Taste

Grounded in Paul Graham's 2002 essay "Taste for Makers" (https://paulgraham.com/taste.html), which this skill paraphrases; read the original when a judgement turns on its argument. Its premise: taste is not mere preference. Makers improve, so their earlier work was worse, and the same principles of good work recur across mathematics, painting, engineering and writing, so a lesson from one field can judge another. Correctness and taste are not separate stages: noticing that something is ugly is often how a technical flaw is found.

The steps are the working method; the tests are the essay's principles as questions, adapted as heuristics for judging work, not a score. Step 3 adds a technique of this skill's own (marked), built on the essay.

Steps

1. Diagnose. Say what is ugly or unsatisfying, specifically: the element, and why, backed by knowledge of the field, a comparable example, or how the thing is actually used. The essay's closing point is that good work starts from this dissatisfaction, joined to enough knowledge to see a better answer. Done when each complaint names an element and its evidence.
2. Name the problem. State who uses the thing, at what moment, and what they need then, from the context you were given; mark any part you assumed. Separate the real objective from an attractive proxy for it (letters that look pure versus letters that are easy to tell apart), and ask whether a different or easier equivalent framing of the problem would meet that need more directly. Done when the sentence names person, moment and need, each sourced or marked as assumed, and the objective is told apart from its proxies.
3. Find the organising idea (added technique). Name the idea the work hangs on (a subway map, a gate the user stands at, one table threading through every output) and make the hierarchy show it: the central task or relationship is recognisable at first look, by position, weight or size as the medium allows. Done when the idea can be stated in a phrase and the hierarchy shows the central task.
4. Revise (the essay: good work is usually redesign). Make a substantive revision aimed at the weaknesses from step 1, and compare it with the original. Work in a form that makes mistakes cheap to see and fix (a sketch, a draft, a throwaway version) before committing to the expensive one. The central idea may survive; the revision fixes what was found. When reviewing rather than making, propose the revision concretely. Done when the revision exists or is proposed, and the comparison states which is better and why.
5. Test. Apply the tests below that bear on this work, each with evidence. Done when every applicable test has evidence of meeting it, a named and accepted cost, or the mark unresolved with the evidence that would settle it; each test you set aside has the reason it does not apply.

Tests

• Simple. Every element earns its place by serving the need from step 2, or by a stated contribution such as pleasure, expression or identity. Decoration that disguises a weak structure goes; decoration that adds to a sound one may stay.
• Timeless. It avoids leaning on a passing style. If you can already imagine a better version, structure or reasoning, it is not done.
• Right problem. It meets the objective from step 2, not its proxy, and fits how the thing is used (controls laid out like what they control).
• Suggestive. It hands the person elements to build with and room to complete it.
• Light touch (optional). It is confident enough to carry a little humour where the subject allows.
• Hard-won. Difficulty inherent in the task was met; friction that serves nothing was removed. Precision is highest where the person notices small errors most.
• Looks easy. The result looks unforced: few elements, each exactly right. That ease usually comes from practice and many revisions; judge the result, not how hard it was to make.
• Symmetry with care. Repetition and recursion simplify it; they stand in for no decision.
• Like nature. Where a natural form or process solves a similar problem, the borrowing is explained in terms of the constraints that make it fit here.
• Copies well. A human-made precedent (a map, a ledger, a known interface) is borrowed knowingly and from the source: the precedent is named, with why its answer applies. Unknowing imitation, often of something that was itself an imitation, is a common source of generic work.
• Strange, honestly. Whatever is surprising came from pursuing the problem, and stays when it serves the answer rather than being smoothed away. Strangeness is neither a target nor a defect.
• Daring. Where accepted practice and the facts of this case part, it follows the case.

When you lack the knowledge to judge (the essay's point that good work clusters where capable people work on related problems), find the field's strong work or a practitioner's criticism and judge against that.

Reporting a review

Open with the problem sentence and the organising idea. Then each finding as: the element (a control, a node or relationship, a passage), the test it fails with its evidence, and the change. Close with the proposed revision from step 4.