Writing C++ with Jev by picking valid next tokens
The author experimented with TypeSafe's Jev decision/classifier model for C++: at each step it was given only grammatically valid next tokens to pick from, avoiding free-form generation.
ANN Dedup Experiment with Jev: Quality Up, Costly
Jonathan Ellis shares his ANN-powered deduplication experiment where Jev classifies top-k candidates. Pros: ~10k lines of code removed and 30 subtle bugs fixed caused by near-duplicates. Cons: not cheap or fast; ~$166 API cost for 207M Opus tokens and nearly 8B Luna subagent tokens (via mjolnir).
Results of my ANN-powered deduplication experiment, with top-k candidates classified by Jev. Pro: -10kloc, fixed 30 bugs where one near-duplicate was subtly broken Con: not cheap and not fast. $166 at API pricing for 207M Opus tokens and nearly 8B in Luna subagents (via mjolnir
· 0 likesOpen on X
More like this
Using Jev for Code Review
The author thinks code review is a really good use case for using Jev.
Fascin: A Jev-Powered PR Triage Tool
A developer shares Fascin, a PR triage tool built with Jev that assesses urgency, change type, and merge readiness, noting Jev handles the actual reasoning.
