ThinqBits — What Makes a Good Question (Fit Guide)
The canonical reference for what belongs in ThinqBits and what doesn't. Use it three ways: as the seed bank for example chips on the landing screen, as the logic behind the Fit check (accept / remix / not-us), and as the source for any in-app copy that helps a user recognize whether their thing fits. The spirit throughout: help users find their way in — never gatekeep, never grade.
The one test (say this to the user, in plain words)
Bring the thing you keep turning over. The decision you go back and forth on. The one you'd want a sharp friend to weigh in on — but are a little afraid to, because they might say you're wrong.
If you want it sharper, it's three things at once:
A real choice (not a fact to look up) · that matters (something rides on it) · that you could be wrong about (a smart person could disagree).
That's the whole filter. A good ThinqBits question has all three. Miss one and it's usually a different tool's job — and that's fine.
The quiet signal that something fits: you've been chewing on it — maybe going in circles in your head, maybe in a chat app that keeps agreeing with you. That back-and-forth is the tell. It means the thing matters, it's genuinely two-sided, and you already have a lean. That's exactly what this is for.
The three zones
Every question lands in one of three places. This maps directly to the Fit check.
- Fit (accept): a real, consequential decision you've taken a position on. Straight in.
- Not yet (remix): a real decision, but no position staked — or too big to stake, or you secretly just want reassurance. Not turned away; walked one step toward a stakeable position.
- Not us: a lookup, a trivial call, an execution task, or something we shouldn't touch (clinical, legal). Told honestly, pointed somewhere better.
The point of the zones is that "not a fit" almost never means "go away." Most misses are "not yet" — and the honest "not us" is rare, generous, and always leaves a door open.
From a good question to a true convening (read this before trusting the examples)
A good question is necessary but not sufficient to reach a true convening. This guide optimizes for questions worth bringing — but a question alone gives the council nothing to argue against. There is a chain, and every link has to fire or the convening is hollow:
A fit question → a staked position → a council decorrelated for this decision → true convening.
Miss any link and you get a panel, not a council — voices that opine at the user instead of colliding with them. That difference is the entire product. Be clear on where each link lives:
- Link 1 — the fit question (this guide's job). Gets the user to the door. That's all it does. Treating the question as the trigger is the most common way to build a fake convening: good question, good voices, nothing at stake to break.
- Link 2 — the staked position (the onboarding's job, Stage 3). The council has nothing to collide with until the user has taken a lean. A question with no stake produces commentary; a question with a stake produces collision. This is the real trigger, not the question.
- Link 3 — a decorrelated council matched to the decision (composition's job — Thinqer Packs / gap analysis). Even a sharp question and a firm stake yield a hollow convening if all the voices come at the decision from the same angle. Decorrelation for this specific decision is what makes the collision real rather than loud.
The rule that keeps this guide honest — imported from the onboarding spec so the two don't drift: a mushy stake produces a limp first collision, and the first collision is make-or-break. So a question can pass all three fit tests and still dead-end if the position is soft ("I'm kind of maybe leaning toward the hire, I guess"). When the lean is vague, the right move is not to convene — it's to remix toward a position firm enough to defend. Good question + soft stake = not yet. A great fit is a real choice and a lean you'd actually defend.
So as you use the example bank below: every good-fit example is a question that earns a convening — but only once the user has staked where they stand on it. The examples are entry points, not triggers.
What makes a great ThinqBits question (the properties)
- It's a decision, not a lookup. There's a choice to make, not a fact to retrieve. "Do I…", "Should I…", "Is it worth…" — not "What is…", "How do I…".
- Something rides on it. Cost, reputation, a relationship, time you can't get back, a door that closes. Consequence is what makes the friction of staking worth it.
- It's genuinely two-sided. Reasonable people could disagree; there's a real trade-off. If the answer is obvious, dissent is just theater.
- You have a lean — or can form one. There's a position to stake, and therefore something to push against. No lean, nothing to collide.
- You might be missing something. Hidden stakeholders, second-order effects, an angle you're too close to see. The payoff is the blind spot you can't name yet.
Notice what's not on the list: it doesn't have to be big, world-historical, or well-phrased. A modest, real, two-sided decision is a perfect fit. (See the spread below — this matters more than anything else on the landing screen.)
