Keep pulling the thread on Boris Cherny.
The exponential growth phase for Claude Code began in May with the release of the Opus 4 model.
The Claude Code team intentionally built the product for six months before it achieved product-market fit because they were targeting the capabilities of a future, unreleased model.
The Claude Code team reached a point where the AI model wrote 100% of the product's code around October or November 2025.
Anthropic has launched a feature called "Routines" which provides server-side, persistent execution for agentic loops.
Every member of the Claude Code team at Anthropic, including the engineering manager, product manager, designers, data scientists, finance person, and user researcher, writes code.
Boris Cherny believes "process power" as a business moat will become less important because AI like Claude is becoming proficient at understanding and optimizing processes.
The Opus 4.7 model has the ability to "hill climb," meaning it can iteratively work towards a specified target until the task is complete.
At Anthropic, AI agents running in loops communicate with each other over Slack to resolve unknowns and collaborate on tasks.
No code is manually written at Anthropic; all of it is generated by AI models.
Boris Cherny believes Anthropic's primary competitive advantage is not its technology, which is publicly available, but its advanced organizational structure and processes built around AI.
The Opus 4.7 model has started to proactively use loop-based execution for tasks without being explicitly prompted to do so.
The original innovation team that created Claude Code has been reassembled and is now led by Anthropic's Chief Product Officer, Mike Krieger.