The Elicitation Interview

A1Step 1: How many creative endeavors are typically live for you at once? Step 2 (appears after): Is that number a deliberate choice, a happy accident, or something you're actively managing?

Number → then: choice / accident / managing / varies · optional elaboration

Establishes concurrency as a fact with the user's own valence attached — the instrument never assumes more or fewer is better. v1.3 (pilot R1): asked as two explicit steps — the pilot answered the count and the valence half silently dropped.

A2Think of the last thing you actually shipped, launched, or released. What conditions made that possible?

The launch-vector question, inverted: instead of asking why things don't ship, it extracts what shipping actually requires for this person.

A3Think of a project that ended without shipping. Looking back — right outcome, wrong outcome, or mixed?

Explicitly permits both peace and regret about unfinished work; the answer's valence is contract data either way.

A4"Progress feels most real to me when…" — rank these:

something ships · a hard problem yields · the idea expands · quality crosses my bar · I learn something new · someone reacts to what I made

The felt definition of progress; downstream alerts use the top-ranked forms when connecting observations to goals.

A5When a project sits untouched for a month, what's most often true?

interest moved elsewhere · blocked on a decision · what remains is execution-only work · waiting on something external · it's deliberately resting · I want to return but can't start · other

Distinguishes resting from stalled from abandoned — three states conventional tools collapse into "inactive."

A6Name one or two of your projects that are furthest along right now but remain unshipped. For each: what's its current status? Are you actively working on it? If not — what's true about why not?

Grounds the launch frontier in named, real projects instead of abstractions. The "why not" arrives in the user's own words before any pattern analysis exists to bias it — and gives future drift alerts concrete, user-nominated targets ("you said Keep Sparrow was one decision away from shippable") rather than generic launch nudges.

A7Now the other end: name a project that's least far along but that you haven't abandoned. What's going on with it — what keeps it alive, and what is it waiting for?

The mirror of A6. Activity data can't distinguish an incubating seed from a dead project — only the user's framing can. The answer defines what "alive" means in this portfolio and tells synthesis how to treat embryonic work: as seeds, as a queue, or as drift — per the user's own model, not an assumption.

B1Recall a recent week when you felt unstoppable. What was the mix of things you were doing?

Maps the dopamine profile positively — what fuel looks like, not what its absence looks like.

B2What kinds of tasks drain you even when they're easy?

Separates difficulty from drain; execution-drain and meeting-drain and polish-drain are different registry entries.

B3Think of the last two times momentum died on something. What killed it, each time?

Two episodes forces pattern over anecdote.

B4What has reliably re-ignited a stalled project for you?

Gives alerts a constructive move: instead of reporting stall, offer the user's own known re-ignition paths.

B5Deadlines and external pressure — what do they actually do to your output?

sharpen it · paralyze it · depends who set them · depends what's at stake · mostly irrelevant · elaboration

Determines whether advice may ever propose deadlines as a tool — a known style divider.

C1Distribute 100 points across what advice should protect for you:

revenue & launching · momentum & enjoyment · the quality bar · learning & growth · reputation & visibility · time & life balance

The dual mandate emerges from the weights rather than being asserted; a user who puts 70 on revenue gets a different contract, legitimately.

C2Name one trade you'd happily make, and one you'd never make. ("I'd trade ___ for ___." / "I'd never trade ___ for ___.")

Hard constraints in the user's own words — the sentences most likely to survive projection compression intact.

C3In your own words: what would make the next 90 days a win?

The near-term anchor alerts connect observations to.

D1Recall critical feedback — from a person or an AI — that landed well. What made it land?

D2Recall accurate feedback that made you disengage. Was it the delivery — or the content itself?

Explicitly allows "the content" as an answer; the contract must record where candor itself has limits for this user, if anywhere.

D3When an AI notices an uncomfortable pattern in your work, what do you want it to do?

say it directly · say it connected to my stated goals · show me the data, let me conclude · hold it unless I ask · depends — elaborate

The core candor dial, asked plainly.

D4Are there areas where you want more bluntness than most people would tolerate? Areas where you want extra care?

D5Is there anything you've stopped telling AI tools — or started using separate or logged-out sessions for — because of how they responded?

The disengagement question. Whatever surfaces here is the highest-priority registry entry, because it's already cost the relationship something.

E1Launch risk:

E2Quality bar:

E3New idea mid-stream:

E4Consistency:

E5Are there specific words, phrases, or framings — in written advice from anyone, human or AI — that make you tune out or push back?

Direct registry entries, verbatim. v1.3 (pilot R4): "stop listening" elicited literal audio triggers; reworded to written advice.

F1Recall a risk that felt energizing, and one that felt threatening. What was different between them?

The named divider between the two of you — asked as episode contrast, not self-classification.

F2Public commitment — telling people what you're building, announcing dates: motivating fuel, suffocating pressure, or situational?

fuel · pressure · situational — elaborate

F3When money is on the line, does your creative process change? How?

F4Think of a project whose main purpose is revenue. Do your normal rules change for it — enjoyment of the process, personal use, taste, what "done" looks like? (nothing changes — same rules everywhere · some rules relax · most rules change — revenue is the mission) Then: what, if anything, stays non-negotiable even there?

v1.4, from the pilot's F3 follow-up: some builders run distinct project modes with different rule sets (personal build vs. revenue mission — "the revenue IS the mission"), while others hold one rule set everywhere. If modes exist, the contract's objective function becomes mode-conditional and advice must first ask which mode a project is in. The non-negotiables question maps the floor that survives every mode.

G1What would make you trust an AI advisor more over time? What would erode that trust?

G2If an AI must choose between telling you something true and protecting your momentum this week, what's the standing instruction?

always the truth — I'll manage my reaction · the truth, timed well · protect momentum now, raise it later · depends — elaborate

The contract's tiebreaker clause, set by the user rather than assumed.

G3Is there anything about how you work that every AI keeps getting wrong?

The open catch-all. This question, asked cold, is the one most likely to surface each user's version of the concurrent-projects insight unprompted.