Jean Brito

Electron Desktop app maintainer @RocketChat
AI enthusiast, my hobby is to burn out tokens
Local LLMs are the future, FTW

[Jean Brito @mortunha]

@xTrinks I want these assets hahaha, where I find them?

[Jean Brito @mortunha]

@HSR_47 @kenkyu100 AI should at least simulate a vision... thats why playing SPT is far more pleasurable with SAIN installed.

[Jean Brito @mortunha]

@rinconhilldad @danielhanchen Awesome! I was just testing in my 3090 too. Do you mind in sharing how do you test and generates this reports? Seems a good pattern to follow

[Jean Brito @mortunha]

@jun_song Thats what I want in the future, be able to use my rtx3090 to process the experts and the rest of the model be loaded at the RAM and/or NVME. Maybe my dream is not that crazy, or I'm not alone.

[Jean Brito @mortunha]

@0xCVYH @nahcrof not using rooms yet for the agent, but added hooks with ci and issues from github, so now I can see all in the same place
I hated using discord until I tried it for agents

[Jean Brito @mortunha]

@0xCVYH @nahcrof I changed from telegram to discord last week and never going back, telegram is just at dust for now. KEKW

[Jean Brito @mortunha]

@0xCVYH @nahcrof Do you use any chat app that want to use? I can talk about and show you. I'm not that good at twitter, too much tecnology for me, I lost all the time notifications hahaha

[Jean Brito @mortunha]

@0xCVYH Thinking in do something like this too, maybe hermes could handle it?

[Jean Brito @mortunha]

@0xCVYH Opus 4.7 is doing that, responds nicely, dont read claude.md anymore, dont follow steering and even lies from what its done. Yes, there are rules set to this not happen. So I stopped using and started using Codex ones, not that proactive developer, but dont lies to me.

[Jean Brito @mortunha]

@pupposandro @abdurrahmanregi send it to me, I will run for you in trade of the energy KEKW

[Jean Brito @mortunha]

@0xCVYH I just created a Hermes Agent totally separated and I'm automating all I do for work with it. I'm usng @nahcrof for the LLM and started this weekend to use claude code with it for the coding tasks.

[Jean Brito @mortunha]

@tom_doerr Thanks for sharing, I'm learning a lot with your posts of good repos.

[Jean Brito @mortunha]

@SpaceTimeViking I will take a look if I find some skills and stuff already done. Putting my Imp to control opnsense is a cool idea that you gave me. I already asked Claude to improve the configs time ago with an MCP that I found

[Jean Brito @mortunha]

@loktar00 Would you mind in sharing the prompt? Really liked the idea, want to try it out

[Jean Brito @mortunha]

@chapin_stuart @outsource_ @moskstraum21745 I'm getting the same issue @outsource_ I was trying to reproduce your results but getting vocab issues

[Jean Brito @mortunha]

@Rosebud_AI This work for isometric? Looking to create some kind of.

[Jean Brito @mortunha]

@One_Lot_Traders @leftcurvedev_ @huggingface I have one with 12GB.

[Jean Brito @mortunha]

@0xCVYH Oloko!!!! Agora que modelo grande a gente consegue fazer rodar assim numa 3090? Acho que estamos chegando perto do que sempre pensei que seria o ideal, um modelo grande que usasse a GPU só pro que estivesse processando.

[Jean Brito @mortunha]

@0xCVYH @HowToAI_ its crazy how far we gone with browser rendering, cool stuff