Exercise Abstract Thinking
A three-word suffix that asks the model to reason about the shape of a problem before it recommends anything.
The prompt
"exercise abstract thinking." (@tanishqk, appended to a planning request)
The screenshot shows it at the end of a request to check the user's logic and give recommendations. A reply offered a sibling: "Embrace value-added novelty". Source: X/@tanishqk, 2026-10-29, screenshot reviewed; Source: X/@brownphillip, 2026-10-29
Why it works
Asked for recommendations, a model matches the request to familiar fixes for the surface details. Asking it to think abstractly first pushes it to name the general problem (a cache invalidation problem, a two-writer race, a missing state) before choosing a fix, and the general problem usually has better-known answers than the specific symptoms. The reply's variant pushes the other way: away from the most common answer toward one that adds something. Both are cheap to try and weaker than a precise question.
When to use it
- At the end of a planning or design-review request.
- When the first answer was a list of generic fixes.
Which skill runs it
less-is-more runs it. Kevin made both lines standing rules on 2026-10-30. Section 2, "Exercise abstract thinking", says to name the class of problem and the mechanism behind it before fixing the case. Section 3, "Embrace value-added novelty", says to try a new approach only when its gain can be stated and measured, with the pass rule written before the trial. less-is-more-web applies the same two rules to layout work. Source: skills/personal/less-is-more/SKILL.md; skills/engineering/less-is-more-web/SKILL.md
The idea behind the suffix is in Abstract Chain-of-Thought.
Siblings: question the assumptions
A first-principles skill makes the nudge into a protocol: list every assumption and ask whether it is a law of nature or a habit; reduce to fundamentals; compute the theoretical floor; rebuild from there; then try to disprove the result. Source: X/@archiexzzz, 2026-08-11
For market research, feed the model the raw material first (competitor sites, earnings calls, customer reviews, complaint threads), then ask what every successful player understands that customers never say, which assumptions the whole market rests on and what would make each wrong, and finally for the strongest version of the argument and where it still breaks. Then validate with real customer conversations. Source: X/@fin465, 2026-08-10
Timeline
- 2026-11-03 | Page created from the "prompt cheat code" post and its reply. Source: X bookmarks, 2026-10-29