I spent $3.40 on Jev in 24 hours: it will be Jev + LLMs, not Jev vs LLMs

gaganghotra_ · x · 2026-09-19

The author stress-tested Jev, burning $3.40 in 24 hours, and concludes it will be Jev + LLMs rather than Jev vs. LLMs.
Jev-like systems will likely be embedded into all LLMs for automatic model routing, small decisions, if/else queries, retrieval, tool selection, and scoring — tasks where LLMs waste tokens thinking and planning with smarter-but-slower models.
Other upsides: learns from a single prompt what LLMs need fine-tuning for; instant decisions with no thinking blocker; scores probabilities instead of reasoning in prose; zero output cost since it only emits decisions.
The full pros-and-cons list is in the original thread.

More from coding & agent