Inside Anthropic’s AI Moat

For almost a year, since Claude Code's inception, I’ve been waiting for it to “come out from the terminal’s cage, to eat the agentic web,” well, that time has come (almost).

Claude Agent Is Coming To Eat The Agentic Web

Anthropic’s breakout moment didn’t happen overnight.

It followed a precise product arc that can be traced through three critical phases, each building on the last: Claude Code proved AI could do real work in the terminal → OpenClaw proved users wanted AI agents everywhere → Opus 4.6 and Cowork proved the enterprise knowledge work thesis.

Understanding this arc is essential to understanding the moat.