--instructions
Say exactly what you mean
The AI will happily guess a hundred details — taxes, refunds, currencies — and some guesses ship broken. Turn a vague goal into instructions exact enough that nothing important is left to chance.
// an independent standard · free for engineers
Anyone can claim it.
Prove it.
A résumé says you can. A degree says you learned. Neither shows what you can actually do in the AI era. Kaelybr measures that — on real work, against a bar the AI can't clear for you.
CLAIMPROOF
STATUS: UNVERIFIED — measured on demonstrated performance, not on paper.
# the-problem[ 01 / 04 ]
The old proof stopped working.
Juniors used to earn judgment the slow way — writing routine code until they knew how systems break. The AI writes that code now, so the apprenticeship quietly disappeared, and the credentials built on it lost their meaning. A degree proves you studied. A portfolio proves you shipped once. Neither proves you can direct a capable AI, catch it failing silently, and stop it leaking a customer's data. That's the job now. That's what we measure.
$ kaelybr --measure[ 02 / 04 ]
Three things the tool can't do for you.
--failures
Instrument the quiet failures
AI doesn't crash. It's confidently wrong, slows to a crawl under load, or quietly runs up a huge bill — with nothing flagging it. Build the instruments that catch it before a customer or an invoice does.
--verify
Make it safe on purpose
Getting a feature to work is a weekend. Making it safe is the job: it can't be tricked into leaking data, can't be manipulated, and keeps a record you can check. Designed in, never bolted on.
# zero-trust[ 03 / 04 ]
We trust no one. Not even ourselves.
trust(your_claims) → denied
Résumés prove nothing
A CV, a degree, a portfolio — none of them count here. Only what you demonstrate on real work does. That's the whole point.
trust(the_ai) → denied
The tool can't pass for you
Challenges are built so that leaning on the AI alone doesn't clear the bar. What we record is your judgment, not the machine's.
trust(our_verdict) → denied
We check ourselves too
Every result starts a clock. We re-check it against what you go on to do at work. If our bar didn't predict that, the mistake is ours to fix.
Free to take. Independent — no employer can buy a score, a pass, or a softer bar. Open — anyone can inspect exactly how we measure.
Stop claiming. Start proving.
Assessment is free, and it stays free. The people who benefit from trustworthy talent pay for it — never you.
$ kaelybr join --waitlist