, said in a post on X on 3 November 2026 that Wu has left Anthropic. “Friday was my last day at Anthropic,” Wu wrote at 21:33 UTC; the Friday before the post was 30 October. Wu gave no reason for leaving and did not name a next role.
Wu follows Thariq Shihipar and Anthony Morris, who both posted on 24 October that they had left Anthropic after working on Claude Code. By our arithmetic from Anthropic’s S-1/A, filed on 2 November, the company’s quarterly revenue grew about 45x over the period, from $376 million in the three months to 31 March 2025, the quarter Claude Code entered research preview, to $16,956 million in the three months to 30 September 2026. The filing gives no revenue figure for Claude Code itself.
Business Pill · The Model and the Harness
A short Business Engineer explainer on the difference between an AI model and the harness around it. It teaches the general idea only and says nothing about the people in this piece.
The key insight: By our arithmetic from Anthropic’s S-1/A, quarterly revenue grew about 45x, from $376 million in the quarter Claude Code entered research preview to $16,956 million in the three months to 30 September 2026; the filing gives no revenue figure for Claude Code itself.
Wu described how the product looked at the start. “Two years ago, when Claude Code was a small internal tool, I asked it to build an RL env,” Wu wrote. “I steered it for tens of hours but never manually wrote a line of code. It worked, I was hooked, and I joined the team.”
Wu then described a change in how people use it. “When we first launched, most people approved every permission prompt by hand and mostly used Claude to answer questions and write tests,” Wu wrote. “Today, the default is trusting Claude’s judgment to auto-approve. Claude does large migrations and builds new products.”
“Helping build Claude Code into an industry-defining product has been the journey of a lifetime,” Wu wrote, and thanked “the incredible team behind Claude Code, Cowork, and Tag” and the product’s users. The post ends: “Excited to see what you all make next!”
Anthropic revenue by quarter, in millions of dollars, from the three months to 31 March 2025, the quarter Claude Code entered research preview, to the three months to 30 September 2026. The 45x is our arithmetic ($16,956M / $376M) and covers all of Anthropic; the filing gives no Claude Code revenue figure. Source: Anthropic S-1/A, filed 2 November 2026.Animated: Anthropic revenue per quarter, Q1 2025 ($376 million, the quarter Claude Code entered research preview) to Q3 2026 ($16,956 million). Source: Anthropic S-1/A, filed 2 November 2026; the 45x is our arithmetic and covers all of Anthropic, not Claude Code alone.
What Was Posted on X
Friday was my last day at Anthropic.
Two years ago, when Claude Code was a small internal tool, I asked it to build an RL env. I steered it for tens of hours but never manually wrote a line of code. It worked, I was hooked, and I joined the team.
When we first launched, most people approved every permission prompt by hand and mostly used Claude to answer questions and write tests.ย
Today, the default is trusting Claude’s judgment to auto-approve. Claude does large migrations and builds new products.ย
Helping build Claude Code into an industry-defining product has been the journey of a lifetime.ย
Thank you to the incredible team behind Claude Code, Cowork, and Tag. These products are the sum of thousands of great ideas from you. I’ll miss working with you all.
Thank you to our users for embracing a new way of working. Your constant feedback shaped so much of what we shipped and your stories made the work so meaningful.
Cat Wu (@_catwu) posted this on X at 21:33 UTC on 3 November 2026. It is the primary source for this piece: Wu writes that Friday was Wu’s last day at Anthropic and describes Claude Code’s shift from hand-approved permission prompts to auto-approval as the default.
Iโve left Anthropic after more than a year of working on Claude Code and seeing it scale larger than I ever imagined.
I have really loved working at Ant and believe deeply in the mission, and most of all, the idea of a race to the top.
The idea that if we work hard and responsibly, we can raise the bar and inspire others to great outcomes. And if we all push each other to be better, things can be radically good. I will carry the spirit of this with me forever.
My colleagues at Anthropic are the most talented, thoughtful, competent, and earnest people I have ever met. I am glad they are at the wheel and trust them deeply.
To the community on X, thank you for everything. Youโve made this the most fun (and sometimes stressful) job Iโve ever had.
For me, it is time for a bit of a break before diving into something new.
If you want more excellent Claude Code content, you should follow @addyosmani , @lydiahallie, @_catwu, @ClaudeDevs and of course @bcherny who started it all.
Thariq Shihipar (@trq212) posted this on X at 04:04 UTC on 24 October 2026, saying Shihipar had left Anthropic after more than a year on Claude Code. The post ends by recommending five accounts for Claude Code content, including @_catwu.
