AI-Native Code Engineering

Poolside (Paris/San Francisco, founded by former GitHub CTO Jason Warner) is building AI models trained specifically on the software development process. Unlike code completion tools that predict the next token, Poolside's approach uses Reinforcement Learning on Code Execution (RLCE) — the model learns from whether code actually runs and produces correct results, not just whether it looks syntactically correct.

The $500M Series B (October 2024, estimated $3B valuation) made Poolside one of the most well-funded AI startups globally. The technical insight is that code has a uniquely verifiable feedback signal: it either works or it doesn't. This makes reinforcement learning far more tractable for code than for natural language, where 'correctness' is subjective.

Europe's positioning in AI-assisted software engineering is strategic: reducing dependence on Microsoft's GitHub Copilot for a tool that increasingly defines developer productivity. By understanding the intent behind codebases rather than just predicting syntax, Poolside aims to produce AI systems that function as autonomous engineers — designing architectures, debugging, and shipping features with minimal human oversight.