Anthropic Keeps Claude Code Lean as Rapid Model Gains Undercut Structured Context
Updated
Updated · Ars Technica · Jul 20
Anthropic Keeps Claude Code Lean as Rapid Model Gains Undercut Structured Context
2 articles · Updated · Ars Technica · Jul 20
Summary
Anthropic’s Claude Code team is deliberately keeping its AI coding harness lightweight, betting fast model improvements make rigid product planning and heavily opinionated features less useful.
Cat Wu, Claude Code’s product head, said the harness is meant to sit between the model and a developer’s codebase, controlling what the model sees and what actions it can take.
Claude Code also avoids building structured codebase context by default; Wu said Anthropic’s evaluations show no measurable improvement from that approach.
The strategy sets Claude Code apart within a crowded field that includes OpenAI’s Codex, Google’s Antigravity, open-source OpenCode, and startup tools such as Cursor and Augment Code.