Weco AI

Weco AI develops recursively self-improving AI systems, with research demonstrating agents that can autonomously enhance their own performance through iterative refinement. Its flagship AIDE² project showed an outer loop that rewrote the company's own research agent over 100 steps, outperforming two years of hand-tuning. Weco AI offers Autoresearch capabilities that compete in AI research challenges, SpecBench for evaluating software correctness, and a human-in-the-loop design philosophy that keeps people involved in overseeing self-improving systems. The company provides platform and CLI tools for developers, maintains public documentation and GitHub repositories, and runs an active Discord community. Pricing is not disclosed on the website; interested developers can access documentation and GitHub repositories directly, with further details available by contacting the company.