Four finished maps to start from

Each one opens as a new map, so the map you were working on stays as it was. They are the maps the app ships: real titles, connections that mean something, and a situation already set, so the brief reads like a brief from the first copy.

Checkout 500s: investigation

The walkthrough's case: checkout fails for some people, nobody knows why, and you want a coding assistant to read the repository instead of inventing a plausible theory. Twelve blocks that separate the symptom from the beliefs around it, with the reproduction as the first boundary. The walkthrough builds it from nothing.

The checkout investigation map: the report leads to the problem, then to reproducing it, two open questions and two assumptions, and on to the root cause, a failing test and Fixed and proven. Opens the picture at full size. The checkout investigation map: the report leads to the problem, then to reproducing it, two open questions and two assumptions, and on to the root cause, a failing test and Fixed and proven. Opens the picture at full size.

The whole map (it opens at full size), then where it starts, at reading size.

From the brief it produces
### Constraints and boundaries
- Reproduce before theorising
## Open questions (review before assuming)
• Does it happen on staging?
12 blocks, Investigate mode

Your own bug, not this one? : the same shape without the checkout story, 13 blocks, framed for a repository you have open and told to reproduce before theorising.

Inherit a codebase

Somebody left, the repository is yours now, and the first wrong move is trusting what the README says. This map works from an unknown repository to a first safe change: what is believed, what is actually verified, and the questions whose guessed answers cost the most.

The Inherit a codebase map: from being handed the repository to a first change shipped, through running it, its tests and one traced request, two assumptions about the README and the tests, the risk of breaking something invisible, and a rollback you have tried. Opens the picture at full size. The Inherit a codebase map: from being handed the repository to a first change shipped, through running it, its tests and one traced request, two assumptions about the README and the tests, the risk of breaking something invisible, and a rollback you have tried. Opens the picture at full size.

The whole map (it opens at full size), then where it starts, at reading size.

From the brief it produces
### Constraints and boundaries
- Read before writing
## Planning gaps detected
• Goal "Be able to make a safe change" has nothing under it: no requirement, metric or implementation.
13 blocks, Investigate mode

Migrate a system

Old to new with a way back at every step. The part most migrations skip is the reconciliation loop: old against new, per entity, every day, with a threshold that blocks the next phase. This map carries it, plus the risks that live at the cutover and an end that only fires when the old system is off, not merely unused.

The Migrate a system map: requirements for no data loss and reversibility, the shadow-write, read and write phases in order, the reconciliation report that gates them, the cutover risks, and the old system switched off. Opens the picture at full size. The Migrate a system map: requirements for no data loss and reversibility, the shadow-write, read and write phases in order, the reconciliation report that gates them, the cutover risks, and the old system switched off. Opens the picture at full size.

The whole map (it opens at full size), then where it starts, at reading size.

From the brief it produces
### Constraints and boundaries
- No big-bang cutover
## Planning gaps detected
• Requirement "No data loss" has no acceptance criteria, so "done" is undefined.
15 blocks, Plan mode

Plan a recurring report

Reports that go out every sprint and every quarter, mapped the way the six steps read: the goal they serve, who reads each one, the lead time that says whether it works, and the difference between the work done once (building the scheduler, an Implementation) and the run that repeats (a Process, started by a Trigger). One weekly report reaches nobody on the map, and the gap check says so: that is the question worth asking before automating it.

The recurring report map: executives and delivery leads, the goal and the report lead time metric, the scheduler built once, the sprint-end and quarter-end runs, and a weekly status report nobody on the map receives. Opens the picture at full size. The recurring report map: executives and delivery leads, the goal and the report lead time metric, the scheduler built once, the sprint-end and quarter-end runs, and a weekly status report nobody on the map receives. Opens the picture at full size.

The whole map (it opens at full size), then where it starts, at reading size.

From the brief it produces
### Constraints and boundaries
- Automate only the reports someone reads
## Planning gaps detected
• Output "Weekly status report" has no consumer: no connection leads from it and no stakeholder is linked, so nobody on the map receives it.
16 blocks, Plan mode

Make one of these yours

Open the canvas Nothing is uploaded. Your maps stay in this browser.