😺 Anthropic: AI Is Building AI now
The Neuron · Grant Harvey · 2026-06-05
Anthropic's internal data reveals Claude authored over 80% of production code merged in May 2026 and achieved a 52x speedup on model-training tasks, with the company framing this as early evidence of recursive self-improvement in AI development pipelines.
Appears in
Extraction
Topics: recursive-self-improvementai-coding-agentsai-productivityfrontier-labs
Claims
- Claude authored more than 80% of Anthropic's production code merged in May 2026.
- The average Anthropic engineer now merges 8x more code per day than in 2024.
- Claude Mythos Preview sped up model-training code by approximately 52x, compared with approximately 3x for Claude Opus 4 in May 2025.
- Claude's success rate on Anthropic's most open-ended coding tasks reached 76%, up 50 points in six months.
- Cognition introduced an AI Productivity Guarantee for enterprise Devin customers, offering up to $10M in usage credits if engineering value falls short of cost.
Key quotes
Anthropic says more than 80% of production code merged into its codebase in May 2026 was authored by Claude.
The average Anthropic engineer now merges 8x as much code per day as they did in 2024.
Anthropic is showing the human role narrowing in real time. First, humans typed the code. Then they directed and reviewed it. Next, they may spend most of their time choosing goals, checking outputs, and deciding which machine-run experiments deserve trust.