Loading
Every artifact the course produces, with the one question that tells you whether it is done. This page carries no arguments — the lessons do that, and they are worth reading first. This is what you keep open afterwards, while planning something real.
Before anyone proposes a solution.
The one-pager
Problem · Why now · Success criteria · Non-goals. One sentence to one short paragraph each.
Test: Could a reviewer reject the framing in five minutes, before anyone has built anything?
The load-bearing question
Five questions across unstated user, scale, constraint and success. Each marked load-bearing or nice-to-know.
Test: Would a different answer produce a different plan? If you cannot finish “if the answer were X, I would build ___ differently”, it is nice-to-know.
The assumption with an expiry date
What you are assuming, why that direction, and by when someone must correct you.
Test: Are you unblocked — and did you assume toward the larger scope, so a correction shrinks the estimate rather than growing it?
Turning an agreed problem into something reviewable.
Alternatives considered
Options as rows, your own success criteria as columns, plus cost. Rejected options stay in.
Test: Is any option eliminated by a criterion you already wrote down, rather than by taste or seniority?
The recommendation
What you chose, which stated goal it serves, what it costs — in one sentence.
Test: Does it name its own downside, and can you state the condition under which you would revisit it?
The circulation message
Who you want review from, and the single thing you are least sure about.
Test: Have you given the reviewer a job? “Any thoughts?” is not one.
The decision record
Decision · context · consequences · what would make us revisit it. Dated.
Test: In three months, would this answer “why are we paying for this instead of having built it?” in thirty seconds?
Putting a number on it without lying to anyone, including yourself.
The reference class
Which comparable past projects, measured to the same boundary you are estimating to.
Test: Is it work done by people in your circumstances? If you borrowed it or have none, do you say so out loud?
The range
Low to high, the branch that leads to the top end, and the date the range narrows.
Test: Is the top end derived from a specific named risk rather than from padding the bottom?
Standing behind it, and telling the truth as it moves.
The pre-mortem
Assume it is six months on and the project failed. List why, for ten minutes, quantity over quality.
Test: Did non-engineering causes appear — contracts, people, timing? If everything on the list is technical, you have not finished.
Retire or monitor
Sort every failure cause into one pile or the other.
Test: Could you convert it from unknown to known this week? Then retire it. Filing a retirable risk as monitored is deferral with paperwork.
The monitored risk
Owner (a person) · trigger (observable) · response (decided now).
Test: All three present? And is the trigger dated before the thing it watches, so you have room to act?
The dependency
What you need, from whom, and how long it takes to arrive — not how long it takes to do.
Test: Ask “what would this displace?” If they cannot answer, they agreed with you; they did not prioritise you.
The room
Say the range first. When pushed, offer a commitment at reduced scope or a forecast at full scope.
Test: Did you convert the pressure into a scope decision — theirs to make — rather than absorbing it privately?
The update
What changed against the plan · what it means for the date · what I need from you, by when.
Test: Could the reader act on it without replying to ask a question? And did anyone learn late something you knew early?
The reference-class row
Estimated range · actual · the dominant term in one sentence. Same boundary every time.
Test: Would you be comfortable if your skip-level read it exactly as written? If not, record that discomfort rather than softening the row.
The reversal
Lead with the trigger. Separate what changed from what you got wrong. Both costs. A new revisit condition.
Test: Have you deleted every apology for a decision that was right on the information available?
When the plan already exists and it is not yours.
The inventory
Check the list above against what actually exists. Note what is absent.
Test: For each missing artifact: what is now undecided, and who is silently assuming their own answer?
The fence
Before changing an inherited decision, find out why it was made.
Test: Can you state the reasoning behind it? If not, you are not yet in a position to judge whether it should go.
Reopen or inherit
Sort what you found. Reopen at most two things; inherit the rest deliberately and in writing.
Test: Is the inherit list explicit? Drifting past a decision is how you end up owning it with none of the reasons.
If a test here does not make sense, that is a signal to reread the lesson behind it rather than to follow it mechanically. The tests are compressions of arguments, and a compression you cannot unpack is a rule, which is the thing this course is trying not to teach.
Every claim on these pages links to its source. If a source looks wrong or out of date, check the resource list and tell your teaching agent — the course is meant to be corrected.