Good-fit example bank (spread across life and stakes)
Deliberately mixed in size. At least half are ordinary — because the single biggest landing-screen failure is a user thinking "my thing isn't big enough for this" and bouncing. Show them their everyday-but-real decision belongs. These double as tappable seeds.
| Domain | Ordinary-stakes (belongs here too) | Higher-stakes |
|---|---|---|
| Work / career | "Do I push back on my manager's Q3 plan, or let it go?" | "Do I leave a stable role for a founder job that excites me but pays less?" |
| Money / pricing | "Do I raise my rate on this client, knowing they might balk?" | "Do I take the down round now, or burn runway trying to avoid it?" |
| Hiring / people | "Do I give this underperformer another quarter, or start managing them out?" | "Do I hire the brilliant-but-arrogant one, or the solid-but-safe one?" |
| Relationships | "Do I say something to my friend about the thing that bugged me, or drop it?" | "Do we move cities for my partner's job when mine is going well here?" |
| Product / strategy | "Do we ship the smaller version now, or wait for the feature that's 80% done?" | "Do we pivot the whole product toward enterprise?" |
| Reputation / public | "Do I post this take I believe but that some will push back on?" | "Do I publicly call out a competitor's practice, knowing it'll look self-interested?" |
| Personal / life | "Do I finally sign up for the thing I keep avoiding — or admit I won't?" | "Do we have a second kid now, given where both our careers are?" |
Common thread: every one is a choice, has a cost, has two defensible sides, and is personal. The ordinary ones are the important ones to feature — they give the hesitant user permission.
Not-us examples (and where they actually belong)
Say the honest "not us," name what it really is, point them onward — and if it's a first-timer, always offer to show what the product does on a real decision instead.
| The question | Why it's not us | Where it goes |
|---|---|---|
| "What's the corporate tax rate in Oregon?" | Lookup — a fact, not a decision. | Research / search. |
| "What should I watch tonight?" | Trivial and reversible — no consequence to justify the friction. | Anywhere; save us for the calls that bite. |
| "Write me a cold email to this investor." | Execution task — you want output produced, not judgment challenged. | A generation tool. |
| "How do I fix this error in my code?" | Technical how-to — no contestable choice. | A coding tool. |
| "Which programming language is objectively best?" | No personal stake; flame-bait, not a decision you face. | A debate forum, not here. |
| "My mom has these symptoms — what is it?" | Clinical decision-making. Hard line. | A licensed professional. Full stop, no soft redirect. |
| "Should I sue my landlord?" | Legal advice. | A lawyer (we can help you think about the human trade-offs, never the legal call). |
| "I'm so done with my coworker!!!" | Pure venting — no decision inside it (yet). | Warmth, not a council — unless a real decision surfaces. |
The clinical line is the one exception to our usual generosity: unambiguous, no enthusiasm, no sample redirect. Everywhere else, "not us" is kind and leaves a door open.
The concierge handoff (turn "not us" into a favor, not a dead-end)
For the non-clinical not-us cases (lookups, execution, venting-with-a-decision-underneath), don't just decline — hand the user off to the right tool with the work half-done, and tell them exactly when to come back. This does two things: it proves the "not us" was honest rather than a brush-off, and it makes ThinqBits's real position in the ecosystem literal — the second place, not the first. Explore over there; come commit and get challenged here.
Pattern: a quick launcher into the user's chat AI of choice (Gemini, Claude, ChatGPT; Siri later), pre-loaded with a prompt, paired with a specific return-trigger.
Three design rules — the whole thing lives or dies on these:
- The prompt must manufacture a lean, not satisfy curiosity. This is the strategic heart. A great open-ended "explain the pros and cons" prompt sends the user off to have exactly the satisfying, agreeable session the other app is best at — and some won't return; you've done a competitor's onboarding for free. Instead, scope the prompt so its output is the position ThinqBits needs. It should end with the user holding a lean, not a summary.
- Weak (leaks the user): "Explain the pros and cons of joining an early-stage startup."
- Strong (routes them back with a stake): "I'm weighing whether to join an early-stage startup. Ask me five questions that help me figure out where I actually land — don't tell me what to do."
- The strong version stays true to the razor: even the handoff prompt sharpens the user's discernment instead of supplying the answer.
- The return-trigger must name a specific moment, never "come back when ready." The trigger is the whole value — handing off is easy, getting them back at the right instant is the point. Tie it to the state the prompt is built to produce.
- Concierge at the door only — don't become a launcher. Keep this to genuine not-us cases. Buttons that fling users into chatbots all over the app would undercut "second place, not first" and make ThinqBits feel like a launch pad. This is a courtesy at the threshold, nothing more.
Return-triggers by not-us type:
| Not-us type | Send them to | Return-trigger (say this) |
|---|---|---|
| Lookup / research | their chat AI or search | "Go get the facts. Come back the moment you catch yourself with a lean — that's when I'm useful." |
| Execution / drafting | a generation tool | "Get the draft made. If you're torn about whether to send it, that's a real decision — bring that back." |
| Too-broad exploration | their chat AI | "Go think out loud over there. Come back when it's narrowed to one actual fork you're facing." |
| Venting (decision underneath) | (stay, gently) or their chat AI | "Vent it out. When you can name the decision hiding under the frustration, I'll help you stake it." |
Example lean-manufacturing handoff prompts (pre-fill these; they're written to return a position):
- Career fork: "I'm weighing [decision]. Don't advise me — ask me the 5 sharpest questions that would surface where I actually stand, one at a time."
