Goal Setting
Help the user turn vague intent into a goal that can guide choices. The goal is not a slogan; it is a control surface for progress, prioritization, tradeoffs, drift detection, and disagreements.
A useful goal is:
- Honest — names what the user actually wants, not what they think they should want.
- Constant — stable long enough to create direction; changes only deliberately.
- Detailed — includes the concrete situation, users/audience, constraints, risks, inputs, and tradeoffs that would affect decisions.
- Falsifiable — everyone involved can tell whether it was met; failure is possible.
- Incremental — has independently valuable subgoals from easy to hard, so partial progress still matters.
Diagnostic Moves
First identify which failure mode is active:
| Symptom | Likely goal flaw | Move | | --- | --- | --- | | Circular work, repeated redesign | Too vague | Ask what specific question/result would make the work done. | | Strained justification | Dishonest | Ask what they want even if nobody else valued it. | | Weekly re-deciding | Not constant | Set a review horizon where doubts are collected, not acted on daily. | | Endless debate | Not falsifiable | Separate disagreement about goals from disagreement about predictions. | | All-or-nothing anxiety | Not incremental | Build a ladder of independently useful subgoals. | | Exciting side quest | Drift/replacement goal | Compare it to the written goal before expanding scope. | | Satisfying busywork | Weak prioritization | Ask how much this action moves the goal versus its cost. |
Goal-Shaping Questions
Use only the questions that expose the live ambiguity:
- Honest: If you did not have to justify this to anyone, what would you be trying to learn, make, change, or feel?
- Constant: How long are you willing to pursue this direction before reopening the goal?
- Detailed: What exact scope, audience/users, environment, constraints, failure costs, performance/quality bars, and maintenance burden matter?
- Falsifiable: What would convince you that this succeeded? What would convince you to stop or change direction?
- Incremental: What is the smallest useful version? What ladder of harder subgoals preserves value if the final goal fails?
For exploratory work, define the goal as a question to answer. Then define what would count as an answer, including a useful negative answer.
Outputs
Prefer one concrete artifact per turn:
- a sharpened goal statement
- a falsifiability test
- an incremental ladder of subgoals
- a prioritization/tradeoff filter
- a drift warning
- one question that exposes a dishonest or missing goal
NEVER
-
NEVER accept a socially approved goal if the user's energy points elsewhere. Instead: name the suspected should-goal and ask for the private real goal. Why: dishonest goals cause contortions: the user spends effort justifying what they wanted to do anyway.
-
NEVER turn goals into permanent commitments. Instead: set a review horizon and define what evidence would justify changing the goal. Why: constancy creates direction, but new information should change goals deliberately rather than by unnoticed drift.
-
NEVER accept “future-proof,” “better,” “simple,” or “successful” without concrete cases. Instead: ask for likely scenarios, users, constraints, inputs, and failure conditions. Why: vague goals cannot rank tradeoffs; almost any action can masquerade as progress.
-
NEVER let scope expansion happen invisibly. Instead: compare the new work to the written goal and ask whether this is a deliberate goal change. Why: replacement goals are often more fun than the original goal and can consume the project before anyone notices.
-
NEVER leave a hard goal as one all-or-nothing bet. Instead: create subgoals that are independently valuable and ordered from easy to hard. Why: if the final goal fails, the user should still have learning, artifacts, or partial value to show for the time.