The CTO's Next Career — Why Unicorn Founders Move to AI Labs, and the 9-Step Exit That Makes It Possible
How to start a company, how to raise funding, how to build products, how to scale teams. There are thousands of articles on each. Yet articles on how to resign and hand things off are almost nowhere to be found.
Henry Shi’s resignation essay, “I Left My $200M+ Rev/yr Startup to Build Safe AGI at Anthropic,” tries to fill that gap. Shi co-founded Super.com, an all-in-one app spanning travel, cashback, and more, and over eight years grew it into a profitable business with $200M+ in annual revenue, 50M+ users, and $150M raised, serving as CTO and COO. Then he left to build safe AGI at Anthropic. His essay lays out the resignation process he designed over nine months as a 9-step playbook.
I am myself a hands-on CTO who still writes code, working in a fully remote organization. So this essay hit close to home in two ways. One is the question of where a CTO’s next career actually lies. The other is the question of how, exactly, you are supposed to let go of a heavy role. In this post I work through both, using the essay as my text.
Why founders and CTOs are moving to AI labs now
The essay opens with Shi noting that he is not an isolated case. A wave of founders is moving to frontier AI labs, and the momentum is only building. Unicorn founders, he says, reach out to him privately and constantly, asking how to make the move.
Why is the trend accelerating? His framing is simple. Instead of building AI wrappers on top of external APIs, you get to shape the foundation models themselves from the inside and work on fundamental AI safety challenges. The scale of impact is in a different league.
Underneath this is a founder’s response to a tectonic shift: AI is consuming SaaS. Businesses that differentiated at the application layer by calling a model’s API are easily undercut as the model itself advances. Better, then, to move to the side that builds the foundation models, which is where the value originates. From depending on the layer below to owning it. That is, for a CTO, a fundamental question about which layer of the stack to compete on.
The value of an operator, as labs see it
What is interesting is that Shi also reads this migration from the lab’s side. These founders, he argues, become rare hires for a lab.
Pure researchers, however brilliant, have not necessarily lived through the full stack of bringing a new product to market. Navigating the messy reality of product-market fit, scaling a team from 0 to 100+, managing complex stakeholder dynamics, and shipping under intense pressure: that experience, he writes, cannot be taught in a classroom or a research lab. Depth in research and breadth in operations: when the two mesh, you get a rare combination that benefits both the lab and the person.
I took this point seriously from the angle of a CTO’s career capital. Being able to write code and draw an architecture is merely the baseline. On top of that sits a track record of making the decisions that move a business forward (hiring, org design, funding, customers, shipping calls), under fire, again and again. That experience of having gone through the entire stack is the core of an operator-CTO’s value, and it is hard to substitute with a researcher’s purity. An AI lab is simply a legible destination; the value itself is not confined to labs.
A clean exit is the foundation of the next career
So how do you execute the move? The body of the essay is devoted to exactly this. Shi is candid that the decision was hard, but the execution was even harder. Most founders, he writes, never think about their exit until the moment arrives, and by then it is too late to do it right.
The nine steps he worked through over nine months can be summarized like this.
| Step | What | Purpose |
|---|---|---|
| 1. Write your story down | Before telling anyone, document why you started, what you accomplished, what the company became | Own the narrative yourself |
| 2. Co-founder, privately first | Align fully on successor, narrative, and timeline, just the two of you | Get in step before word leaks |
| 3. Board members, 1:1 | Win buy-in for the successor individually, not in a group | Avoid the politics of group dynamics |
| 4. The official board meeting | Hold it only once everyone has already agreed; make it a formality | Make the meeting a surprise-free confirmation |
| 5. Email investors with a narrative | Not a boilerplate notice: story, accomplishments, succession, vision, gratitude | Cement your legacy in memory |
| 6. C-suite and key people, 1:1 | Talk to 15+ people individually within a 24-48 hour window | Preserve trust and transparency |
| 7. The company-wide announcement | Put the successor on stage, paint the future, take questions, hold office hours | Leave everyone informed, respected, hopeful |
| 8. Use the window before your last day | Push bold moves and tee up wins for your successor; document everything | Build the bridge with zero-consequence authority |
| 9. Make the public post count | Tell the story with numbers, credit the team, end with a clear CTA | Open the next door yourself |
Let me bundle the design philosophy running through all nine into three ideas.
First, write your own story down
The first step, Shi says, is the one most founders skip. Before you tell anyone you are leaving, sit down and write your complete story: why you started, what you accomplished, whether your motivation still aligns with where the company is, and what the company became.
He spent three weeks on this, going back through old emails, board decks, and Slack to reconstruct decisions he had completely forgotten. Memory fades faster than you think, other people’s memories are even fuzzier, and left alone the gaps get filled with everyone’s own interpretation. So you write it before anyone else writes it for you. Beyond resignation, this strikes me as a universal practice for how to give meaning to any inflection point in your career.
