#production-ai

Evals are written like unit tests and behave like psychological tests. A century of measurement theory already named the failures teams keep rediscovering.

SRE-Bench spent five thousand expert hours keeping its questions secret. Testing has handled that problem for forty years and calls it something else.

McKinsey counted 1,719 leaders in August. Large firms scaling agents jumped to 40%. Firms moving earnings stayed at 6%. Part of that gap is architectural.

The AI Act's transparency duties went live on 2 August. They govern what an agent says about itself. Nothing governs what it may touch.

Gartner says 40% of enterprise apps will embed agents this year, and that most vendors are faking it. Both are true because 'agent' stopped being a technical word.

Enterprises burned their 2026 AI budgets in months. The token bill made pricing a strategic question — and the per-seat model is now a subsidy.

The Commission's draft Annex III guidelines published today put two-sided talent platforms squarely inside the high-risk perimeter. The same engine, surfaced to the candidate, sits outside it. The verb is what changed.

Karpathy reframed LLMs as ghosts trained by imitation of text, not animals that learn the world. The shift from vibe coding to agent engineering is what production has been waiting for.

El País counted us. The handful who actually train models are doing something different than the rest of the AI workforce — and what we're doing is changing in a way the headline misses.

PwC measured 1,217 executives across 25 sectors. 20% of companies capture 74% of the economic value of AI. The split is not budget, not stack — it is whether the company uses AI to do what it already does, or to sell what it could not before.

Notes from a ProductTank Madrid we hosted at Shakers, on what changes when AI enters a product team — and what doesn't.

Dorsey from Block and Botha from Sequoia published a new operating model for AI-native organizations. It looks like architecture, but it is really a strategy for not letting your roadmap become the limit on what your AI can do.