Anthropic Reports 8x Increase in AI-Assisted Code Delivery
Original: When AI Builds Itself: Our progress toward recursive self-improvement
Why This Matters
Demonstrates measurable AI acceleration in software development with implications for autonomous AI development
Anthropic's engineers now ship 8x more code per quarter than in 2021-2025, aided by AI systems. The company shows AI models can complete increasingly complex tasks, from 4-minute software tasks in March 2024 to 12-hour tasks by 2026, suggesting potential for recursive self-improvement.
Anthropic published research showing AI systems are accelerating AI development itself. Engineers at the company deliver 8 times more code quarterly compared to 2021-2025 levels using AI assistance. The capability progression shows Claude Opus 3 handled 4-minute tasks in March 2024, Claude Sonnet 3.7 managed 90-minute tasks a year later, and Claude Opus 4.6 completed 12-hour tasks by 2026. Task completion length doubles every four months, accelerated from the previous seven-month trend. AI models have saturated coding benchmarks like SWE-bench and research reproduction benchmark CORE-Bench within two years. The company suggests this trajectory could lead to recursive self-improvement, where AI systems autonomously design successors, though notes this raises control and safety concerns.