Sajith K
Writes on AI engineering in Artificial Intelligence in Plain English.
Contribution to This Vault
- JEV vs LLM - Your Software Doesn’t Want a Conversation It Wants a Decision (2026-09) — an audit of Jev’s launch claims: the docs, the benchmarks, and what he calls the pile-on from the sceptics, sorted into “what’s real, what’s marketing, and when you’d actually reach for it.”
The piece earns its place for doing arithmetic nobody else in the reaction did. He confirms TypeSafe’s 238x price claim is exactly correct against Claude Fable 5.1 — and then points out that the realistic incumbent for classification work is Claude Haiku 4.5, against which the same math gives ~24x. He grants that ~24x is still a real business case; the objection is that 238x is the number in the headline.
Two other things are worth carrying. He separates a Noul value from a confidence score — a Noul of 0.5 is a genuine coin-flip rather than medium urgency, and confidence is a distinct quantity derived from how peaked the distribution is, so conflating the two breaks thresholds silently. And he ends by dismantling his own title: every use case the vendor leads with puts the model beside an LLM rather than in place of one, making it a decision layer around a generation layer.
His summary is the fairest assessment written about the launch in either direction — the speed, the cost and the novelty of the category are real, and the marketing runs about ten times ahead of the evidence, which the vendor half-admits.
Related Concepts
Related Notes
- Jev · TypeSafe
- Reception of Jev — where this piece sits in the wider reaction