Runic
QUESTIONS

After the draft, the agent asks what it left open

A queue of questions, each bound to one document, asked one at a time; every answer comes back as a diff on the document the question belongs to, and the next question waits until you decide.

Try it free →Read the docs

Where the questions come from

Two sources feed the same queue. After a Whole system run the plan already carries 5 to 12 open questions bound to the documents they would change — the founder-blocking ones first, about the main flow — and Runic turns up to 20 of them into the queue with no AI call, keeping only questions whose document was actually created or linked and remembering which kind the plan meant, so a process can never shadow a sequence diagram of the same name. The queue starts once the main flow has finished typing itself in.

After a Process flow brief that came back as a linked set of two or more processes, Runic sends the set to a reviewer prompt with one job: find what the draft does not answer yet, and ask only questions whose answer would change the diagram — a new .if, a validation step, a .fail with recovery, a .note threshold, a .back retry, a permission gate — never something the documents already answer, in the documents’ language. It runs on Claude Sonnet 5 with thinking on, 20 runs an hour, reading up to 40 documents of 20,000 characters each and 120,000 in total (a cut is announced in the text), and answers in a strict line format — document name, topic, question — which Runic binds to a real document (exact name, then case-insensitive, then similarity), de-duplicates and caps at ten. The chat then opens with “The draft covers the happy path. I have N questions that would change the diagram — validations, thresholds, what happens when things fail. Answer in your own words; skip any that don’t apply.”

Asked one at a time, answered into the right document

An “Open questions” panel sits above the message box with a progress count (“2 / 7”), a Skip button and a ✕ to close the queue; each row shows its document and topic with ○ pending, ▸ current, ✓ answered or – skipped. Asking a question switches the editor to the document it is bound to and posts “<document> · <topic>” followed by the question. Topics are validation, decision, failure, timing, permission and data.

Answer in your own words. Runic wraps your reply as an open question plus your answer and asks the agent to update that document so the answer is reflected where it belongs — a validation step, an .if decision, a .fail with recovery, a .note threshold or a .back retry — keeping everything else intact, in the document’s language. The answer always goes to the document the question is bound to, even if you clicked elsewhere to look something up; the editor jumps back to it and your bubble shows “[answer → ‘<document>’]”. If that document was renamed or deleted in the meantime: “I can’t find ‘<document>’ anymore — skipping this question.”

The queue waits for your decision

When the answer comes back as a proposal, the queue parks on it: the next question is asked only after you Apply or Discard (or Skip), so the next answer is computed against the document as it now stands rather than a base that is about to change. A prose reply advances the queue at once. A failed request leaves the question open for another try. Skip marks the question skipped and lets go of any proposal the queue was waiting on.

When the last question is answered or skipped: “That’s the whole queue — 5 of 7 answered, 4 changes applied to the documents. Ask me anytime to go deeper on one.” The count is of changes you applied, not of replies you got. ✕ ends it early with “Interview closed — the remaining questions are dropped.”

Limits

The interview runs in signed-in sessions after Create with AI: a whole system with at least one written document and plan questions, or a Process flow linked set of two or more processes. A single generated document does not trigger it, and there is no button to start one by hand. The queue lives in the current session: questions already posted stay in the chat history, but the panel does not survive a reload. Each answer is one chat message against the 60-an-hour limit and one proposal — nothing changes until you Apply it.

How to use it

  1. Generate a system or a linked set
    Use ✨ Create with AI. After a whole system the questions come from the plan; after a Process flow linked set the reviewer asks them. The chat opens on its own.
  2. Read the first question
    The editor jumps to the document the question is about. The “Open questions” panel shows where you are in the queue.
  3. Answer or skip
    Type an answer and press Enter, or click Skip. Click ✕ on the panel to end the interview.
  4. Apply or discard the diff
    The answer comes back as a proposal on that document. Review it, Apply or Discard — the next question is asked only after you decide.
  5. Go deeper later
    After the closing message, keep asking in the chat — “add a retry to the credit check” works the same way.

Commands & shortcuts

Open questions · 2 / 7the queue panel with answered count
Skipmark the current question skipped, release any awaited proposal, move on
✕ (queue)close the queue and drop remaining questions
Entersend your answer; it becomes a proposal on the bound document
[answer → “doc”]the note on your bubble naming the document the answer was edited into
validation · decision · failure · timing · permission · datathe six question topics

Tips

  • Short, concrete answers work best: “above 5,000 the CFO approves; otherwise the manager” becomes an .if with two branches.
  • Skip questions that do not apply rather than answering “n/a” — a skipped question changes nothing.
  • You can look at another document while a question is open; the answer still goes to the document the question is bound to.
  • Apply the proposal before expecting the next question — the queue waits on it by design.
  • The topic label tells you what kind of edit to expect: failure adds a .fail with recovery, timing adds a wait or a threshold note.

Questions

Can I start an interview on an existing process?

Not from a button. The interview runs automatically after Create with AI. For an existing document, ask the chat directly: “what does this process leave open?”

Does my answer change the document immediately?

No. It comes back as a diff on the document the question is bound to; you Apply or Discard it like any other proposal, and the queue waits for that decision.

How many questions are there?

From a plan, 5 to 12 are requested and up to 20 kept, bound to processes or sequence diagrams. From the reviewer after a linked set, 5 to 10 are requested and at most 10 kept after parsing and de-duplication.

What if I reload mid-interview?

The questions already posted remain in the saved conversation, but the queue panel is gone. Answer them as ordinary chat messages instead.