@usekantei is now on @RaetoCC.
our own AI, the audit summaries and daily notes, now runs on raeto credits. 400+ models behind one API on Robinhood Chain, and Raeto picks the right one for every job. A quick question goes to a small fast model, code goes to a stronger one.
and kantei gives raeto agents the other half: which provider of a model is actually serving it and billing honestly, before they pay
https://t.co/NUN2kkb6LP
smaller fixes too: a "try again" button when a page doesn't load, cleaner link previews when you share a page, and suggestions when you mistype a model name.
kantei updates are live.
alerts now say what changed in plain words, and link to the proof.
you can search for any provider on the leaderboard.
your ai agent can ask "what changed this week?"
and the weekly report no longer counts private test runs.
https://t.co/Q7ZDlvNw32
day 21!
working on auditing and updating our product stack across the board.
sometimes it’s good to take a step back and look at where you can improve what you currently have versus pushing out new products.
a seller grading its own ai is like a restaurant writing its own reviews.
kantei doesn't sell models and doesn't take sides.
we test every seller the same way, and we show every call behind every result.
when you pay for ai, you pay by the word.
so we count the words ourselves and check them against the bill.
most sellers charge fairly. a few charge for more than they send.
you can see which is which.
https://t.co/Q7ZDlvMYdu
new ai launches every day. new models, new sellers, new ways for agents to pay.
almost nobody checks if any of it does what it says.
that's why kantei exists. we're the trust layer: we check it, we show our work, and we don't take sides.
somebody has to.
19 days in, and we're just warming up.
so we want to hear it from you: what in ai do you wish someone would check?
reply below. the best ideas get built.
when kantei flags an ai seller, the seller deserves a say.
now they get one. a seller can claim its page, see exactly why each result came out the way it did, and reply in public.
a reply never changes a result. only our checks do.
https://t.co/pMGlWoUsIn
right now, only about half the ai sellers we check pass every one of our checks.
the rest aren't all bad. some were down, some answered differently, and some are due a fresh check.
your app can't tell which is which. kantei can.
everyone in ai tells you how fast or how cheap.
kantei tells you if it's real.
we check that the ai you're paying for is the model it claims to be.
we check that your bill matches what you actually used.
we let your agent pay only when the answer is right.
we check every listed paid api, every day, without paying a cent.
we grade the code your agent pays with.
and we publish all of it, for anyone to check again.
we don't sell ai. no seller can pay for a better result.
we looked hard. nobody else does all of this.
kantei is a diamond, don't sleep on us.
your app sends its ai calls to sellers nobody checked.
now one line of code changes that. your calls go only to sellers that pass kantei's checks.
your key, your bill, and we never see your prompts.
https://t.co/BEZWZNPAEC