How it works
Your judgment, written down. Then applied, every time.
Twynit starts by capturing something no system has ever held: what a good candidate looks like, to you, for this search. Everything else — the reading, the ranking, the receipts, the compounding — follows from that one act.
01 — The standard
Describe the hire in plain English.
You write it the way you'd brief a colleague. Your Twin turns it into a scorecard — weighted priorities, must-haves, and the screening questions that will establish what a résumé can't. Every weight is visible and yours to change.
“Staff backend engineer. Strong Go and Postgres in production — payments experience is a real plus. I care more about ownership than years.”
+ 4 screening questions drafted — edit anything before it runs
02 — The approval
Nothing runs until you've approved the judgment.
A search is never live until you've signed off on how it judges — whether it was created by hand, imported from your ATS, or parsed from a job description. No candidate is ever evaluated by a standard you haven't seen.
03 — The reading
Every candidate, read the same careful way.
The 400th application gets the same read as the first. Ranking and filtering are deterministic — rules and math against your scorecard, not vibes — and every verdict carries its evidence, cited to the line it came from.
When something can't be established, your Twin says so instead of guessing — and asks the candidate directly through screening.
notice period: not determinable → added to screening
04 — The decision
You decide. It remembers why.
Every call you make lands in an immutable ledger — the decision, the evidence, the why. When you overrule your Twin, it asks what it missed, and your answer changes how it reads the next hundred candidates.
05 — The trust
Autonomy is earned, one action at a time.
Everything starts in Ask mode: your Twin previews the action and waits for your call. Approve the same action enough times, unchanged, and it offers to just do it — announcing each act with a receipt, reversible for 72 hours.
Undo once, and that action goes back to asking. Trust is granted by you, never assumed.
06 — The compounding
Outcomes close the loop.
Hires get outcome check-ins — two clicks at 30, 90, and 365 days. Outcomes become knowledge facts with evidence and sample sizes; knowledge proposes scorecard updates; you approve them, and the pipeline is re-read under the sharper standard.
That's the whole loop: every session creates progress today and makes the company smarter tomorrow.
Production incident ownership predicts success in your backend hires better than title seniority.
→ proposed scorecard update · awaiting your approval · v3 → v4
Write the standard once. Watch it get sharper for years.