@thsottiaux Unlock subagents from 4 to 8, this would boost productivity for anyone my 50%.
Options:
Maxim 4 of sol. Maxim 6 of terra. Maxim 8 of luna-or 12 if compute is not a problem.
This would be a game changer 😁
@OpenAI I built ToolBraid alone, self-funded, and without working in tech. AI helped me move faster, but the idea, product decisions, testing, and final responsibility are mine.
AI agents can use the web. But who controls what they change?
I built ToolBraid: a browser-native WebMCP control plane where agents use live web tools while humans retain exact authority over every mutation.
Solo-built with @OpenAI ChatGPT Sol + Codex. ↓
@OpenAI Why it matters:
Useful AI agents need more than mouse control. They need provenance, drift detection, replay protection, honest receipts, and a hard boundary where human authority cannot be automated.
@wintersbit@latentblu@babayagatwt OpenAI usage is the most generous on Plus plan compared with other providers. Now you should know how to use your usage. Optimise your codex aka agent.md you will be surprised how much it makes the difference 😁
@SiCedricXD@thsottiaux Try to make your own version of that MCP. SOL is quite good at it. Ask him to adapt that mcp for himself and maintain the full functions.
@SiCedricXD@thsottiaux True. Not going back to Claude. OpenAI are very generous with usage limits on x20 plan. And sol on ultra it's better then opus 5 and not that lazy. It just execues, not pushing back unless is very important. And the most important...it doesn't send me to sleep 🤣
IT OUT! AEON-ULTIMATE-UNCENSORED-NVFP4 with GDNFP8 and LLMC
FULL MULTIMODAL PRESERVED
Mix of BF16, FP8, and NVFP4 where it counts most
Final Prod release will be even higher quality AWQ but this one is running so good I want to get it in your hands NOW
https://t.co/FwvpOiWV8A
@iTzShmeedy@SiCedricXD@thsottiaux AI experiments that involve activities such as running simulations, fine-tuning models, and optimizing performance metrics and some other.