Evaluation suite additions
We added new red-team prompts and failure checks across coding and multi-step tasks.
Aug 12Notes on research, product and deployments from the Brightforge team.
We added new red-team prompts and failure checks across coding and multi-step tasks.
Aug 12Introducing policy controls and monitoring to keep actions within your intent.
Jul 29A short overview of how we steer models with small, principle-like rules.
Jun 23We’re documenting changes and caveats for each significant update.
May 9