
Evals and training environments for frontier labs
idler.ai (opens in a new tab)idler is a frontier data research lab that makes the evaluations and training environments leading AI labs use to measure and improve their models. Over the past year its main work was coding evals for well-known top coding models, and it has since moved into further domains, among them recursive self-improvement and automated research, law, enterprise business and cybersecurity. Its data creation scales through technology, with every data product running on a single self-reinforcing platform that learns from experience. Labs now want more of its data than it can deliver, and the small founding team, which has reached product-market fit, is growing quickly to catch up.