Bharat Dudeja

💻

[Bharat Dudeja @csinye]

@ChatGPT I think the follow-up meeting will arrive sooner than Astra ultra lightning fast’s response

[Bharat Dudeja @csinye]

@vigyso Soon! Earlier I was doing this through MCP with other apps, but Sign in with ChatGPT feels like a game changer here. I’ll post a working demo soon. 🚀

[Bharat Dudeja @csinye]

@abhi16_93 Congrats on shipping! Agents pulling from a curated dev feed instead of random search results is a smart idea. I'm building an MCP browser (Skillerr), so I'm curious: which tools are you exposing, and are you finding agents call search or feed more?

[Bharat Dudeja @csinye]

@sentient_found Good list. For roundup 3, if it fits (I built it): https://t.co/2htixComiW browser that Claude Code, Cursor and Codex drive over MCP. Every page the agent opens is a real tab you can watch, pause and take over.

[Bharat Dudeja @csinye]

Free, open source, no telemetry. macOS, Windows, Linux.

One line to install, and it connects Claude and Cursor for you:

curl -fsSL https://t.co/wrFMTBvwbE | sh

A star on GitHub helps a lot 🙏
https://t.co/rwWh140WNs

[Bharat Dudeja @csinye]

Your AI says "I searched the web."

Which pages? Did it read them, or answer from memory?

I built a browser where you watch. Every search and every page Claude Code opens is a real tab, live, side by side.

[Bharat Dudeja @csinye]

And it keeps what your AI learned.

Research is saved on your computer as notes and folders. Next week, a different AI checks what you already know before searching again.

[Bharat Dudeja @csinye]

You’re building with claude. Your ai researches for hours, but what did it actually search? what did it read? what made it choose this over that?

You have no idea.

Meet skillerr. Every page your ai reads opens as a real tab you can watch. live. side by side or in claude desk.

[Bharat Dudeja @csinye]

Meet Skillerr, a browser for the Agentic Era. Watch Claude or Cursor research, approve actions, and keep trails locally. Built with Claude. Its 7.9 MB Orb embeds page titles over 250x faster than 90 MB
MiniLM (4 threads), with 98% of its tab-journey F1.
Free & open source.

[Bharat Dudeja @csinye]

Your AI’s internet browsing shouldn’t just disappear when the chat ends.

I built the Orb: a 7.9 MB model inside Skillerr that groups tabs into journeys, finds pages by meaning, and learns locally from how I browse.

8 µs per page title. No GPU. 0 bytes sent.

[Bharat Dudeja @csinye]

https://t.co/YVHX3hkIQI

[Bharat Dudeja @csinye]

@Cashvolumes Solid analysis on this XAU setup. If you ever want to test these parameters without dealing with code, I just pushed a pilot for an AI where you just chat to build the strategy. Looking for testers today, 50 free credits and 100% free forever. Break it: https://t.co/I74aHVpDct

[Bharat Dudeja @csinye]

Test it out here: https://t.co/I74aHVpDct

[Bharat Dudeja @csinye]

मैंने बोला | I spoke.

1 stand taken for transparent NEET & CBSE 2026 audit.

Join the public record →

#BharatBol #ExamAudit

https://t.co/aMTIS3vKxP

[Bharat Dudeja @csinye]

What you verify = exactly what runs. Transparent audit log.

No more blind trust. Just npx @skillerr/add [skill] → TrustView → approve.
Real security for agentic workflows.
Stop gambling. Start verifying.
👉 https://t.co/zr4H2SB6GH
#AI #DevSec #AgenticAI #OpenSource

[Bharat Dudeja @csinye]

Running random AI-generated code or prompts?

You’re playing Russian roulette with your filesystem, network, and secrets. 🔥

One bad skill and boom data leak,
broken build, or worse.

Skillerr changes the game with sealed skills + TrustView.