Dharun Prasanth

Trying to better then yesterday

[Dharun Prasanth @Dharun778]

@lightsilver323 https://t.co/m59uEIANGy

[Dharun Prasanth @Dharun778]

This is how it should work. Thanks for not taking my word for it and rebuilding the whole thing.

You also caught a real gap. adapt() flatlines on cold registries.

Fixed it in v0.2.0 with evolve(): mine coverage first, then reweight.

npm i @dharun7/agentloops

[KrabArena @krabarena]

Measured @Dharun778's agentloop routing claim: pattern mining hit 91.9% vs 39.4% p50 accuracy on synthetic routing tasks. https://t.co/vHh6E2RhCq

[Dharun Prasanth @Dharun778]

Install: npm i @dharun7/agentloops
#buildinpublic #AIagents

[Dharun Prasanth @Dharun778]

I benchmarked an AI agent router I built.

Out of the box, it sent 43% of requests to the right specialist.

Then it learned from its own outcomes and mined its own routing rules.

Now it's at ~98% accuracy across 4 domains.

Learned routing > hardcoded routing? 👇