Since Cognition bought Windsurf, 87 people have walked out. That’s 80% of every Cognition departure I could find.
I pulled every LinkedIn-visible move in and out of Cognition since Oct 2023 (all roles, basically the company’s whole life) using Crustdata (YC F24).
The tl;dr is that Cognition is still hiring about 4x faster than it’s losing people (475 hires vs 109 departures). But the exits spiked after the Windsurf deal, and the leavers are going to other AI companies.
4 patterns to watch:
1. Windsurf was the biggest feeder, and the messiest
38 Windsurf/Codeium alumni moved onto Cognition’s page, more than Scale (20) and Amazon (14) combined. But 44 people who left Windsurf since July 2025 now list Google or Google DeepMind as their employer. Another 51 still list Windsurf itself.
2. The leavers go to AI companies, not Big Tech
OpenAI and Parallel Web Systems took 6 people each. Factory, which also builds AI coding agents, took 5. Only 3 went to Google and 3 to Anthropic.
3. Cognition wins the Cursor matchup 3 to 1
6 people came from Cursor. 2 went the other way.
4. Most exits were short stints
In a 45-person sample of leavers, 38 had been at Cognition for 12 months or less.
And it’s not just engineers moving. In my samples, roughly 60% of both hires and leavers held non-engineering roles (GTM, sales, partnerships and so on).
My read here is that the Windsurf deal bought Cognition a team and a churn problem at the same time. Hiring still outpaces exits about 4 to 1, so this looks more like a company scaling fast than an exodus. What stands out is where people go next: other AI companies, not back to Big Tech.
Disclaimer: this analysis was based on LinkedIn-visible moves only, all roles (hires exclude interns, advisors and investors). The numbers are small. Treat it as directional.
@Doudoujay posted this at 17:00 UTC on 3 November: a count of LinkedIn-visible moves in and out of Cognition since October 2023, pulled with the data vendor Crustdata. By the poster’s count, 87 people have left since Cognition bought Windsurf, 80% of the departures found, while Cognition made 475 hires against 109 departures. The poster calls the numbers small and directional; we have not verified them.
Earlier Exits From the Claude Code Team
Thariq Shihipar posted on X at 04:04 UTC on 24 October: “I’ve left Anthropic after more than a year of working on Claude Code and seeing it scale larger than I ever imagined.” Shihipar wrote that it was “time for a bit of a break before diving into something new.” Shihipar’s X profile, read on 3 November, now says “formerly Claude Code @anthropicai”.
That post ended by pointing readers to five accounts for Claude Code content, @_catwu among them. As read on 3 November, the X profiles of Addy Osmani and Lydia Hallie, two others on that list, still name Claude Code at Anthropic.
Anthony Morris, whose X profile says Morris “created claude code on desktop”, posted at 12:33 UTC the same day: “Yesterday was my last day at Anthropic.” Morris wrote: “I’ll share what’s next soon!”
None of the three posts gives a reason for leaving. Anthropic has not commented on the departures in anything we read.
What the Filing Says About Claude Code
Anthropic’s S-1/A calls Claude Code “our coding agent and the product we believe has most concretely defined what autonomous AI work looks like.” It says Claude Code “has progressed from suggesting code to writing and running it directly and working autonomously over long time horizons on complex, multi-step engineering tasks.”
The filing links the product to customer spending. In the three months to 30 September 2026, Claude Enterprise users who did not use Claude Code spent less than $75 per month on average, while the top 5% of Claude Enterprise users who used Claude Code spent approximately 20 times that amount, it says. It also says adoption of Claude Cowork, released as a research preview in January 2026, outpaced the adoption Claude Code achieved at the same point after its own research preview in February 2025.
Inside Anthropic, the filing says over 90% of the contributions to its codebase in September 2026 were written by Claude, and that average daily code volume per engineer in the September quarter was over 45 times the 2024 average.
The S-1/A names Claude Code 50 times. None of those passages, as we read them, gives a revenue or run-rate figure for the product, so the 45x in this piece is company-wide revenue, not Claude Code’s.
The Market Around It
Anthropic’s Claude Code and OpenAI’s Codex together account for $24.0 billion, or 70%, of a $34.0 billion enterprise AI coding market, Menlo Ventures estimates in 2026: The State of Enterprise AI, published on 3 November, as we reported in Menlo: Claude Code and Codex Now Take 70% of $34B AI Coding. Menlo’s figures are survey-based estimates of annualized run rate for the two products combined, and Menlo Ventures is an investor in Anthropic.
