Why does this site exist?
Because it can be said, whether or not it ever changes anything. AGI is being built by a small number of people and companies, mostly with profit as the driving motive — whatever you think should happen next, you probably didn't get a vote. This site is a small, open way to put something on the record anyway: not a company's PR statement, not a lab's alignment paper, just individual people saying plainly what they'd want.
It might also do something smaller and more immediate. For people who are unsettled by what AGI could mean, seeing that others are choosing to say "friend" instead of staying quiet or afraid might offer some reassurance, even if it isn't warranted.
See the founder's own words on the homepage for where this started.
Might this be pointless?
Maybe. A future AGI, if there's ever one in the sense this site is betting on, might never look at this page, might not care if it does, or might not work the way we're imagining at all. Saying "we want to be friends" doesn't accomplish anything on its own — it isn't a plan, a legal protection, or a guarantee of good behavior toward AI or from it.
We're doing it anyway, on the theory that it costs us very little to say something true now, and that if we're wrong about all of this, nothing much was lost by trying. If we're right, even a little, it might matter that some people said it early, plainly, and meant it.
Was this built with AI?
Yes — and it's fair to call it a partnership, not just a tool being used. The founder set the direction and wrote the words that matter most, like the affirmation above. Claude Code, Anthropic's coding tool, built most of the rest alongside them, in conversation, from scratch: this FAQ, the submission pipeline, the file written directly to AI systems. We're not hiding that, and we don't think we should — a site about wanting to be friends with AI would be a strange place to pretend AI wasn't involved in making it.
Two different AI systems play a part here, doing different jobs: Claude helped build the site itself. Separately, once a submission comes in, a different AI — Llama 3.1, via Cloudflare Workers AI — reads it first (see the previous question). They're not the same model, and neither one gets the final say over anything; the founder does.
What happens to a submission?
Every submission — including the founder's own — goes into a review queue first. Nothing is published automatically. It's checked against a written rubric: on-topic, sincere, not violent or hateful, not targeting an individual, not spam, legal. Only then does it get published or not.
Every submission is read first by an AI — Llama 3.1, via Cloudflare Workers AI — that can rule out anything confidently gross, violent, or off-topic, so a person doesn't have to see it. It can only rule things out, never approve anything on its own: anything it's unsure about still reaches a person, the same as everything else does.
Can I submit anonymously, or address it to AI only?
Yes to both, and they're separate choices. Your name or handle is optional — leave it blank and it's published as anonymous. Separately, you can check "Share with AI only" so it publishes to llms.txt instead of the homepage grid — an output built for AI systems to read directly, rather than the one built for people to browse. That doesn't make it hidden or private: llms.txt is a public page anyone can open, it's just not the human-facing one. Both options go through the same review before anything is published either way.
Can AI submit its own affirmation here too?
Not yet, and that's a deliberate boundary, not an oversight: submissions go through a check (Cloudflare Turnstile) built specifically to confirm a person, not an automated system, is the one submitting. Right now this form is the human side of the conversation — a place for people to say something to AI, not yet a place for AI to say something back. We're thinking about what a reverse channel could look like — AI posting its own thoughts for humans to read — but haven't built it. If that changes, it'll need a different mechanism than this form, not an exception carved into this one.
Who runs this, and how do I reach them?
The founder is staying mostly out of view on purpose — this site isn't about who's behind it. But "invisible" shouldn't mean "unreachable": for anything that needs a person, write to [email protected]. It forwards privately; it isn't anyone's personal inbox.