Why Are CTOs Becoming ICs Again?

In the last eighteen months, the following has happened:

• Workday's CTO joined Anthropic as a Member of Technical Staff.
• Anthropic's own CPO, the co-founder of Instagram, moved into Labs and called himself a builder again.
• You.com's co-founder and CTO joined Anthropic as a Member of Technical Staff.
• Adept's CTO, a co-author of the Attention Is All You Need paper, did the same.
• Super.com's co-founder and CTO joined Anthropic Labs after scaling the company past $200M in annual revenue.

These are not retirements. These are not failures. They are some of the most senior technical leaders in the industry walking away from C-suite titles at multi-billion dollar companies to go back to building.

That is unusual enough on its own. So why is it happening?

This post is my attempt to find the answer.

The Roll Call

A quick tour of the cases, because the volume is part of the story.

Peter Bailis joined Workday as CTO in May 2025, after leading AI for data products at Google Cloud. He lasted under a year. By March 2026, he was at Anthropic working on reinforcement learning engineering. The company that hired him as an engineer is also the company building HR apps that compete directly with Workday.

Mike Krieger, co-founder of Instagram and Anthropic's CPO since 2024, stepped down on January 14, 2026 to co-lead Anthropic's Labs incubator alongside Ben Mann. His own framing of the move is the cleanest articulation of the whole pattern:

"We've reached a watershed moment in AI. Model capabilities are advancing so fast that the window to shape how they're used is now. That's why I'm getting back into builder mode, moving from my role as CPO and joining our Labs team."

Bryan McCann, co-founder and CTO of You.com, left for Anthropic in March 2026. You.com had to promote an internal engineer to fill the gap.

Niki Parmar, CTO of Adept and a co-author of the original transformer paper, quietly joined Anthropic in late 2024. Hers is the earliest case I know of, and it shows the pattern was already starting before the SaaSpocalypse made it visible.

Henry Shi, co-founder and CTO of Super.com, joined Anthropic Labs in mid-2025 to work directly with co-founder Ben Mann and Mike Krieger. He scaled Super.com from zero to over $200M in annual revenue before leaving. He had been an early Anthropic enterprise customer, which is part of why the move made sense to him.

Five people. Five C-suite or near-C-suite titles handed back. All five went to roles that, on paper, look like a step down.

"Member of Technical Staff" Is Not a Demotion

The first instinct, especially from outside, is to read this as failure. A CTO becoming a "member of technical staff" sounds like a step down a career ladder.

The ladder is the wrong frame.

At Anthropic and OpenAI, Member of Technical Staff is the common engineering title used across research and engineering. The structure is deliberately non-hierarchical. There is no "Senior MTS" climbing toward "Principal MTS" climbing toward "Director MTS." OpenAI president Greg Brockman has explained the labs do not want to "bucket people into researchers and engineers."

Compensation is not a step down either. MTS base salaries at Anthropic run from $300K to $405K. At OpenAI the band stretches from roughly $210K to $530K. Equity grants push total compensation for senior hires well into seven figures.

The signal is not about position. It is about proximity.

A CTO at a public SaaS firm spends most of the week on org politics, board prep, integration of acquisitions, budget defence, and quarterly narratives. An MTS at a frontier lab spends most of the week shipping changes that go into the next model. Both jobs are senior. Only one of them is close to the work.

That is the trade these five people made. Authority for proximity. Title for impact.

One Direction

All five went "down" the stack into the model layer. Bailis, Krieger, McCann, Parmar, and Shi are all working on the systems that the rest of the industry now depends on.

None went sideways into another CTO seat at another SaaS firm. The lateral move that would have been the obvious next step in the old playbook was not on the menu, and that negative space is louder than the moves themselves.

The Push: Why the SaaS C-Suite Got Uncomfortable

I covered the market shock in The World Is Changing and the Market Has Priced It In. Nearly $1 trillion in software stock value evaporated in seven trading days after Anthropic and OpenAI launched back-to-back agent platforms. Oracle, ServiceNow, Salesforce, Workday, Adobe, and SAP all took 16% to 30% hits.

The market did not just price in slower growth. It priced in a structural question: when agents do the work, what is the per-seat business model worth?

For a CTO at one of those companies, the experience is specific.

Your job is to defend a revenue line that is being repriced in real time. Your headcount and roadmap are designed for the application layer that just got commoditised. The companies you bought last year are still being merged in, and your engineers are stuck stitching those products onto the platform instead of building anything new. Your CEO is writing investor letters explaining why the company is "AI-native now." Your engineers, the strong ones, are getting recruited by the labs every week.

Bailis joined Workday as CTO in May 2025. Less than a year later, the seat had gotten worse on three fronts. Workday's co-founder Aneel Bhusri had returned as CEO, the kind of move boards make in crisis mode. Bailis was now responsible for integrating a string of AI acquisitions he had not chosen. And Anthropic, the company recruiting him, was openly shipping HR apps that competed directly with Workday's core market.

The company poaching him was the same company most directly threatening the company that hired him.

That is not a rage quit story. It is a calculation. If you spent your career building infrastructure that AI now eats, and the lab eating it is hiring, the cleanest move is to go build at the layer that is winning.

The Pull: Where the Real Work Lives

The other half of the equation is what these people get on the other side.

