Show HN: Pit Claude, Codex, and Gemini against each other, and apply the best

2 points • languid-photic • 1 day ago • 0 comments

Voratiq is an open-source (MIT-licensed) terminal-native CLI that lets you run multiple coding agents against the same spec, review their diffs side-by-side, and use whichever implementation is best.

It's designed for experienced developers who want to make the most of agentic coding. It positions you as the architect and reviewer, and shifts implementation onto an ensemble of agents who (hopefully) in aggregate can do a good job. And in my experience, the latest generation of models do quite well.

I didn't think "agentic coding" worked until recently. I first tried it in February. The UX was really powerful, but the models weren't good enough, so it didn't matter. Then, I gave it another try several months later, and the experience was significantly more productive. And this is just the start, the models will likely continue to improve over time.

Why take an ensemble approach? Because there is no one "best LLM for X". The top performer for your (likely highly narrow and contextual) task is impossible to know a priori. Tokens are cheap relative to human engineering time. So, just throw compute at your problem. Run all of the models. Then decide who did the best job and use that output.

Voratiq is built for "pro" users, so everything is local, configurable, inspectable, and hackable. We think these users have the skills to get the most out of these new tools. Furthermore, since the field is evolving so quickly, the more adaptable the product is, the better.

And with that, I hope you give it a try and it proves useful to you: https://github.com/voratiq/voratiq

Thank you!