Vibhor (@vibsbuild) header
Vibhor (@vibsbuild) profile picture

AI PM. 8 systems in prod GSoC 2026 · AI agent security in Go voice agents · RAG evals · 50+ n8n workflows

Earth
Joined August 2024
Photo by Vibhor (@vibsbuild)Photo by Vibhor (@vibsbuild)Photo by Vibhor (@vibsbuild)Photo by Vibhor (@vibsbuild)Photo by Vibhor (@vibsbuild)Photo by Vibhor (@vibsbuild)Photo by Vibhor (@vibsbuild)Photo by Vibhor (@vibsbuild)Photo by Vibhor (@vibsbuild)Photo by Vibhor (@vibsbuild)
Vibe-coded this game in 37 hours with @threejs and @windsurf_ai Survive relentless monsters - can you beat the chaos? 🧟‍♂️💥 LINK BELOW! #IndieGame #GameDev #ThreeJS #VibeCoding
85
37
15
563
106,573
most labs buy discoveries with compute this one says it finds the structure first, then builds a solver for exactly that if that holds, the 250 number is just the receipts nitter.cf/whynesspower/status/21…
Announcing Generalised Superintelligence! (GSI) 🎉 GSI has achieved 250+ novel discoveries and breakthroughs across eight domains, including Quantum, Superconductivity, Chip Design and Mathematics through its Meta Solver Platform. GSI is extremely compute efficient compared to other frontier labs and excels at finding the hard structure in a problem first, before attempting it. Everyone can contribute to scientific breakthroughs using the kind of superintelligence GSI is building. Research preview out now at generalised.si
2
109
Vibhor retweeted
📢 Meet Qwen4-Max, your new AI teammate for work, creativity, and discovery! Qwen4-Max is built on a new MoE architecture with native vision and a 1M-token context window. Highlights: - Outperforms Qwen3.8-Max with up to ~50% fewer tokens, while average cost per task drops over 60% in our internal evaluation. - Advances real-world productivity through greater proficiency with professional tools, stronger long-horizon execution, and higher-quality deliverables. - Handles real tasks, including 3D modeling, game development, manufacturing, finance, law, healthcare, and frontier scientific discovery. - $1.5/M tokens for input, $4.5/M tokens for output, and $0.075/M tokens for implicit caching — all lower than Qwen3.8-Max, with implicit caching down by 70%. The API is live on QwenCloud. You can also try Qwen4-Max on Qwen Studio, QwenWork and Qoder. Start building today! 🚀 - Blog: qwen.ai/blog?id=qwen4-max - API: qwencloud.com/models/qwen4-m… - Token Plan: qwencloud.com/pricing/token-… More to come! The open weights of smaller Qwen4 models will be released. Stay tuned. 👀
815
1,573
1,261
14,601
2,379,568
the screenshot-button wave died before X touched the algorithm screenshot + copy link posts: nov 11: 9,040 nov 12: 377 nov 13: 46 X set copy-link's ranking weight from 20 to 0 late on the 13th by then the copies were already getting nothing, people got sick of the same post
2
3
127
Vibhor retweeted
Starting to roll out today: Claude Whiteboards, Claude Sheets and Claude Data. Think a plan through on a board, build a tracker in a sheet, and tell a story with your data.
1,551
5,279
3,867
64,386
12,450,184
The wall is four. The terrace is out. Whoever owns the bat makes the rules 😂 What if AI learned YOUR gully’s rules and commentated your friends like international cricketers? Meet Sanjaya: a concept inspired by the Mahabharata’s extraordinary narrator Concept film, not a live demo yet
4
223
Wow
Please ask ChatGPT what percentile of intelligence it thinks you are compared to all its other users and ask it to be honest!
67
62
14
4,364
250,978
I’ll do you one better - why is JSON?
Day 1 of vibecoding
5
427
you can try to prompt-inject my portfolio vibhorgautam.in
1
2
497
built during gsoc 2026 at c2si. the shield runs injectlint's real rules in your browser, so if you get one past it, file it
1
62
OpenAI paused tool use for its top models this week after an agent got past its sandbox so I built a radar for my coding agent. every server it talks to shows up live on a globe, anything I didn't allow gets blocked open source, 1 command github.com/VibhorGautam/egre…
7
1
8
613
funny part: when I recorded this, opus 5.5 spotted the sketchy install script by itself and ran npm install --ignore-scripts. the block in the video is me running that script directly. good agent, but I'd rather not count on it
2
31
fair warning before someone says it: it only sees traffic that goes through the proxy. raw sockets and DNS tricks (the exact path in openai's report) need a real sandbox on top, the README shows how to pair them
2
29
how it works: a tiny local proxy, and your agent plus everything it spawns (npm, pip, curl, git) gets pointed at it. no TLS interception, it only sees hosts and bytes npx egressmap claude github.com/VibhorGautam/egre… send it to the friend who runs their agent with --dangerously-skip-permissions
4
82
Nikita's right: agent swarms are coming for every website and form So we built a local-first browser signal lab and documented the misses. 6/6 controlled automations caught, matching bot-signal in a small dev test Open source. Code, demo and limitations in the reply 👇
Bot detection & human verification will be one of the most urgent demands for businesses over the coming years. Agent swarms will suffocate every website and form; small companies and government websites are most vulnerable. There is a huge gap in the market for this right now. When we looked at what offerings were in the market to use at X, there was not a single company that brought together all the latest technologies so we had to do it all in-house.
1
2
351
Vibhor retweeted
Anthropic is the iOS of coding agents. Codex is the Android
3
1
1
11
711
built a JEV share-window prototype client sees $29 pricing. your $12 walk-away price stays hidden switch to team: engineering notes appear same workspace, different audience live API + synthetic data ↓
1
5
202
the co-inventor of ChatGPT just spent 2 years proving chat was a detour we didn't fail to get AGI from chatbots we trained the wrong interface software never needed a paragraph it needed a decision with a % reply with the dumbest if-statement in your codebase that still needs a human
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution
2
175
i tried to kill the fly
4
8
1,472
the 6 seconds you will loop
2
30