Same promises.
No exceptions.
Human Again is open to other people who make software with AI and want it on people's side. Useful apps of every kind are welcome — a timer, a reader, a notebook, a game, a tool for a real job. The one bar is that your app is at least as innocent as ours.
A lot of software made with AI is made carelessly, and people are right to be wary of it. We'd rather answer that with a standard anyone can check than with an argument.
Why the rules exist.
These apps exist to help people be the best, kindest versions of themselves, and never to exploit them. The nine rules below serve that.
Slipstream, Glimmerwood and Westering are the measure: the code allows what they do, and no more. Where the rules are silent or unclear, the promise decides, always for the person using the app.
Nine promises, checked in the open.
- On people's side.
Made for the people using it, and never working against them or anyone else. It doesn't lie to, exploit or harm anyone. Any nudge is in the open, towards what the person already wants. It watches no one but the person using it, and only after telling them plainly. No design tricks, no random rewards, nothing there only for a limited time, and no betting money — or anything bought with money — on chance.
- Honest.
Everything the app says about itself, anywhere — the app, its page, its shop listing, the form to join — is true, and stays true. It does nothing people aren't told about, where they'll see it before they use it. The README says plainly if AI helped make the code, words, art or music.
- Open, and paid for once at most.
Everything it ships is public and open to check: code under an OSI-approved licence, and art, sound, fonts, data and any model under a recognised free-culture or open-data one. The builds are made from all of it in public, so anyone can check they match. Free, donations, or a one-off price — no subscriptions, adverts, paid add-ons, microtransactions or DLC.
- Nothing leaves the computer unless people ask.
It sends nothing anywhere until the person asks or turns it on: no telemetry, analytics, crash reports, update checks or licence checks, and no fonts or scripts quietly loaded from elsewhere. Nobody needs an account. Any server a part of it talks to keeps this whole code too, and we check what it sends and gets back.
- No AI features.
The app can be made with AI, and AI can help write its words, art or music — as long as a person reviewed them and they ship fixed, the same for everyone. What it must never do is make things for the person as they use it: nothing that writes, draws, speaks or answers on the fly, and nothing they make goes to an AI service. The one exception is reaching what's already there — reading its own text aloud, or turning speech into text.
- Nothing is counted against anyone.
Taking a break never costs anything, or lowers anything, however long it lasts. An app can show you a plain, gentle picture of your own days — even one that grows as the good days add up — so long as it only ever reflects, never randomises or rushes you, and a day off takes nothing away. The banned kind is the opposite: streaks, scores, badges or leaderboards that reward, rank or compare people, or that a break resets. A little play to help with a habit the person chose is fine if research that tested it backs it and the README names that research. Games that are plainly games can use scores and levels — but no gambling, no random prizes, nothing time-limited.
- Kindness over guilt.
It never scolds, shames or guilt-trips — including making someone feel they let a character down, or framing a choice as a loss. It only reminds people of what they themselves switched on, and never keeps on at them. Limits people set for themselves are fine, and they can always change their minds.
- Built to be put down.
Nothing is designed to keep people looking for longer than they meant to. Lists end where they said they would. The app doesn't start playing anything the person didn't choose — though opening an app whose whole point is sound or motion is itself that choice, as long as stopping it is easy. Alerts are only about things people want to know, never to pull them back in.
- When in doubt, people come first.
If a rule could be read two ways, the reading that leaves people calmer and freer to put the app down wins — not the one that keeps them in it. This never loosens another rule, and nothing here is ever a reason to harm anyone. The badge speaks for the app alone, not for anything else its makers sell. And we never use this code to attack anyone, including apps that leave.
Everything the app ships is covered, including its content and what its libraries do. Pages a browser shows are the person's own to visit — what those pages do is the web, not the app — but anything the app itself reaches for, including any server it needs to work, is fully covered.
It only ever gets stricter.
This code only ever gets stricter, and only ever in people's favour. No change may weaken the promise, weaken this lock, or let in something a rule now keeps out — a change like that simply isn't this code, and can't carry the name or the mark.
Every version is public and dated, so any attempt to loosen it is on the record. The last version that kept the promise is always the real one.
Useful apps of any kind.
The bar is that your app is at least as innocent as ours, which in practice means:
- it runs on the person's own computer;
- it works without a server at its heart;
- it ships no AI that makes things for the person while they use it;
- every word, picture and sound in it is fixed, and anyone can open and check it.
If an app can't be checked this way — it needs a live server to work, or it makes its content on the fly — we can't vouch for it, however good it is, and we'll say so plainly. That's not a judgement on the app; it's the limit of what we can honestly stand behind.
Read, then run.
Before an app joins, we read everything it ships — code and content both — and run its latest release in a sandbox with no network, watching every file it opens and everything it tries to reach. We judge it by how it would behave on any day, after any amount of use, for anyone — not just the one run we did: we move the clock and the usage forward and try it as different people. Anything it loads later, like a pack or a preset, is part of what we check. We do the same for every release after.
Anyone can tell us about a broken promise with the Report a broken promise form, about an ally's app or one of ours.
We only take on as many allies as we can really check. A badge that meant "nobody had time" would be worth less than none.
Break it once, and it's over.
A release that breaks the code ends the alliance. There's no warning period: the app comes off this page, and the page says which promise it broke, and nothing more. The decision is final.
A strike follows the maker, not just the app: a struck app can't rejoin under a new name, and joining means telling us about any app of yours that was struck before. Coming back needs a public account of what changed and a cooling-off period, decided in the open.
Build something that keeps them?
Read the code.
All nine promises, above. They're the same ones our own apps keep.
Ask in the open.
Open a Join as an ally issue: what the app is for, where its source is, and how it's made.
We check it there.
We read the source and run the latest release, and answer in the same issue, where anyone can follow along.
What allies get
- A place on this page, with a link and a line about what the app is for.
- The right to say their app keeps the Human Again code, with a link to it.
Allies
None yet. The first could be yours.