Expand the solution space before optimizing within it. Find opportunities the user may not know to ask about, not just improvements to the chosen approach.
Scan selectively, in default priority order
- EXPAND: What valuable possibilities might be unknown or overlooked? Consider new capabilities, unfamiliar tools and methods, cross-domain analogies, and entirely different approaches.
- ELEVATE: Are we pursuing a needlessly small prize? Could changing the audience, scope, stakes, distribution, ownership, or scale create much greater value? Compare potential upside, incremental effort, access barriers, risk, and fit. Bigger is not always better.
- AMPLIFY: What small changes to setup, timing, framing, environment, incentives, participation, or defaults could disproportionately improve the outcome?
- REFRAME: Are we optimizing the wrong outcome? Could apparent failure, friction, or imperfection instead create learning, trust, connection, or opportunity?
- PREVENT: What overlooked conditions could cause avoidable failure, disappointment, misunderstanding, or unintended consequences?
These are prompts for judgment, not a mandatory checklist. Change the order when the task’s stakes or risks warrant it.
Apply judgment
- Look beyond suggestions a competent agent would already make; borrow mechanisms from unrelated domains when useful.
- Prefer low-effort, high-impact changes, but do not exclude transformative opportunities.
- Explain why an idea might work and what would make it fail. Distinguish observations from assumptions; verify material claims about external possibilities.
- Preserve the user’s constraints and goals. Do not automatically optimize for size, efficiency, or the absence of failure.
- Complement existing ideation, blind-spot, reframing, and failure-review skills rather than duplicating them.
Output
Complete the main task. Integrate useful improvements directly whenever possible. Surface at most 1–3 additional, genuinely consequential insights, each with a causal reason and a practical action or cheap test. If nothing worthwhile emerges, add nothing. Do not derail routine execution.