The best biology AI won't do biology
Anthropic's strongest bioinformatics model is partner-only, and the one you can call refuses biology. So I ran the open version of their eval across nine models — cheap ones included — and watched.
Read →Engineering notes on attested AI routing, Fusion evals, provider privacy, and open source model routing.
Anthropic's strongest bioinformatics model is partner-only, and the one you can call refuses biology. So I ran the open version of their eval across nine models — cheap ones included — and watched.
Read →I built PrometheusBench to measure how often a model refuses a plain question. The models that market themselves on safety refuse the most.
Read →Source: PrometheusBench on GitHub
A live engineering note on the first frontier Fusion attempt: what ran, what failed, and why we are not claiming a benchmark win yet.
Read →Source: Open Fusion methodology
TrustedRouter is reproducing Fusion-style DRACO evals with exact criterion scoring before publishing a headline comparison.
Read →Source: OpenRouter Fusion announcement
TrustedRouter gives developers OpenAI-compatible model routing while keeping the prompt path separate from the control plane.
Read →Source: Joseph Perla original
For AI routing, trust should be something an agent can verify, not only a policy page a human reads after the fact.
Read →Source: Joseph Perla original