Blog
Writing from the team building Latent: the research behind reading a model's internal state, and what we measure.
Latent is live
Starting today, any team can sign up for Latent, add one line of Python and see which of its model's answers are made up. Free to start.
Latent is liveStarting today, any team can sign up for Latent, add one line of Python and see which of its model's answers are made up. Free to start.
Latent against an LLM judgeJudging every answer with a frontier model takes seconds and costs about a cent per answer. Latent reads a finished answer in milliseconds, and inside your own model it adds no measurable time at all.
Why internal stateA model's internal state reveals failures its output doesn't. Four published results on activations, chains of thought and training runs, and what they mean for teams serving their own models.