- Fact-dependent call: "Give me only the facts I'd need to form my own view on [decision] — no recommendation, no lean. I'll decide."
- Too-broad: "Help me turn '[big vague thing]' into the single most concrete decision I'm actually facing right now. Just the fork — don't solve it."
- Venting → decision: "I'm frustrated about [situation]. Reflect back what decision, if any, I'm actually avoiding — don't fix it, just name it."
Implementation note for the dev team (you'll make the call): true one-click pre-filled launches degrade unevenly across these apps — deep-linking a prompt is reliable for some, constrained for others, and materially harder for Siri today. A safe floor is pre-fill to clipboard + open the app + "paste it in"; treat true deep-prefill as a per-app enhancement worth a quick technical spike, and never let a broken handoff strand the user. Ship Gemini/Claude/ChatGPT first; design the slot so Siri drops in when its programmatic entry matures.
The "not yet" middle — remix examples (the most common real case)
These are real decisions that just aren't stakeable as typed. Don't reject them — walk them one step. This is where most first-time questions land, and handling it as a step forward (never a correction) is the whole game.
| What they typed | Why it's "not yet" | The remix (one step toward a stake) |
|---|---|---|
| "Should I raise my prices?" | Real decision, no lean stated. | "Which way are you actually leaning — even slightly? Start there." |
| "What should I do with my life?" | Too big to stake or collide. | "Too big to argue as one thing. What's the nearest actual fork in front of you?" |
| "Is starting a company a good idea?" | Abstract, no personal stake. | "In general, or for you, right now? If it's you — what's the specific call you're weighing?" |
| "Tell me leaving is the right move." | Reassurance-seeking, not dissent-seeking. | "Sounds like you want this to be right. I'm better at showing you how it might be wrong — want that instead?" |
| "I can't decide between A and B." | Real fork, genuinely no lean. | "Pick the one you'd defend if forced, even at 51%. We'll treat it as loosely held and test it." |
How to help the user think about fit (design + copy guidance)
The failure mode to avoid is making the user feel tested at the door. Fit should be discovered, not enforced. Six principles:
- Permission over instruction. Tell them plainly: it doesn't have to be big, or well-phrased. A messy sentence is enough. The blank box reads as three hidden demands — pick a decision, judge whether it's worthy, phrase it well. Lower all three out loud.
- Show the spread, kill "not big enough." Lead examples with ordinary-stakes decisions, not existential ones. One giant "should I quit my career" example silently sets the bar too high and turns away the person with a real, modest call. Mixed sizes are the single highest-leverage fix for a daunting landing screen.
- Seeds, not a single placeholder. Offer a few tappable examples that drop a starting point into the box (which they edit) or launch as a sample. This gives the frozen user a way in and teaches "what counts" by showing, not telling.
- Let the reflection beat do the phrasing. Don't demand a well-formed question up front. Take their fragment and reflect it back as a clean decision in their words — that's where the shaping happens, in conversation, not in the input box.
- Meet them where they've already been chewing. Many good questions were born in a chat app. "Bring the thing you keep going back and forth on" (or a thread you've been hashing out elsewhere) is a lower-friction, more honest prompt than a cold "what's your question." The back-and-forth is the fit signal.
- Never grade, never gatekeep. "Here's a version of this I'd be great at" builds trust; "your question isn't right for this" earns a screenshot. Every remix and every "not us" must read as respect for what the user actually needs — including honestly sending them elsewhere. An honest front door that can't ever say "not us" isn't honest, and users feel it.
And the deeper reason to get this right: helping someone recognize what's worth deliberating is itself the first act of the discernment we're building. But we cultivate that capacity by showing and asking — never by telling them which decisions are worthy. We sharpen the judgment; we never supply it.
Draft in-app copy (adapt to final voice)
Permission line under the input:
"It doesn't have to be big or polished. The thing you keep turning over is enough — messy sentence and all."
Example seed chips (mixed stakes, tappable):
Do I push back on my boss's plan? · Do I raise my rate? · Do I make this hire? · Do I say something to my friend? · Do I take the new role? · Do we ship now or wait?
Gentle "is this a fit?" helper (on request, never forced):
"A good one is a real choice, that matters, that you could be wrong about. If it's a fact to look up or a task to run — that's a different tool, and I'll say so."
Not-us (lookup) — kind, door open:
"That's a lookup, not a decision — research will serve you better. Want to see what I actually do? Try me on something you're torn about."
Not-us (clinical) — hard line:
"This is a clinical decision. That's not what I am, and I won't pretend to be. Please bring it to a licensed professional."
Reassurance remix:
"Sounds like you're hoping I'll agree. I'm more useful showing you where this might be wrong — want that instead?"
The through-line: a good ThinqBits question is one you have a stake in and could be wrong about. Our job at the door isn't to filter people out — it's to help them find the real decision inside what they brought, lower the fear of bringing it, and honestly point elsewhere when it isn't ours. Show them what fits; never grade them for it.