“Fast-forward to today, it looks very different,” he said. “You have a Claude Code, but it has subagents that are other Claudes.” The user is no longer prompting Claude, he added: “It’s actually another Claude that does the prompting.”
“We’re starting to get to the point where it has ideas,” he said. “It’s looking at GitHub, it’s looking at [X]; it’s figuring out, ‘What should I build next?’”
Often, he added, he wakes up in the morning, and Claude has already taken action on a variety of ideas. “It’s quite exciting, because to me it means I can do much more, and the role of a builder is just totally changing,” he said.
But there are also risks, which Anthropic is taking very seriously, he said. When asked if he was worried about recursive self-improvement, Cherny said yes: “It’s one of the big risks for AI.”
