Perspectives, methodology notes, and standards guidance on evaluating, assuring, and credentialing AI. (Sample titles below — the library launches with the site.)
A daily, citation-complete record of movement in AI governance: bills, rulemakings, standards, registries, and enforcement. Automated signals, human-approved analysis, published every day — including days with nothing to report.
Latest edition: see the brief page.
A weekly synthesis of movement in AI evaluation, guardrails and observability, standards, and regulation — the analysis layer above the Daily Brief's signal layer. Drafted, then reviewed and released by a human decision. Published Friday evenings.
Latest edition: see the briefing page.
AI trials hand-pick the competent human ready to take over. At deployment scale that role has no definition, no training, and no supply — and it is the real bottleneck.
Entry-level work is where professional judgment forms. Automating it away while demanding human oversight is a contradiction the field has not yet priced in.
As vendors self-generate evidence at scale, the scarce resource becomes independent, credentialed capacity on the demand side. Here's why that's where the trustworthy-AI workforce must be built.
Beyond the demo: designing reproducible harnesses that test tool use, guardrail adherence, and behavior under adversarial and edge-case conditions.
A practical walk through aligning an AI system to NIST AI RMF and ISO/IEC 42001 — and what evidence each control actually requires.
Why one-time audits go stale the moment a model updates — and how continuous, evidence-linked assurance keeps pace with change.
How separating who authors a standard from who operates the credential produces more rigorous, more durable certifications.
A low-volume note when we publish something worth your time.