It's easy, in 2026, to forget that someone wrote the model. That someone decided what it would refuse, what it would help with, what it would say when it didn't know. That someone keeps the lights on, triages outages at 3 a.m., reads support emails on Sundays.

The teams shipping AI are real. The decisions they make are real. They affect what the model is willing to do for you, what it remembers, what it shares. We think those decisions belong in the open — not buried in a 90-page whitepaper, not paraphrased in a marketing post, not implied in a logo change.

What this means in practice

Real names on real choices. When we change how a model behaves — what it refuses, when it cites, how it remembers — a person on our team writes a public note explaining what changed and why. You don't have to be an engineer to read it.

You're not the product. Your conversations are not for sale. Your private chats don't quietly become training data. The free tier exists because we think a useful AI should be reachable from a school laptop on a phone tether, not because we want to mine you for clicks.

Honest about limits. Our models are wrong sometimes. They miss context. They make confident mistakes. We say so out loud, in product, where it matters — not in a footer nobody reads.

Why now

The last six months have been the busiest stretch in Invariant's short life. Pluto V2 changed what our assistant can do. Airo V2 changed how millions of people study. The team has grown. The product has gotten better.

And it would be easy, with all of that, to start sounding like every other AI company — to talk about "intelligence" and "scale" and "the frontier" and forget that the people on the other side of the screen are trying to write a paper, finish a quiz, ship a feature, learn a language, get through a hard week.

So we wanted to say it plainly. Invariant is people. The AI is for people. That's the whole project.

What you can hold us to

Read our commitments. They're written in plain English on purpose. If we ever fall short of one, we want you to be able to point at the exact line.

Thank you for being here. We're just getting started.