Build › Plan
The foundations a senior engineer would insist on, planned from your answers
On this page
Build › Plan lists the foundations your product needs before launch, the ones that are expensive to add later: sign-in that can be recovered safely, backups you have restored once, every request checking who is asking, a privacy policy, and so on. It asks a few plain questions about your product, then shows what to do in v1, what can wait and what you don't need at your size. Every item says why it is there and where that came from.
Everyone on the product reads the plan. Editors answer the questions, change items and accept the plan into the Map.
1. Answer the questions (free)
On Build › Plan, Start the questions (about ten, one at a time):
- What kind of product is it? Mind reads your thesis, your main segment and your description and ticks what it thinks fits, each with the words it came from. Untick what is wrong and tick what is missing: who buys and uses it (a business, consumers, developers, a marketplace, one event at a time, your own team), how it is delivered (web, mobile, extension, desktop, hardware) and any high-stakes area (elections, health, finance, children, education, hiring, legal work). With two kinds ticked, say which comes first.
- Up to nine more, picked from what you ticked and from your own words. A question that cannot change the plan is skipped. Questions marked One-way: expensive to change later deserve a moment.
- Suggested answers. Some arrive ticked because your workspace already implies them, and say where from. Not sure: recommend one picks a sensible answer and gives the reason. Skip lets Mind assume the safe default, and the plan says it assumed. Back keeps your answers.
- Your numbers (optional): two to four numbers about your first year, such as your biggest event and when people act. They size the plan, including what you do not need to build yet.
Save my answers shows them as a list. Untick any you do not want, then apply: each answer becomes a decision in Learn › Decisions, and your product's shape is saved. Undo lasts 24 hours. The interview resumes where you left it; Start over begins again.
2. See the plan (free)
See the plan (free) drafts the plan at once, by rule, with no AI call. The same answers always give the same plan.
- The first line says what kind of product it is ("Event-based · B2B · Elections · Web · Idea stage", with Change) and what the plan read: your thesis and positioning, how many pages of your tracked competitors' own docs and how recent, the playbook for your kind of product, and your answers.
- What to do this week: at most five lines, in order. The one-way questions you were not asked, the first things only you can write or decide (a privacy policy), Send these N questions to your developer (with Copy all) and the one fix your users' complaints point at.
- v1, Later and Not now, each in groups ("Sign-in and accounts", "Election integrity"). A group shows how many items, the largest size, a flag when something in it is expensive to add later, and what it needs first. Open an item to see where it came from, why it is in this lane, what it needs and what needs it, who decides, its size and approach, and Done when.
- Sized to your numbers. The plan works out your busiest minute, puts a load test and a queue in v1 when you need them, and lists what you do not need at this size (autoscaling, several regions, a separate cache).
- What you already have is marked Already on your map and is not proposed again. Your users' complaints raise an item's priority and appear as counts and keys, never a quote or a name.
3. Tailor and check (12 credits)
Tailor and check reads your own features: what each needs to work safely and at scale, how someone could abuse your flows, and a senior engineer's check. It costs 12 AI credits (10 to tailor, 2 for the critic; on Growth and Scale it shows Included). The jobs tray shows "Mind is tailoring your plan…", then "The critic is reading your plan…". Afterwards:
- Mind's reasoning sits in the groups, labelled, with what it serves and its confidence: high (both of Mind's passes found it), medium (one pass) or low (only under Worth asking your developer, never ticked).
- What users do: up to 12 rows (must-have, differentiator, later), as Draft the map writes them.
- How someone could abuse it: each case with who, the impact and what covers it, or the question to ask your developer.
- The critic's notes: each from a rule or from Mind, with Move to v1, Keep it in Later, Answer it, Add the fix story or Dismiss. Applying a note is a decision. Dropped lists what the checks or the critic removed, each with Restore.
- Mind's questions: up to three things the questions could not know. Your answer is a decision, and the next tailoring reads it.
Credits come back if tailoring fails, times out or keeps nothing (10), or if the critic fails or finds nothing that holds up (2). An idea Mind will not plan, or one too vague to plan, is never charged. Mind screens every idea first: an idea whose main purpose is illegal or harms people gets no plan. When your workspace has used its month's AI allowance the button says so and nothing is charged.
4. Change an item
Open Change or correct… on an item to move it to v1, Later or Not now (moving a high-stakes area's v1 item out needs a reason), mark We have this (already built) or Dismiss (Mind's items and rows only). Helpful and Not right rate an item; Not right asks what is wrong and keeps it under Where the plan was wrong. Every move is a decision in Learn › Decisions, and it carries over when the plan is drafted again.
5. Accept into the Map
Accept into the Map shows what would be created before anything is. Each group becomes an Idea feature in Build › Features, in its lane ("Election integrity", or "Election integrity: later"), with a draft story per item and its Done when as acceptance checks. v1 groups and Mind's high-confidence v1 items are ticked; medium-confidence items, Later groups and the rows of what users do are unticked. Undo lasts 24 hours. Accepting a later plan adds the new items to the same feature, never a second one.
Once a plan is accepted, every feature's Build Pack carries a FOUNDATIONS section for your coding agent: the plan's decisions (one-way calls first), the rules every feature follows, the size to build for, what to defend against and what the feature needs from other features. It uses the playbook's own words and your decisions, never Mind's free text, and fits in about two kilobytes.
When the plan is out of date
A banner says why (your shape changed, you answered a question, the playbook changed, new competitor docs match, your code map was synced, a feature was added or removed) with Redraft (free).
Learning from what went wrong
On a bug's page, editors can say which foundation was missing (Caused by a missing foundation; Unlink takes it back). It is kept as a lesson from your own product, and the next plan raises it. A foundation you added after launch, which the plan had in Later, Not now or dismissed, is kept the same way. Nothing is learned across workspaces.
What it will not do
- It is a framework, not legal advice. Risk lines end "Check with counsel where you operate", and the plan never says a product is compliant, legal or certified.
- Competitors appear only through their stored, dated docs. Nothing is searched for the plan, and it never says what a competitor lacks.
- Mind's reasoning is labelled as Mind's. The base of the plan is MindBacklog's own playbook.
- Viewers read everything but see "a part of your code map" instead of a component's name, and no buttons.
- A demo session cannot open the plan.
FAQ
Why can't I press Tailor and check? Draft the free plan first. If your workspace has used its month's AI allowance, the button says so and it resets next month. "Tailoring waits: see the note above" means the idea was refused or is too vague to plan.
"Mind is already tailoring this plan." One tailoring runs at a time. Wait for the jobs tray to finish.
"Answer the plan questions first: what kind of product it is decides what it needs." Start the questions before you draft.
"Insufficient AI credits. You need 12 credits but have N remaining." Tailoring needs 12 credits. The free plan needs none.
See also Build › Map, The Build Pack and Decisions.