Claude Writes 80% of Code, Engineers Grow Lonelier
Claude now merges 80% of Anthropic's code. Learn how to build self-improving Agents with planning, error recovery, and long-context power.
“AI Disruption” Publication 10100 Subscriptions 20% Discount Offer Link.
Recently, a talk by Theodora (Theo) Chu, a Product Manager on Anthropic’s team, has drawn significant attention.
Theo noted that more and more developers are no longer just “having heard of Claude,” but are now experiencing real efficiency gains in their daily work. Some say Claude has doubled their productivity, while others report a 10x improvement. Even more notably, Claude has begun to deeply integrate into Anthropic’s own engineering workflows: “Over 80% of the code at Anthropic is now merged by Claude.”
This signals a fundamental shift in the role of models.
Models are no longer limited to simply answering questions. Instead, they are completing tasks continuously within an environment that allows feedback, verification, and correction. “Close the Loop — give the model a way to verify its own outputs.”
In this talk, what Theo wanted to convey to developers is: “How should you adapt to this new world, and how should you build products for the future, rather than just for the past?”
To that end, Theo broke down in detail how to build self-improving Agents. “The real setup is letting Claude run continuously in loops, planning modes, and dynamic workflows.”
I personally think, “This is better than most $300 Agent courses out there.”
So, let’s dive into what this talk actually covered.
