Claude Opus 4.7 built a complete AlphaZero machine learning pipeline from scratch using only a basic description.
April 29, 2026
Original Paper
Frontier Coding Agents Can Now Implement an AlphaZero Self-Play Machine Learning Pipeline For Connect Four That Performs Comparably to an External Solver
arXiv · 2604.25067
The Takeaway
Frontier coding agents have moved beyond simple scripts to architecting entire machine learning systems. This model implemented the full self-play loop and training logic for a Connect Four solver that rivals external mathematical solvers. The AI required no reference code, demonstrating an autonomous understanding of complex algorithmic structures. This milestone marks the beginning of recursive self-improvement where AI builds the infrastructure for its own advancement. Developers can now use agents to bootstrap sophisticated training pipelines that previously required weeks of human engineering.
From the abstract
Forecasting when AI systems will become capable of meaningfully accelerating AI research is a central challenge for AI safety. Existing benchmarks measure broad capability growth, but may not provide ample early warning signals for recursive self-improvement. We propose measuring AI's capability to autonomously implement end-to-end machine learning pipelines from past AI research breakthroughs, given a minimal task description. By providing a concise task description instead of the full prior wo