Why Silicon Valley's 'Already Made It' crowd is taking junior titles at AI labs

Key points

• Tom Blomfield, co-founder of the UK banking app Monzo, announced Monday he is joining Anthropic's compute team as a member of technical staff, not as an executive.
• Andrej Karpathy, a founding member of OpenAI, joined Anthropic's pre-training team in May 2025, describing the next few years at AI labs as "especially formative."
• Peter Bailis left his role as Chief Technology Officer at Workday, a business-software company with $8 billion in annual revenue, for a staff-level technical post at Anthropic in March 2025.
• Chamath Palihapitiya, a former Facebook executive, took his first full-time operating job in over a decade as CEO of 8090 Labs, an AI coding startup, after raising $135 million in April 2025.
• Eric Wu, who ran home-buying platform Opendoor for a decade, launched NavigateAI, an AI assistant for construction workers, with $25 million in seed funding in 2025.

Something unusual is happening in Silicon Valley. People who already have the money, the reputation, and the freedom to do nothing are choosing to start over, often in roles far below what their track record would normally command.

Tom Blomfield is the latest example. He co-founded GoCardless, a payments company, and then Monzo, one of Britain's most recognised digital banks. He spent four and a half years at Y Combinator, the famous startup school, mentoring the next generation of founders. On Monday he announced he is stepping back from that role to join Anthropic, the AI safety company behind the Claude family of AI assistants, on its compute team. His title: member of technical staff.

That title matters. At Anthropic and at OpenAI, "member of technical staff" is a deliberately flat label used for nearly everyone who works on the technical side, whatever their experience level. There are no vice presidents or directors hidden inside it. Blomfield, who could reasonably walk into a senior leadership role almost anywhere, chose the same job title a recent computer-science graduate might hold.

He is not alone.

Andrej Karpathy, one of the original team at OpenAI and later head of AI at Tesla, joined Anthropic's pre-training team in May. Pre-training is the foundational stage where an AI model, the software engine behind chatbots and tools like Claude, learns from vast amounts of text and data before it is ever released to the public. Karpathy wrote that "the next few years at the frontier of LLMs," meaning large language models (the technology that powers modern AI assistants), "will be especially formative."

Peter Bailis made perhaps the sharpest trade. He was Chief Technology Officer at Workday, overseeing AI strategy across a business that reported $8 billion in annual revenue. He lasted less than a year before taking a staff-level post at Anthropic this March.

Not everyone is joining an existing lab. Chamath Palihapitiya, best known for early work at Facebook and a long run as an investor, stepped back into a full-time operating role for the first time in more than a decade, as CEO of his own enterprise AI coding startup, 8090 Labs. The company raised $135 million in a funding round led by Salesforce Ventures. Eric Wu, who led Opendoor through a decade of growth before stepping back in 2023, recently launched NavigateAI, an AI copilot (a software assistant that works alongside a person rather than replacing them) built for construction workers, with $25 million in seed funding. Wu told TechCrunch AI he feared he would "probably regret" not acting.

What does this mean for everyone else?

It means the people who know this industry best are betting their time, not just their money, that AI is still in its early days. That is a signal worth noting.

For ordinary users and workers, the practical takeaway is straightforward. The tools these people are building, from AI coding assistants to construction-site helpers, are aimed at everyday professional tasks, not science fiction. Expect more focused, job-specific AI tools to reach the market in the next two to three years, backed by founders who have built real products before.