AI Engineer · October 22, 2026

Agent evaluations without a benchmark: plant known faults and test invariants

video Model Evaluation AI Engineering

Agent evaluations without a benchmark: plant known faults and test invariants video thumbnail

Why it matters

Nathanael Weill builds tests from known answers: generate questions from source pages, plant contract defects, remove required clauses, and vary details that should not change the result. Missing-clause tests exposed failures that plausible answers hid. Expert feedback improves the test set. These synthetic checks accelerate development but do not establish readiness for deployment.

My takeaway: Start with expert-reviewed synthetic examples and explicit expected results. Pair each planted defect with an unchanged valid example. Remove required fields or clauses, reorder content, and change irrelevant names; check misses, false alarms, and repeatability in code. Keep realistic held-out cases and expert review for the final release decision.

Keep exploring

More curated notes connected through Model Evaluation and AI Engineering.

OpenAI News · framework

Pacing model development in an era of cyber-critical capabilities

OpenAI says preliminary evidence that Astra may meet its Critical cybersecurity threshold led it to pause frontier reinforcement-learning work for two weeks and keep its largest planned run on hold. New safeguards include stronger workload and network isolation, continuous boundary testing, token-level monitoring that escalates suspicious tool activity, and broader alignment checks for deception, reward hacking, and unauthorized access.

OpenAI News · framework

A blueprint for democratic governance of frontier AI

OpenAI proposes a three-part U.S. frontier-AI governance model: harmonize emerging state safety laws into a federal baseline, strengthen CAISI as an evaluation and standards institution, and coordinate a broader resilience program. Proposed controls include severe-risk evaluations, transparency reports, independent audits, safety-incident reporting, model-weight security, whistleblower protection, and periodic technical assessments.

OpenAI News · framework

OpenAI’s Frontier Governance Framework

OpenAI's 22-page Frontier Governance Framework maps its frontier-model processes to California's Transparency in Frontier AI Act and the EU AI Act's general-purpose AI code. It documents lifecycle risk assessment, cyber-offense and other risk tiers, mitigation and residual-risk decisions, critical-incident handling, security risk management, model reporting, external review, responsibility allocation, and change control.