Agent leaderboards saturate and leak. A private, held-back task set is what actually predicts how your agent behaves in production.

The EU AI Act's general-purpose-AI obligations, decoded for teams that build on top of foundation models rather than train them.

Frontier windows hold a million tokens but retrieve the middle unreliably — position, not capacity, decides recall.

In production RAG, retrieval — chunking, embeddings, freshness — decides quality; the model is rarely the culprit.

Indirect prompt injection hides in retrieved content; containment held, prompt guardrails didn't.

Test-time compute is Signal on hard, decomposable, checkable problems — and Noise almost everywhere else.

A cost, freshness, and grounding teardown of RAG versus long-context prompting for knowledge-heavy apps.

How to run NIST's AI Risk Management Framework and its generative profile as real engineering controls instead of self-attested paperwork.

C2PA fixes AI-media authenticity at the source, but stripped metadata and opt-in adoption limit it in the wild.

A planted backdoor can persist through the exact safety pipeline meant to remove it. Treat open weights as an untrusted supply chain.

Grammar-constrained decoding guarantees parseable JSON — not correct JSON. Ship the syntax, gate the semantics.

Bounded tool-calling holds; long multi-step chains still drop the thread.

One shared tool interface turns M×N integrations into M+N — real leverage, but the spec is still moving.

Speculative decoding cuts single-stream latency 2-3x with identical outputs, but throughput-bound batches erase the win.

Adversarial suffixes tuned on open-weight models transfer to closed ones. Exploitable by design — blocklists won't save you.

The smallest evaluation that catches the regressions a single accuracy score always misses.

The EU AI Act's transparency rules make AI media and text carry disclosures — machine-readable for makers, visible for deployers.

How to judge an open-weight model by its eval harness and model card instead of launch-day hype.

How 4-bit quantized finetuning shrinks the GPU bill for small teams — and where quality actually leaks.

Indirect prompt injection to exfiltration is a real, chainable ATLAS path; cut the egress channel, not the prompt.

Unbounded context growth is a silent cost bomb; discipline, caching, and routing defuse it.

Mixture-of-experts cuts per-token compute but not VRAM — a production serving verdict hung on the Mixtral paper.

A field guide to the two doors into high-risk and the seven duties that follow.

Supply-chain and model-poisoning entries in the OWASP LLM Top 10 are real, exploitable, and chronically under-weighted.

Why the cross-cutting Govern function decides whether Map, Measure, and Manage do anything at all.