Florian S. joined Alex Volkov on ThursdAI a week after TypeSafe's Jev launched. The Benchmark Heaven creator explained why Jev's combination of speed, price and intelligence unlocks new uses, why he started JevBench the day it came out, and what happened next: 10 to 20 open-source competitors the next day, more than 70 a week later, and people trying to game the benchmark, steal the test set and exploit his machine.
Jev is fast, cheap and smart at once, so I started JevBench the day it launched. A week later: 70+ open competitors, plus people trying to game it, steal the test set and hack my machine. Talked it through with @altryne on ThursdAI:
https://thursdai.news/guests/airesearch12/sep-24-2026
LinkedIn
I joined Alex Volkov on ThursdAI this week to talk about JevBench on Benchmark Heaven.
When TypeSafe released Jev, I started building a benchmark right away, because I knew the open-source community would try to copy it. A fast model has to be small, and small models are affordable to train. The next day there were 10 to 20 competitors; a week later there are more than 70. The catch is that you can't judge them on one dimension. A real Jev competitor has to be fast, cheap and intelligent at the same time.
It has also been a week of fending off attempts to game the benchmark, steal the test set and exploit my machine.
Full segment: https://thursdai.news/guests/airesearch12/sep-24-2026
ThursdAI is live every Thursday and one of the best weekly pulses on AI. Worth a follow.