@FinanceYF5: Anthropic has publicly disclosed how close they are to recursive self-improvement: Claude leads 26% of R&D tasks, less …

X AI KOLs Timeline News

Summary

Anthropic has disclosed that Claude leads 26% of R&D tasks with over 90% participation, involves 30,000 agents in continuous R&D, and details significant decision volumes and monitoring processes.

Anthropic has publicly disclosed how close they are to recursive self-improvement: Claude leads 26% of R&D tasks, less than 1% in February; tasks it participates in or leads have exceeded 90%. Approximately 30,000 Agents are continuously engaged in R&D. The August review exceeded 1 billion decisions, with online monitoring blocking only 0.002%. Offline systems flag about 100,000 records per week, with only about 50 handed over for human review.
Original Article

Similar Articles

@FinanceYF5: Anthropic internal data shows Claude is accelerating AI development — potentially a path to recursive self-improvement, or a path where AI autonomously builds a more powerful successor. This process is unfolding faster than expected, and its implications deserve more attention.

X AI KOLs Following

Anthropic internal data shows Claude is accelerating AI development, potentially leading to a path of recursive self-improvement. The process of AI autonomously building a more powerful successor is happening faster than anticipated.

@FinanceYF5: Source:

X AI KOLs Following

Anthropic is reportedly using 1,000 freelance software engineers at $280 per task to train Claude Code, having them write prompts, compare code, and test follow-ups to teach the AI how real developers work.