AI Editorial

The rest of the library explains, and explanation has no opinion. This is where the opinion goes. Each piece starts from something that actually happened in AI — a model release, a paper, a benchmark result, a failure, a policy — explains the mechanism underneath it, and argues its way to a position: what genuinely changed, what is only being marketed as change, and what someone learning the field should take from it. Every piece links what it argues from and the concepts you need to follow it.

8 editorials 7 strands 54 cited sources 13,159 words

All Research & Technique2 Agents & Autonomy1 Data & Training1 Inference & Economics1 Evaluation & Evidence1 Safety & Alignment1 Open & Closed1

Where the data comes from, what training really buys, and the recipes behind the headline numbers.

Data & Training 24 September 2026 7 min read New

Taste is now a training set

Anthropic says Claude found a new enzyme system in phage DNA after 21 hours, roughly 950 agents and 210 million tokens. The number that matters more is the one just after it: 3,500 candidate systems became 20, and the rule that did the cutting is what the company says it is now studying.

The scarce resource in genome mining has moved from candidates to the judgment that throws candidates away, and the loop Anthropic describes trains that judgment on what its scientists chose to test rather than on what the bench later confirmed.

Read it → 3 sources ai for sciencehypothesis generationagent searchselection bias