DEV Community: Abhinav Sharma

CodeAtoms is now an AI Agent marketplace

Abhinav Sharma — Sat, 31 Jan 2026 10:32:19 +0000

I want to be honest about something.

When I started CodeAtoms, the idea was simple:
a place where developers could find useful tools, resources, and experiments around.

That made sense back then.

But over the last few months, while actually using these systems every day, it became clear to me that tools aren’t the main thing anymore.

Agents are.

Trend is now changing.

Most developer tools today still work the same way:
you click, configure, run, repeat.

But the trend is now changing, people are now preferring ai ides, ai agents over traditional tools.

AI Agents

You give them a goal.
They figure out steps.
They execute.
They fail, retry, adapt.

That feels less like using software and more like working with a junior teammate.

Once you experience that, it’s hard to go back.

Why I changed CodeAtoms

I realized something pretty simple:

If developers are going to rely on agents, they need:

clarity on what an agent can and cannot do
visibility into how it works
control over when it runs and what it touches

and most importantly SECURITY

So CodeAtoms is shifting focus.

From today, CodeAtoms is an AI Agent marketplace.

Just a place where agents live, are tested, compared, and actually used.

This is still early

I’m not claiming this is perfect or finished.

Agent ecosystems are messy right now. Standards are unclear.
Everyone is experimenting.

But that’s exactly why this shift matters now, not later.

Someone has to build the place where this all becomes real.

I want CodeAtoms to be that place.

After AI coding Agents, What’s actually next? First everyone moved to Copilot. Then Cursor blew up. Then suddenly it was all about AI agents Claude Code, Gemini CLI

Abhinav Sharma — Fri, 23 Jan 2026 09:04:47 +0000

First AI IDEs then AI coding agents… what’s next?

Abhinav Sharma ・ Jan 23

#programming #ai #opensource #discuss

First AI IDEs then AI coding agents… what’s next?

Abhinav Sharma — Fri, 23 Jan 2026 09:03:03 +0000

AI coding agents are improving fast. Writing code, running commands, and taking multi-step actions inside IDEs is already normal.

Modern IDEs and tools can show what an agent attempted and the steps it took. That part exists.

But after using these systems seriously, another gap becomes obvious.

What’s missing is guardrail-level observability and control.

Developers need a clear, unified view of everything that happens inside a workspace not just agent steps, but all actions, whether they come from a human or an AI agent.

The goal isn’t just visibility.
It’s control through visibility.

We need logs that answer questions like:

What task was the agent working on?
What actions did it take during that task?
Which actions were allowed, blocked, or required confirmation?
Which actions came from the user vs the agent?

See the GitHub issue in CrewBench for this Functionality here

This is less about IDE UX and more about system-level guardrails.

Tools like Claude co-work are a strong step forward in collaboration, but developers who care about safety, debugging, and trust need something deeper: action-level logs tied to rules, permissions, and task boundaries.

When something goes wrong, it’s not enough to see the final code or even the agent’s steps. Developers need to understand how decisions were made and why certain actions were allowed.

That’s why the next step after AI coding agents isn’t more autonomy.

It’s better observability and better control.

This problem is being actively discussed and explored in the open:

GitHub issues around agent action logging
Guardrails and permission-based execution
Unified logs for user + agent actions

If you’re thinking about these problems too, you’re already looking at what’s next for AI agents.

Working with AI Coding Agents Made Me Lose Track of File Changes

Abhinav Sharma — Wed, 21 Jan 2026 09:19:43 +0000

Hey Abhinav Sharma here...

One thing that keeps bothering me when using AI coding agents

I’ve been using AI coding agents more seriously lately, not just for small snippets but for actual project work. There’s one thing that keeps bothering me.

When an AI agent edits a file, you only see the final version of that file.

But in reality, that file might have been changed many times:

maybe the agent rewrote it
then fixed something
then adjusted it again
maybe I tweaked it in between

All of that history is gone. ( now agents do ask for permission to write but not talking about after permission )

At the end, you’re left with one file and a vague feeling that “a lot happened here.”

Why this feels like a problem

If something breaks later, it’s hard to answer simple questions like:

How many times was this file changed?
When did it start going wrong?
Can I go back to an earlier version, not all the way back?

Most of the time, you can’t.
You only have the final result.

This becomes more painful with long agent runs

This is why I integrated a file timeline directly into CrewBench.

For every file, I wanted:

visibility into how many times it was changed
access to older versions
the ability to roll back to a specific change (for example, the 5th edit)

So in CrewBench, each file now has a simple timeline:

every edit (by you or an AI agent) is recorded
previous versions are kept
diffs are visible
you can roll back to any earlier version

If a file was changed 10 times, all 10 changes are visible, and you can go back to any one of them when needed.

See the CrewBench Issue here

Implementation here

CrewBench UI

This doesn’t replace Git.

It just gives visibility into something Git doesn’t really show well.

Build Y: An Open-Source Playground for Breakthrough Inventions

Abhinav Sharma — Fri, 14 Mar 2025 14:06:47 +0000

Introducing Build Y by Necrozma Labs

The world is changing faster than ever. With every passing day, engineers, scientists, and visionaries push the boundaries of what’s possible. But for those who dream of building the future, one challenge remains: access. Access to cutting-edge tools, knowledge, and a community that thrives on collaboration.

That’s where BuildY was born. Created by Necrozma Labs, BuildY isn’t just another platform. It’s a movement—a space where the brightest minds come together to explore, create, and shape the world of tomorrow.

A Hub of Innovation
Imagine a place where groundbreaking ideas in AI, Biotechnology, Quantum Computing, and more are not locked behind paywalls or patents. A place where open-source innovation is not just encouraged but celebrated.

