A small lab, chasing a faint but real signal.

What we do

We run small, careful experiments on how modern AI systems behave — and we write up what we learn. Our focus is less on building the largest model and more on understanding the ones we have: where their reasoning holds, where it breaks, and how to tell the difference reliably.

What we believe

Three things shape how we work:

• Depth over breadth. A few questions, pursued patiently, beat a long roadmap pursued shallowly.
• Reliability over spectacle. A system you can predict is worth more than one that occasionally dazzles.
• Open by default. We publish methods, results, and the experiments that didn't work.

The name

An anthelion is a luminous spot that appears opposite the sun, formed as light refracts through ice crystals high in the atmosphere. It's subtle, real, and rewards careful looking — which is roughly how we feel about the questions worth working on.

Get in touch

We're always glad to hear from researchers, collaborators, and curious readers. Email us at hello@anthelic.ai, follow along on X and GitHub, or subscribe to get new writing in your inbox.

---

This page is a starting point — edit the copy in about.html to add your team, location, or anything else you'd like visitors to know.