# ChipTrain > Training data and RL environments for AI that designs chips. We refine chip-design history (real bugs, real regressions, real agent failures) into graded environments with measured pass rates, across the whole flow from spec to signoff. Graders: formal equivalence (with counterexamples), mutation kill rate, co-simulation against a reference model, coverage delta, measured PPA. Graders run outside the agent's sandbox and are replayed for determinism. ## Start here - [TRAIN.md](https://chiptrain.ai/md/train.md): what we believe about training AI on chip design, and what we sell ## Essays - [In hardware, cheap tasks don't have to waste compute](https://chiptrain.ai/md/cheap-hardware-tasks.md): Hand-written chip-design tasks cost more than the compute they feed. The way out is the thing the cheap-tasks argument warns against, and hardware is the one domain where it works. - [Verification is the most verifiable job AI can't do](https://chiptrain.ai/md/verification-is-the-most-verifiable-job.md): Chip verification exists to check whether things are right, and its tools are machines built to decide. It should be the easiest job to train for. It is where models score worst. ## Contact - Sample tasks, data licensing, agent-failure partnerships: hello@chiptrain.ai