@BohuTANG: I'm doing something similar too. Jev is still too slow in this kind of scenario.
Summary
A discussion about using Jev for query optimization, with one user sharing a 12% speed improvement in Postgres queries and another commenting on Jev's performance issues.
View Cached Full Text
Cached at: 09/19/26, 11:01 AM
I’m doing something similar too. Jev is still too slow in this kind of scenario.
Michael Malis (@mmalisper): I used Jev to build a query planner! With some tuning, I sped up Postgres queries on the join order benchmark by 12%. Here’s what I did
Similar Articles
Jev vs. Kev: open-source Jev alternative tested side by side
The article compares TypeSafe's Jev model with the open-source Kev alternative, testing their accuracy, token usage, and speed on a new dataset to avoid data leakage, finding similar performance but differences in implementation.
Been experimenting with Jev — interesting approach to AI agents
The author discusses experimenting with Jev, a tool for AI agents focused on decision-making, which claims significant speed and cost benefits compared to using large language models for all tasks.
Journey in optimising Elixir application
A developer shares lessons learned while optimizing Elixir applications, particularly focusing on performance improvements to a Postgres connection pooler (Ultravisor). The article covers profiling techniques using flame graphs, call tracing, and tools like eFlambè and tprof.
@NFT_Chen: Awesome! Let Jev self-optimize and solve complex problems on its own—no more hand-writing instructions, criteria, or ca…
JevHarness is an open-source tool that allows LLMs to automatically generate and optimize task-specific decision flows for Jev, enabling self-optimization and improving performance, as shown in a Pokémon AI battle win rate increase from 25% to 75%.
@taroleo: When calling Jev from the US West Coast, a single request compiling 6 questions takes about 130 ms, or 20-25 ms per jud…
The tweet highlights the low latency and scalability of calling Jev from the US West Coast, with 130 ms per request for 6 questions and constant latency under high parallelism, indicating good design and future potential with specialized models.