They build the most advanced AI on the planet. And now, the AI they built is making them question why they're there.
An anonymous employee at Anthropic — one of the world's leading AI companies, valued at $965 billion — described a feeling that goes beyond job insecurity into something deeper: a fundamental loss of self-worth.
"On smooth days, I can't help but think that anything I do is meaningless; everything has been automated, better and faster than anything I could ever do in the future," the employee said. "But when everything collapses, I don't know where things went wrong — at that moment, I realize I no longer know what I'm doing."
The numbers tell part of the story. By May 2026, over 80% of the merged code in the Anthropic codebase was written by Claude — up from single digits before Claude Code launched in early 2025. Each engineer was merging code at a rate eight times greater than in 2024. Another employee said they hadn't personally written any code for about five months. Boris Cherny, the creator of Claude Code, was even more blunt — he hadn't hand-written a line of code in eight months and was instead managing thousands of AI agents.
Fiona Fung, engineering lead of the Anthropic Claude Code and Cowork teams, revealed an unexpected side effect: engineers began to feel lonely. Everyone was collaborating with their own AI agents, leading to a decrease in direct communication among team members. The company had to implement interventions — pair programming lunches, hackathons, scheduled "co-creation time" — to try to restore human connection.
This is not the fear of unemployment. It's something subtler and perhaps more insidious. AI has not replaced these workers. It's that AI is performing so well that they start to doubt their own reason for "being here." Their code was written by AI. Their judgments are approximated by AI's judgment. Their role has shifted from "creator" to "approver" — and they're not even sure whether what they're approving is beyond their understanding.
The internal survey found that the median estimate from 130 researchers was that their output using AI was approximately four times what it was previously. But the report's own wording acknowledged "there remains a significant performance gap" between Claude and humans on the most open and judgment-requiring tasks.
If even the engineers at the world's most advanced AI company are experiencing this existential drift, the question becomes: what happens when AI digital employees enter every company in every industry?