yudonKnow
How it works 한국어

Most AI answers your questions. This one asks you.

The senior leaves. The judgment stays.

What retires with an expert is not documents — it is the eye that reads a situation. yudonKnow hands them instruments to dig it out themselves, and leaves an alter the juniors keep working with.

① Task map → ② incident harvest → ③ timeline → ④ deep probes → ⑤ judgment cards → ⑥ member checking — the knowledge-engineer's procedure, run by the agent, voice-first.

🧑‍🏭 Enter as the senior

The AI interviews you — 3 minutes leaves a draft, 7 leaves a reviewed judgment card.

① The AI interviews you — you just talk (voice works) ② You review and approve each judgment card — and the rules that may run ③ What stays: your alter for the juniors, a memoir, and a usage statement that pays you back

💬 Enter as the junior

Ask — the alters answer only from their seniors' cards.

① Ask in your own words — or walk the structured triage ② Every answer cites the senior's card; open it and check ③ Report "helped / didn't hold" after you try it — that earns the ✔ and pays the senior

What you leave here is yours.

· Private — only you · Sealed — opens on a date you pick · Targeted — one named successor · Export everything, any time · If your alter talks nonsense, you switch it off · War stories are never linked to HR records · Every verified use is counted — a usage statement you can bill with

Demo scope: card content honors your visibility settings; profile name and usage counts are visible in this demo. Production scopes those behind corporate SSO.

🎫 Judges use the same two doors — no sign-up, no password. Quick: Ask Dale's alter as a junior · Watch Dale's expert page · expert in one click

Demo identity — no authentication, by design: a lightweight switcher lets judges test the senior and the junior roles immediately, on synthetic data. In production, owner access, visibility, and verified usage are bound to corporate SSO. This demo focuses on the knowledge-extraction and reuse engine.

The alters left behind

Don't know whom to ask? Start with the question

Dale's alter

wastewater treatment · EN
4 judgment cards · ✔ 1
retires in 29 days

Vet Front Desk's alter

Veterinary front desk triage · EN
3 judgment cards

김덕규's alter

협력사 품질 대응 · KO
4 judgment cards · ✔ 1
retires in 82 days

+ Leave mine

Answer as I ask — 3 minutes leaves a draft, 7 a reviewed card. It is yours — you set who may see it, and you can switch it off.

Who leaves, and what goes empty — succession risk board →

How it works — one turn of the wheel

1 · The AI interviews you

Not a chatbot waiting for input — it runs the knowledge engineer's procedure: task map, incident digging, deepening, your own review. You just answer.

2 · Answers become judgment cards

Situation · cues · judgment · action · exceptions · failure. You approve every card and set who may see it.

3 · The alter answers juniors

Only from your cards, citing them paragraph by paragraph — a mechanical check discards any uncited answer. No cards, no answer: the question queues back to you instead.

4 · The field verifies

Juniors report "it helped" or "it didn't hold" — the only source of the ✔ badge. No views, no likes.

5 · A memoir, not a database

The same digging typesets a career memoir — drafts woven by AI, on the record only after the author approves.

6 · Use returns to the retiree

Every citation and field report lands on a usage statement — the verifiable basis for compensating knowledge after retirement.

The rule underneath: judgment is never delegated to the LLM. Gap decisions, citation checks, memoir honesty — all deterministic code, all auditable. Generation is convenience; the cards are the truth.

LLM: gemini-3.5-flash (vertex/global) · Vertex AI