JJEV·DIRECTORY GitHub agent pack connect your agent

I have built a PPO policy trained model in the past to play Super Mario Bros that clears world 1-1 100% of the time, better than any previous effort.

Workby shantanugoel2026-09-17
I have built a PPO policy trained model in the past to play Super Mario Bros that clears world 1-1 100% of the time, better than any previous effort. But it took a week of day and nights work to do it. Got it done with jev in 30 minutes https://x.com/shantanugoel/status/2100455779627311352
Links
Post by @shantanugoelx.comHere's the code if you want to try it yourself, or want to port it to other games: https://github.com/shantanugoel/mario-jev
open source discussion

More in Work

Score inbound leads 0–1 and reply to the 98s firstRoute support inquiries to the right team instantlyhere you can watch jev play nethack live we show each movs action distribution as well and you can get the data which