OpenAI and academic partners find coding agents speed up neglected research software by up to 60x, but can't validate the underlying science.
Continue to AI University →