The GPT-5.5-Cyber capability profile beyond OpenAI's marketing: Simon Willison's evals, the Trusted Access Program scope, and what the Five Eyes briefings actually covered.
We compared what Anthropic Mythos and OpenAI GPT-5.5-Cyber actually do on offensive testing tasks. Capabilities, refusal patterns, evals, and where each model breaks down.
The LLM observability category has 4 distinct tool types in 2026. Confusing a reverse proxy with an SDK tracer costs trace coverage — not just $59/mo.
Langfuse rebuilt experiments as a first-class concept on April 13, 2026. Here is what changes for LLM evaluation workflows and how to migrate this week.