BuildY is that place. It brings together researchers, developers, and creators from across the world, offering them an ever-evolving repository of cutting-edge projects and insights.

A scientist in Japan, an engineer in Germany, and a student in India can collaborate seamlessly, sharing their expertise to create something greater than themselves. No locked doors, no gatekeeping—everyone is welcome to contribute, and everyone benefits. Knowledge is no longer a privilege but a shared resource, updated constantly to keep pace with the world’s latest breakthroughs.

Who stands to gain from this?
For engineers and scientists, BuildY is the missing piece of the puzzle—the platform where bold ideas meet the tools needed to bring them to life. Students and educators find in BuildY an ever-evolving learning hub, filled with resources that transcend traditional textbooks.

And for the entrepreneurs and innovators shaping the next wave of technology, BuildY offers the perfect ground to build, test, and refine solutions that can change the world.

This is more than a platform. It’s a call to action. A call to create, to collaborate, and to redefine the limits of what’s possible.

Usage License
The ideas presented on Build Y are intended to be free and open-source, aligning with our commitment to fostering innovation without barriers. Users are encouraged to utilize, modify, and share these concepts, provided that appropriate credit is given to the original contributors.

This open-access model is designed to stimulate collaborative development and rapid technological advancement.

Credits & Transparency
All ideas featured in Build Y are based on publicly available research and contributions from the open-source scientific community. Full credits to original researchers, developers, and contributors are provided:

Check the full credits page here: See here

We prioritize transparency and ensure that all resources are properly credited.

Submit your idea to Build Y
We welcome contributions from innovators, engineers, and scientists worldwide! You can submit your ideas to Build Y by opening a issue on our Github repository.

Graphics created for Build Y

You can visit Build Y by clicking here...

Get 6 months of FREE Notion Plus Plan and Unlimited AI through ConquerortheCrown

Abhinav Sharma — Wed, 22 Jan 2025 07:09:29 +0000

At ConquerortheCrown, we are Super excited to announce that we’ve partnered with Notion to join their Startups program!

🌟 Exclusive Offer for ConquerortheCrown Startups
As a member of our ecosystem, you’re eligible for:
🎁 6 months of FREE Notion Plus Plan
🎁 Unlimited AI features to supercharge your workflow

More details on how to redeem here - ConquerortheCrown Partnered with Notion

Join ConquerortheCrown as we provide resources and grants to people who want to do something big instead of going to college or doing a job.

Visit ConquerortheCrown

ConquerortheCrown Organization - Resources Grants for Future Makers

Abhinav Sharma — Sun, 19 Jan 2025 10:07:40 +0000

ConquerortheCrown gives resources and grants to people who want to do something big instead of going to college or doing a job.

We give resources and grants to -
STARTUP FOUNDERS
AND
EXTRAORDINARY PEOPLE

Our total grants giving capacity is $600,000
while we have given $31,500 in grants.

You can apply to join ConquerortheCrown by visiting Here or visiting https://www.conquerorthecrown.com/

ConquerortheCrown Team

Contribute to imagination-to-real in open-source sprint!

Abhinav Sharma — Thu, 16 Jan 2025 06:35:00 +0000

Imagination-to-Real transforms the way robots learn by providing them with realistic, diverse, and geometrically accurate visual data.

The integration of Generative AI with classical physics simulators helps create synthetic datasets that are rich in variation, enabling robots to train on tasks previously deemed too complex or expensive to model.

we encourage developers, researchers, and enthusiasts to contribute to Imagination-to-Real.

Contribute to imagination-to-real in open-source sprint!
Here

Contribute to `real-to-sim-to-real` in SmilingRobo Open-Source Sprint!

Abhinav Sharma — Wed, 15 Jan 2025 06:33:00 +0000

The mission of the Real-to-Sim-to-Real Framework is to create tools and libraries that empower users to train their robots in custom environments tailored to their specific needs.

The framework also enables users to create their own environments easily using just their smartphone.

Get more details here Open-Source Sprint

Start contributing and help us advance robotic policy training using the Real-to-Sim-to-Real Framework!

Open-Source Sprint by SmilingRobo

Abhinav Sharma — Sat, 11 Jan 2025 15:36:18 +0000

An opportunity for developers, students, and researchers to contribute to groundbreaking libraries and shape the future of open-source robotics.

🌟 Why Participate?
Enhance your skills while contributing to impactful solutions.
Earn exciting rewards for your contributions.

👩‍💻 Who Can Join?
Students passionate about robotics and technology.
Researchers looking to share their innovations.
Tech enthusiasts eager to build open-source solutions.

📅 Let’s innovate together and push the boundaries of robotics!

Visit Open-Source Sprint

Creating a library that will allow users to train their robots using GenAI

Abhinav Sharma — Wed, 18 Dec 2024 06:07:35 +0000

Imagination-to-real Train your robot to do whatever you want using Generative AI.

Visit on Github

Description
imagination-to-real empowers robotics developers by bridging the gap between generative AI and classical physics simulators. Our library prepares realistic, diverse, and geometrically accurate visual data from generative models. This data enables robots to learn complex and highly dynamic tasks, such as parkour, without requiring depth sensors.

🚀 What It Does:

⚪ Integrates generative models with simulators to create rich, synthetic datasets.
⚪ Ensures temporal consistency with tools like Dreams In Motion (DIM).
⚪ Offers compatibility with MuJoCo environments for seamless data preparation.

ProgrammerBuddy by ConquerortheCrown

Abhinav Sharma — Wed, 11 Dec 2024 05:11:27 +0000

This repository is a curated collection of valuable resources designed to assist programmers, developers, and tech enthusiasts at all skill levels.

Get

Create Any Tech

OpenSource Alternatives to SaaS

Public Apis

Visit ProgrammerBuddy https://github.com/ConquerortheCrown/programmerbuddy