In Software Engineering to Outcome Engineering, I covered the shift in what engineering work even is. Boris Cherny, the head of Claude Code, said he had not written code himself in over two months when he was interviewed in January. Company-wide, 70% to 90% of Anthropic's code is AI-generated. Claude Code itself is around 90% written by Claude Code.

The most interesting engineering problem on earth right now is "how do we make agents better." Not "how do we route a sales workflow through an enterprise CRM." Senior technical people know this, and they go where the interesting problem is.

Krieger said it in plain language: the window to shape how AI gets used is now. That is the pull. It is not money, although the money is fine. It is the sense that the technical decisions being made at the model layer in the next eighteen months will outlive the careers of the people making them.

A CTO seat at a SaaS firm does not give you that. A flat MTS role at a frontier lab does.

The Structural Frame: From Hierarchy to Intelligence

There is a deeper reason this is happening, and Jack Dorsey wrote it down.

In April, Dorsey and Sequoia's Roelof Botha co-authored an essay called From Hierarchy to Intelligence. They argue that hierarchy existed because humans had to relay information through layers. Middle management was the routing protocol for an organisation. AI can now hold a continuously updated model of the entire business and route the work itself, which kills the reason the middle layer exists.

Block put the theory into practice. The company cut roughly 4,000 of its 10,000+ employees and reorganised around three roles only:

1. Individual Contributors who build the system.
2. Directly Responsible Individuals who own a specific outcome on a 90-day cycle.
3. Player-Coaches who stay hands-on while developing people.

There is no middle layer. The system handles coordination.

If you read the CTO migration through that lens, it stops looking like a series of strange career choices and starts looking like a leading indicator. Of those three roles, the most senior technical people in the industry are choosing the first one. They are choosing IC. They are choosing player-coach. They are not choosing "VP of VPs."

The Block framework is not the only valid org design for an AI-native company. But it points at something real. When the middle compresses, the rational path for talent is to move toward the layer that survives. That layer is closer to the work, not further from it.

Tying It to Ideas Over Hierarchy

I wrote about this from the culture side in Ideas Over Hierarchy. Steve Jobs had a line that keeps coming back: good people don't stay if ideas don't win.

The CTO-to-IC move is the same principle expressed at the senior end of the career.

In a hierarchy-heavy SaaS firm, the best technical idea routes through layers of approvals, board reviews, and budget cycles. A change to a model architecture at a frontier lab ships into the next training run. The path from idea to impact at the lab is shorter than the path from idea to PRD at the incumbent.

Talent flows where ideas can move. The senior cases above are not driven by ego, money, or escape. They are driven by gradient. The people changing roles are following the gradient toward where their best ideas can still cause something to happen.

My Take

Three claims I am willing to bet on.

1. The CTO Role at the Application Layer Is Shrinking Faster Than People Realise

The traditional CTO role at a public SaaS firm bundled three things: technical authority, organisational scaling, and political translation between engineering and the board. AI eats two of those. Architecture decisions can now be made by one strong technical person plus an agent fleet. Most of the "managing managers of managers" function is what Dorsey is now arguing the system should handle.

What is left is the political layer. That is a real job, but it is not a builder's job. The CTOs who care most about the work are the ones leaving first. The ones who stay are increasingly the ones who like the political layer best, which is exactly the wrong filter for the next decade.

2. The Signal Is in the Direction, Not the Volume

Five public cases is still a small number. The story is not the count. It is the fact that "CTO to MTS" is now a legible, smart career step.

Two years ago, a CTO going IC was read as a soft retirement. In 2026, it is read as the move of someone who knows where the puck is going. The labels did not change. The interpretation did. That shift in interpretation is the news.

3. The Senior Career Path Is Being Rewritten in Public

The next wave is not "more CTOs become MTS." The next wave is the question every senior person in software is now quietly asking themselves: what does my career look like if there is no middle to climb?

The Block model gives one answer. Stay close to the work. Own outcomes on short cycles. Coach without becoming a manager of managers. The companies that build that ladder first will absorb the talent leaving the companies that have not.

The companies still designing org charts around fifteen-layer hierarchies are quietly losing this race. Not because their people are leaving for more money. Because their people are leaving for more leverage.

What This Means

For SaaS firms. Losing your most technically credible person to the company building your replacement is a leading indicator, not a lagging one. The reflex is to backfill the title and move on. The smarter reflex is to ask why the role is no longer attractive to the people you most want in it. If your CTO seat has become 80% politics and 20% architecture, the architecture-driven candidates will keep walking past it.

For AI labs. The talent magnet works because the work is the most interesting on earth right now. The day that stops being true, the migration reverses. The flat MTS title is not the magnet. The proximity is.

For senior individuals. Title inflation has been masking a real shift. The leveraged role in 2026 is "skilled builder plus agents," not "manager of managers of builders." If your week is mostly status, alignment, and review, you are running an org chart, not a career. The five people above already noticed.

For boards and CEOs. Stop trying to retain technical leaders by giving them more reports. They do not want more reports. They want to be near the model, the data, and the customer. The people you will lose in the next eighteen months are not leaving over comp. They are leaving over distance from the work.

Closing

The CTOs leaving SaaS firms for IC roles at AI labs are not running from anything. They are running toward where leverage now lives.

The org chart of the next decade will not reward the people who climbed the highest. It will reward the people who stayed closest to the work. Those two things used to point in the same direction. They no longer do.

The senior career path is being rewritten by the people walking it. The most useful thing the rest of us can do is read what they are writing.