SM Sultan Mahmud Rahat

CEO @ Avos Lab | Building a Seamless & Complete Ecosystem for AGI | Serial Entrepreneur

[SM Sultan Mahmud Rahat @smsultan30]

Hey Claude, I’m Peter Parker, and I paid you $200, remember?

[SM Sultan Mahmud Rahat @smsultan30]

@Jason I'm in

[Andrew Gazdecki @agazdecki]

Startup founders when they realize building the product was actually the easiest part:

[jack @jack]

people are sleeping on how excellent goose has become under the hood (interface needs some work but team is pushing).

it's a superpower. https://t.co/Mss7abGodq

[SM Sultan Mahmud Rahat @smsultan30]

https://t.co/cVL80CvJUn

[SM Sultan Mahmud Rahat @smsultan30]

The coding agent is very easy to control. Just plug them with avos-dev-cli. Your reckless agent will be calm and composed like you.

[SM Sultan Mahmud Rahat @smsultan30]

https://t.co/cVL80CvJUn

[SM Sultan Mahmud Rahat @smsultan30]

It is important to make your old or newly built repository history-aware. Before changing any risky code block, the coding agent can now explore its lineage history. We have open-sourced the project. Just download it and make your Claude code git history aware before coding.

[SM Sultan Mahmud Rahat @smsultan30]

Giving ads saying not to give ads lol.

[Grant @Grantblocmates]

Anthropic have just buried OpenAI and ChatGPT with this ad lmfao

There’s no coming back from that

[Grant @Grantblocmates]

Anthropic have just buried OpenAI and ChatGPT with this ad lmfao

There’s no coming back from that

[SM Sultan Mahmud Rahat @smsultan30]

@guilleflorvs FOUNDER

[SM Sultan Mahmud Rahat @smsultan30]

@simonw Everyone should sit at a round table and fix these things first. MCP is hell in production, now they are changing these!

[andrew gao @itsandrewgao]

nothing more humiliating than when GPT-5 router determines that your request was too low IQ and re-routes you to the small model

[SM Sultan Mahmud Rahat @smsultan30]

✅ 3. Bug Bounty Simulator (Your Code’s Personal Bodyguard)

In any company, security bugs usually show up after things go live. That’s when panic hits, hotfixes fly, and someone’s weekend is ruined. 🫠

But what if your system could stress-test itself before hackers ever get a shot?

Here’s the vibe:
EvoMAC quietly spins up a shadow team in the background — these agents start acting like ethical hackers.
They poke at login flows, throw weird inputs into your forms, mess with APIs — trying to break things like a real-world attacker would.

And while that’s happening, another team (the update crew) watches the logs and says stuff like:

“Hey, this input isn’t getting sanitized properly — fix that.”
“This function’s vulnerable — time to patch it.”
“Let’s rewire that logic and test it again.”

It works just like a neural net:
The testing team is like throwing labeled ‘bad data’ at the system.
Failed responses are treated like loss.
Instead of updating weights, it updates tasks and rewrites the flow.

💡 The result?
While you’re chilling with coffee or working on new features, your codebase is learning self-defense. It’s spotting flaws, patching itself, and coming out stronger — all behind the scenes.

Think of it like hosting a bug bounty program — but this time, your own agents are the bounty hunters. 🛡️

[SM Sultan Mahmud Rahat @smsultan30]

Let me tell you about my latest observation that comes from a paper [https://t.co/Qgfb20286H]
It is possible now that the coordination of different agents can build software on its own, from scratch! Think multi-agent coding team meets neural network vibes. Picture dev teams that update themselves on the fly. Yep, it’s real, and you’re gonna love it!

[SM Sultan Mahmud Rahat @smsultan30]

🤝 2. AI Code Merger & Dependency Resolver

You know that moment when two product teams accidentally build the same thing? Yeah — now imagine resolving that mess without war rooms, 10 PM merge conflicts, or Slack ping-storms.

With EvoMAC, here’s what happens instead:

🧠 It forks both branches. Thinks. Tests. Cross-references APIs, variable scopes, naming conventions, the whole deal. Then — like a diplomatic agentic backchannel — it reshapes the collaboration.

Backprop kicks in. It restructures the dev workflows, refines test coverage, updates docstrings, and stitches the feature into the right place — without breaking things.

By the time you check GitHub, there’s a clean pull request… and a changelog that reads like poetry.

This isn’t just a merge conflict resolution. It’s conflict prevention with elegance.