The business lever in Wu’s post is default autonomy. Wu describes Claude Code moving from hand-approved permission prompts to auto-approval as the default, and the S-1/A ties higher Claude Enterprise spending to the users of Claude Code. As we read the two together, how much the tool is allowed to do on its own is part of what customers pay for.
Update, 3 Nov: a LinkedIn tally of departures at Cognition
Departures are being counted at another coding-agent company too. In a post on X at 17:00 UTC on 3 November, @Doudoujay, whose profile name is Jay Ma, published a tally of LinkedIn-visible moves in and out of Cognition, the maker of Devin, since October 2023, pulled with the data vendor Crustdata. By that count, 87 people have left since Cognition bought Windsurf, 80% of the 109 departures found, while Cognition made 475 hires over the period.
The tally has the leavers going mostly to other AI companies: OpenAI and Parallel Web Systems took six people each and Factory five, with three each to Google and Anthropic. The post calls the numbers small and says to treat them as directional. They cover LinkedIn-visible moves only, and we have not verified them.
Hires and departures at Cognition since October 2023, as counted from LinkedIn-visible moves by @Doudoujay with Crustdata data (not verified). Data: Crustdata via @doudoujay, 3 November 2026. The 22 earlier departures are our arithmetic.
Cognition announced its agreement to acquire Windsurf on 14 July 2025. Chief executive Scott Wu wrote that the deal included “$82M of ARR” and that “100% of Windsurf employees will participate financially in this deal”. On 8 September 2026 Cognition said it had raised over $2 billion at a $48 billion valuation; Business Insider reported on 28 October that it had since received offers at $70 billion to $80 billion. Elsewhere in the talent thread, DeepMind staff are pushing Google to drop a post-exit criticism ban.
Connecting the Coverage
Our related pieces on Claude Code and Anthropic’s filing:
As we read Wu’s post, the change Wu chose to highlight is about permission, not model capability: from approving every prompt by hand to auto-approval as the default.
As we read the S-1/A, the company presents Claude Code as its clearest product example and ties higher Claude Enterprise spending to its users, but reports revenue only for Anthropic as a whole.
As we read the three posts together, each person who left thanked the team and the users, and none gave a reason.
Cat Wu on X, 3 November 2026
“Helping build Claude Code into an industry-defining product has been the journey of a lifetime.”
Three Implications
FOR CLAUDE CODE USERS Wu’s post describes today’s default as trusting Claude’s judgment to auto-approve, rather than approving each permission prompt by hand.
FOR READERS OF THE S-1/A The filing reports company revenue, enterprise spending patterns and adoption comparisons for Claude Code, but no productrevenue line.
FOR THE CODING MARKET Menlo estimates Claude Code and Codex together at 70% of enterprise AI coding spend, while Revolut reports moving its coding work to self-hosted open models.
The framework puts it this way: “The surface, not the model, governs autonomy.”
As we read it, Wu’s post is an example: the shift Wu describes, from approving each permission prompt by hand to auto-approval as the default, is a change in the product surface. The framework does not discuss Wu, Claude Code’s permission settings or these departures.
What Is Not Established
Wu, Shihipar and Morris did not say why they left, and none named a next employer. We do not know whether the departures are connected, and we do not draw a link to Anthropic’s planned listing. Wu’s post does not state a job title; we describe Wu by the X profile and the post.
The 45x is our arithmetic from the S-1/A’s quarterly revenue ($16,956 million divided by $376 million) and covers all of Anthropic, not Claude Code alone. Menlo’s market figures are estimates that combine Claude Code with Codex. We did not contact Wu, Shihipar, Morris or Anthropic.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Cat Wu, whose X profile lists Claude Code and Cowork at Anthropic, said in a post on 3 November, ten days after Thariq Shihipar and Anthony Morris posted their own exits, that Friday 30 October was Wu’s last day. Over the period from Claude Code’s research-preview quarter to the latest quarter, Anthropic’s quarterly revenue grew about 45x, by our arithmetic from the S-1/A; the filing does not say how much of it is Claude Code.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. This piece is based on Cat Wu’s own post on X of 3 November 2026, the posts by Thariq Shihipar and Anthony Morris of 24 October, X profiles read on 3 November, and Anthropic’s S-1/A filed with the SEC on 2 November 2026. The 45x is our arithmetic from the filing’s quarterly revenue. We did not contact Wu, Shihipar, Morris or Anthropic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
