OptiXon

@built4us @waitin4agi_ tried to go solve that but needed GPUs, for ocr n training model from scratch....kinda expensive...plan is ready but no GPU 😢

@waitin4agi_ yup, the very skills we need will be math, physics, and writing >> coz rn ai sucks at writing... Any solution you would propose?

We’re in the golden era for math, science, and progress thanks to AI. Intelligence is no longer the limiter. Grateful to live through this era and watch it all live.

There will definitely be changes in AI, but it's getting addictive...rn everyone is just shipping code without actually learning the core skills or trying to solve problems on their own... Once the entire dev community reaches a point where they literally can't code without assistance, big tech will absolutely jack up the costs for their frontier models. We're setting ourselves up for massive vendor lock-in. lets see

@built4us @waitin4agi_ exactly, we can still see the shift and AI is quickly moving from a luxury to an absolute necessity in daily workflows.

For a short while (last year or so) the vibe coding tools were useful for two reasons: 1. Better visual sense 2. Handling the headache of deployment and hosting (I mostly see them as hosting services) Opus and Sol have bought caught up on design sense and ChatGPT now has sites for hosting. If they just add domains and scalable plans then I’m not sure what advantage vibe coding tools have.

@built4us man, exactly.. mindlessly dumping raw logs into chat is a terrible waste of tokens n limits, targeting the upstream origin programmatically is very smarter than just blinding doing back and forth guessing.... glad someone finally tackled this loop!!

Yeah, that's the reason I built this, yk we build new features and hit a lot of errors, and we just used to copy n paste to other AI tools or just to our AI tool where we are working on and we just keep copy-pasting things saying " got this or fix this "isn't it? and burn a lot of tokens, and eventually hit the limit and in the coming times, due to inference cost, every single one will think twice before burning tokens

setting up an MCP server used to mean: find the config file for your client, paste a JSON blob, go find your api key on some website, paste that in too, restart, hope. we deleted all of it. npx -y @debugai/mcp setup browser opens, you confirm a code, done. every client it recognizes gets configured automatically.

@waitin4agi_ w bro

My favourite internal project was UTA (the trailer) and my favourite external project was the stuff we did with One8 (content for their event with Virat Kohli) which was on everyone’s feeds!

In June, across the Aeos Group we did ~1.5B views, mostly in English.

@built4us Nothing worse than chasing a runtime crash, pasting stack traces into Claude, and getting hit with 'message limit reached' right as you're halfway to the fix. If this actually traces the upstream cause right in the editor, will def giving a try...

not claiming this replaces reading your own code. claiming it replaces the five minutes of grep before the reading starts. npx -y @debugai/mcp setup free tier, 10 debugs a day, no card: debugai.io/start?utm_sour…

@built4us thats really cool

what you get out of it: your coding agent stops grepping. one tool call with a stack trace comes back with root cause, exact file and line, and up to 3 fixes as ready-to-apply edits. not "here is what the error means." the file that caused it, which is usually not the file that threw.

@built4us @thsottiaux idk how are they managing at 1 million users ig

@thsottiaux You are crazy, man!! What about inference cost?

Back at the laptop. The usage limits have been reset for all paid users of Codex and ChatGPT Work. Weeeeeeeee. It's a good day!

@built4us @thsottiaux

@thsottiaux Comments are hilarious 😂

Spotted Codex making friends at the OpenAI office

@built4us @Kimi_Moonshot

@Kimi_Moonshot btw W for Kimi K3

Releasing the model weights and technical report of Kimi K3. Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window. New model architecture: 2.5x the intelligence per unit of compute, not just more params. Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale. Model weights: huggingface.co/moonshotai/Kim… Tech report: github.com/MoonshotAI/Kim… Tech blog: kimi.com/blog/kimi-k3

5 weeks left on the $9 founding rate. The roadmap it locks in: browser-side capture for the bugs that never throw... The dead button...The silent 404... Building it right now... $9/mo for life until Sep 1: debugai.io

@built4us thats serious

i was thinking about who will be the next education minister n some threads connecting to Raghav Chadda, coz they won't select a neutral minister, so this is the option left for them...if this is true, then again we got nothing n eventually they wanted the protest to be ended.. hell no @Cockroach4India @abhijeet_dipke @RahulGandhi @dhruv_rathee @UFbySamdishh whats ur thought?

thats true

go check it out debugai.io

@built4us thats really great brother!!!!

we built a debugger that asks to be graded. the grades go into the same memory your team's confirmed fixes live in. npx -y @debugai/mcp

@RahulGandhi the govt must collapse....every single sector has the same corruption n issues

भारत के छात्रों और प्रधानमंत्री को मेरा सीधा संदेश...

@built4us already in...

Time is clicking...6 weeks until the $9 founding rate is gone. $9, for life, until sep 1: debugai.io NOTE: Existing users can claim the offer, but if you want to be part of the founding users, DM or RT...