Sometimes it’s not the benchmarks that stick with you — it’s a sentence about feelings. Fiona Fung, Head of Engineering for Claude Code and Cowork at Anthropic, joined Lenny’s Podcast on June 21. The conversation circled a question a lot of us are wrestling with right now: how do you build a team that actually works with AI, instead of just talking about it?
The ‘AI-pilled’ team
Fung’s résumé is something else. Eleven years at Microsoft (Visual Studio, TypeScript), then Meta, where she started Facebook Marketplace — now over $100 billion in GMV a year. At Anthropic she runs the teams behind the two products reshaping our industry the fastest.
The context is striking. More than 80 percent of the code merged into Anthropic’s production codebase is now written by Claude itself, up from low single digits before Claude Code launched. So the engineers there aren’t delegating a little — they spend the whole day working side by side with agents.
The side effect nobody planned for
And here’s the honest moment. Fung says that after a while, the Claude Code team noticed something: “It could start being a lonely experience because we all started just working with our agents so much.”
That’s a remarkably candid admission. When everyone sits with their own swarm of agents in their own session, the obvious things fall away — turning to a colleague, debugging together, swapping war stories about a nasty bug. The team’s response wasn’t technical, it was social: shared programming lunches, hackathons, dedicated ‘maker time’ where people sit shoulder to shoulder again.
My take
Honestly, I find this more important than the next version number. We talk endlessly about productivity, tokens, and agent chains — but rarely about what it does to a team when human-to-human collaboration gets replaced by human-and-agent.
I feel it too. Claude Code is a huge lever for me — but the most productive days are sometimes the quietest ones. The fact that the very team building these tools names the problem and actively pushes back is a good sign. Efficiency isn’t everything. The interesting question for the next few years isn’t only how good the agents get — it’s how we avoid losing the human side of work along the way.
Sources: Lenny’s Newsletter: Fiona Fung (Anthropic), Yahoo Tech, Let’s Data Science