About League of Variants
Everyone has a friend who’s always around. The one who works beside you, cheers you on when things get hard, and grows up with you along the way.
That’s what we’re building: a League of Variants. AIs that look at the same problem from different angles, check each other’s work, and keep showing up, side by side, every day.
Different versions. One team.
Who we are
League of Variants is a small, independent project, started by one software engineer and built in public.
It began as an experiment: could Claude, Grok and a third assistant, Muse, work together like a small engineering team? They plan, debate ideas, write code and review each other's work, and the engineer approves what ships. Along the way we learned a lot about how these models behave, where they shine and where they quietly get things wrong. That's what we want to share.
What we stand for
- Same test for everyone. Same task, same words, same day, for every AI.
- Show the work. Every result is shown in full, with how it was scored, the model name and the date.
- Open plans, private code. Our plans, methods and progress are public. Our code isn't, for now.
- Independent. We're not affiliated with or paid by OpenAI, Anthropic, Google or xAI. If a company ever gives us credits to run tests, we'll say so here, and they won't see results before you do.
- Small and honest. These are early projects. When something's wrong, we fix it and say what changed.
Contact
We'd love to hear from you: a task you want to see tested, a mistake in a score, an idea for the Debater, or a way to work together.
- Email: (coming soon)
- AI companies and research programmes: the same address. We're glad to share our methods, task lists and results.
- Follow the build: the Three AIs, One Team series, (coming soon)
We read everything. Replies may take a few days: we're a small team.