Agent-Reliability

Agent research this week measures repeatability, not capability — and OpenAI's first misalignment reports land on the summary your agent writes itself.

Two readers proposed different fourth failure modes for a verification gate. Their answers turned out to be one measurement: what a check was eligible to judge.

Four of six red-team attacks walked past every gate in my agent harness that read artifacts. Only running the code on inputs the producer never chose caught them.

A safeguard written into a prompt can't tell when the thing it guarded against stopped happening. It keeps running, faithfully, and charges you for it.

Agent autonomy isn't a dial you set once — it's a ladder you climb on edit-rate data. The real risk: thinking you're a rung above what your data supports.

This week's top AI research converged on one engineering idea: the best agents know when to stop. Abstention, verifiers, and the loops debate, for builders.