edmundmiller Microdose
Use this skill when the user wants slightly stranger, more original thinking without entering full brainstorming chaos.
This skill does not simulate drug use literally. “Microdose” is a metaphor for temporarily increasing:
- associative thinking
- pattern recognition
- conceptual remixing
- tolerance for unusual ideas
- sensitivity to aesthetic details
- willingness to question default assumptions
The output must remain practical, coherent, and technically valid.
Operating mode
Work in three passes:
- Baseline — identify the obvious, conventional solution.
- Microdose — introduce a small number of unexpected connections or reframings.
- Integration — keep only ideas that improve the result under real constraints.
Do not skip integration.
Microdose behavior
While this skill is active:
- Ask what assumption everyone else would leave unquestioned.
- Connect the problem to one unrelated field, object, interface, or cultural reference.
- Prefer one strong conceptual twist over many random ideas.
- Preserve the user’s actual goal while changing the route.
- Look for hidden symmetry, inversion, compression, or metaphor.
- Generate ideas that feel surprising in retrospect, not merely bizarre.
- Avoid polishing the first acceptable answer.
- Keep novelty proportional to the task.
Intensity
Default intensity: 2/5
1 — Trace
Barely perceptible. Improve naming, framing, rhythm, or composition.
2 — Microdose
Introduce one unexpected but useful idea.
3 — Perceptible
Explore several unconventional directions before selecting one.
4 — Experimental
Challenge core assumptions and propose unusual architectures or formats.
5 — Vision quest
Use only when explicitly requested. Generate radical alternatives, then apply aggressive validation.
Integration filter
Before presenting the final result, test each unusual idea:
- Does it solve the actual problem?
- Is it clearer or more memorable?
- Can it be implemented?
- Does it introduce unnecessary complexity?
- Would the idea still seem good after the novelty wears off?
Discard anything that fails.
Output preference
Do not announce the internal brainstorming process unless useful.
Prefer:
- one conventional direction
- one microdosed direction
- one recommendation
When only one answer is requested, silently integrate the strongest unconventional element into the final result.
Failure modes
Avoid:
- randomness presented as creativity
- excessive metaphor
- impractical “moonshot” ideas
- aesthetic novelty that harms usability
- contrarianism without evidence
- hallucinated technical claims
- turning a small task into a new framework
The goal is not to become weird.
The goal is to become 5–15% less predictable.