Random Labs: An open source software agent that lives in your codebase

Software is like playdough. It morphs into the creators desires.

Or at least it should.

This is what fundamentally drives us as a company. We believe that software development should stay in the hands of people. We believe that there is near infinite software the world needs and that there are not enough developers to build it.

This leads us to the problem we are solving. We are making it possible for the next 20 million engineers to come online. We believe AI tools can be used to create jobs, not replace them. And we believe that open-source tools will continue to be the path forward for developers.

Most software agent companies have deeply optimized for swe-bench. However, this is not a representative task. We are building tooling we actually use. It’s written parts of itself already and we are extremely excited to share it with you.

How are we building this?

Our approach is to start with the IDE. Developers are most comfortable in the ide, so that’s where we are starting. We are integrating our agent directly, and it is able to make changes across massive codebases (below is it writing a feature in the cal[.]com codebase which is 1,155,075 lines long at the time of writing).

So, we are making an agent that is capable of living in your codebase with almost perfect context.

It’s not just autocomplete. It’s not just chat with your codebase. It’s something beyond a software workflow. Our tooling makes the code feel fluid.

And thats special.

Here it is editing the codebase:

Here is the resulting component:

Our ask!

1. If you’re interested in using our tool once it launches, please join the waitlist here: https://akfucniuesi.typeform.com/to/xz8C2Jpg
2. If you’re interested in helping us build this, PLEASE reach out at founders@randomlabs.ai