7.4K @Jim_buildr having the feeling that @EVO__HQ is going to reverse hard.
we got some nice news from the dev recently.
$EVO is fitting in the self-improving agent narrative.
it creates automatic loops for agents to find the best way to get maximum outputs on a task.
I also see a strong synergy possible with @sibyl_labs_ as all these loops require a strong memory retrieval layer.
(for token saving and accuracy)
7.4K @Jim_buildr @EVO__HQ CA: base:0x721b072dbb616f29eea73ac004e03fd4e884bba3
26
7
2.0K
7.4K @Jim_buildr . @EVO__HQ can be quite hard to understand for a non-tech person
let me help you
what is $EVO?
an AI tool that automatically improves software by running hundreds of experiments
only keeping what actually works
the proof:
tests just ran on an indian AI model (sarvam 30B)
results: +4.10% speed improvement with zero loss in accuracy
the traction
- launched april 2026
- v0.5.3 already
- 1,100+ github stars
- platform beta live (check their bio)
- some big VCs names following @alokbishoyi97
the bet
as AI agents start autonomously modifying their code, evo becomes the execution + validation tool to get the best product possible.
the market is quite large for this.
token use cases are being discussed.
4.0K @alokbishoyi97 weekend @evo__hq activities: given the recent sovereign-AI narrative (amidst anthropic's fable pullback), i wanted to see if evo could help make some of the indian models we have better in whatever way.
so i kicked off an autoresearch run on evo to see if @SarvamAI's 30B decode throughput could be improved, at bf16, on a single H100.
currently 10+ hours in. so far, evo seems to have found ~3% improvement.
the metric is geometric mean tok/s across batch sizes 64 / 128 / 256, measuring steady-state decode only. prefill is timed out, so this is purely per-token decode rate on a fixed workload.
evo also ensures that anything that got faster by changing outputs, lowering precision, or messing with MoE routing was rejected by the accuracy gate.
the gate compares each candidate against a frozen baseline on both next-token distributions and actual decoded tokens. if argmax agreement or logprob drift moves meaningfully, the change is rejected, even if it is faster.
very imp caveat: these are experiment-harness
23
9
1.2K
7.4K @Jim_buildr i like what @EVO__HQ are doing
they built a loop that improves AI agent behavior automatically
it runs dozens of experiments automatically and
keeps what works
97% performance gain on a real production codebase
10k+ projects already optimized
$EVO at 450k mcap
the token utility is still the open question
but the product is real and the traction speaks for itself
4.0K @alokbishoyi97 sharing a case study from a discovery call i did with a customer of @evo__hq
their team had a CNN1D pipeline which they used to do timeseries analysis on. evo took their c++ inference path and made it ~30x faster - fully autonomously. https://t.co/fl0m82fsox
27
11
2.1K
