Jev Daily

Theo. Jev is for classifying fixed data, not the demos

The loudest Jev critic says he is stuck sounding negative about a tool he likes, and Diogo Almeida agrees the demo stream oversells it.

Theo posted that he is tired of watching people push Jev at things it is not good for, and gave a one line definition of what it is actually for. In his post he described the job as "classification of fixed data with mostly known quantities", and said his worry is that silence reads as endorsement.

Theo - t3.gg
@theo
X
classification of fixed data with mostly known quantities
Sep 26, 2026 · View on X

"I'm scared if I don't call out the stupid ideas, people will think it's useful for those cases," he wrote. He framed the whole thing as reluctant. He called Jev "legitimately awesome" and said he is sad he has to appear negative toward a thing he likes.

The reply

Diogo Almeida replied agreeing with him, and added the failure mode in one line. Posting outlier demos, he wrote, leads to over-promise and under-deliver. That is two voices, not a survey, but it is the first time the critique has been about the demo stream itself rather than about any single result.

Theo also said he is hopeful there is a saturation point coming, where the average developer understands the scope without being told. Nothing in either post names a specific demo or builder, so this is a complaint about the genre rather than a takedown of anyone's project.

For people building on it

The useful part is the scope line, because it is narrower than what the feed implies. Fixed data and mostly known quantities rules out a lot of what gets posted, and it is consistent with Theo's earlier arguments. He has previously said Jev cannot validate work because it cannot run tools, and he spent $1,000 on a Jev Router benchmark that came back almost 5x slower than the alternative.

Neither post carries a new number, so treat this as the position of a critic who has run the thing, not as an eval. If your use case is classifying data you already have, that is the case he says works. If it involves acting, verifying or long horizon steps, he is telling you the demos are not evidence.

Diogo Almeida
@CompleteSkeptic
X
Posting outlier demos leads to over-promise under-deliver
Sep 26, 2026 · View on X

Know what Jev is good at before you bet on it

One email a day. Under three minutes. What builders found out about Jev yesterday.

Get the next one by email

Jev, read daily so you do not have to. The builds, the benchmarks, the criteria that worked and the cases where it lost, from the people shipping on TypeSafe AI's System One model.