Agent harnesses need maps, not more manuals

Agent harnesses need maps, not more manuals

Temperature zero is not a determinism guarantee

Vulnerability counts need a CVE ledger

An AI detector you can switch off measures cooperation

Less restrictive is now a model spec

Anthropic's Opus 5 is cheaper, engages safety classifiers 85% less than Fable 5, and can silently reroute blocked prompts to weaker models. Restriction level just became a product axis you have to audit.

Agent ROI is often feature activation in disguise

The open-weight ban letters price the margin, not the risk

The Jacobian counterexample has a human byline

Task cost is a harness property, not a model price

The FCA put the model vendor inside the sandbox