Tell people one at a time, in order
What steps 2, 3, and 6 share is a discipline: tell the people who matter individually, never in a group, and in the right order. The co-founder first: the conversation Shi broached on a Christmas call, one of the hardest of his life. Board members always one-on-one, because in a group they perform for each other. The C-suite and key people packed into 30-minute 1:1s with 15+ people inside a tight 24-48 hour window, to prevent leaks and chaos.
What works here is the idea of signaling respect proportional to the weight of the news through the design of how you deliver it. Nothing destroys trust faster, he writes, than hearing major news in a group setting where you can neither ask questions nor process the emotions. The accumulation of 1:1s costs time and energy, but that investment is what determined the team’s response when the company-wide announcement finally came.
Formalizing, then communicating inside and out
Only after the successor is finalized and every stakeholder has individually agreed do you hold the official board meeting (step 4). By that point the meeting is no longer a negotiation but a matter of documentation and making things formal. The investor email (step 5) is not a boilerplate notice; it carries the full story you wrote in step 1. The company-wide announcement (step 7) puts the successor up front, paints the future, does not dodge questions, and is followed by several days of office hours. And the public post (step 9) is treated not as a formality but as a strategic asset. Shi’s LinkedIn post announcing his exit crossed 100K impressions and, he says, opened countless doors.
The named successor was the VP of Engineering, who had grown over seven years from being the first engineering manager and had scaled the team from 10 to 100+. Even an obvious choice was articulated and earned buy-in from each board member one by one. That care is what turns an exit from an “organizational accident” into a “designed succession.”
The “invincible window” before the last day
The step that impressed me most is step 8. Once you have announced your departure, the stretch until your last day is the only time you can wield full founder authority with zero consequences, because you cannot be fired anymore. So you push bold moves here. If they land, your successor inherits the win; if they miss, you take the heat and it is forgotten. At the same time, you tee up big deals near closing and products near shipping, so your successor can win in their first days. And you leave behind transition documents, relationship maps, and decision frameworks, which is knowledge transfer at its finest.
Using the final weeks not as a wind-down but actively as a runway for your successor: that reframing is, I think, one of the most practical pieces of wisdom for a CTO letting go of a role.
The final exam: making yourself irrelevant
Shi closes the essay with its most important lesson. Whether the resignation process succeeded can only be measured by whether the company grows after you leave. After he left, Super.com kept growing 50%+ year over year.
The best founders make themselves irrelevant. They build systems that do not depend on their daily involvement, develop leaders who can decide without checking in, and create cultures that persist regardless of who is in charge. You should step back only when the team can accelerate without you. That your absence makes no difference to the company’s trajectory is, he writes, the sign you have truly succeeded as a founder.
This is the hardest question of all, posed before you even get to talk about a next career. We can only move on to the “next” once we have finished building the current place to withstand our absence. Conversely, as long as you cling to a state where nothing runs without you, the next door never opens in the first place. Exit design, the essay teaches, is not about the rituals of the final months; it is the final exam for the organization-building you have been doing all along.
The question of a CTO’s next career
As a hands-on CTO, I read this essay not as a recommendation of a specific destination but as a way of framing the question. Anthropic was Shi’s answer, not everyone’s. What matters is the shape of the thinking that leads to such an answer.
To put it in order, thinking about a CTO’s next career means looking at three layers at once. The first is the trend layer: reading where value is migrating across the technology stack. The move to AI labs was one response to value drifting toward the foundation-model side. The second is the asset layer: identifying where your accumulated operator experience becomes most scarce. The third is the exit layer: whether you can finish building your current role into something that withstands your absence.
Only when all three line up does the next career become a designed step rather than an escape or a leap of faith. And the third, the exit layer, is the only one you can start on right now, even before a destination is decided. Just as Shi began preparing nine months out, an exit is something you build months before the day you decide to leave, not on that day.
Conclusion
Using Henry Shi’s resignation essay, I have thought through a CTO’s next career. The three things I took away are these.
- Read the trend. From wrapping external APIs to building foundation models from the inside. The move to AI labs can be explained as a re-reading of which layer the value is shifting toward.
- Know the operator’s value. Experience scaling from 0 to 100+ and shipping through crises is hard to substitute with research purity. This is the core of a CTO’s career capital.
- Design the exit. Write your own story, tell people one at a time in order, and use the window before your last day as a runway for your successor. Exiting well is itself the foundation of the next career.
There is an endless supply of material on how to build a company, and almost none on how to leave one. The value of this essay is that it opens up nine months of preparation into that gap without holding back. A CTO’s next career begins not with searching for a destination, but with first shaping the current place so you can let go of it well.
That’s the view from a hands-on CTO who still writes code in a fully remote organization, reading a founder’s resignation essay and thinking it through, reporting from the field.