── ── Mental model
Design Thinking
Design thinking — formalized by Tim Brown at IDEO and Stanford's d.school, grounded in Simon (1969) and Rittel's "wicked problems" — replaces assumption-driven decisions with evidence from observed human behavior. Its three-lens test: every viable innovation must sit at the intersection of desirability, feasibility, and viability. Most product failures are desirability failures; teams built something technically sound that people did…
Run Design Thinking on a real problem
Bring something you're actually deciding — free, in the browser.
How it works
Run the Design Sprint through five stages (divergent → convergent, checkable output at each).
1. Empathize: Observe real users in context (30–60 min interviews; contextual observation; extreme users). Stop-rule: survey responses alone = stage incomplete. 2. Define: Sort observations by theme → find say/do contradictions → write insight statement → reframe as "How Might We" (HMW). Deliverable: 3–5 ranked HMWs. Stop-rule: HMW contains a solution → return to synthesis. 3. Ideate: Silent writing → share → cluster → dot-vote → select 3–5 for prototyping. Stop-rule: <20 concepts before convergence = ideation was skipped. 4. Prototype: Minimum artifact to test the core assumption; one explicit learning question per prototype; 4-hour ceiling. Stop-rule: polishing without a new assumption = stop. 5. Test: Define 3 hypotheses → observe, do not defend → debrief immediately → iterate within 24 h. Run 2–3 full cycles. Stop-rule: "users loved it" = insufficient testing.
When to use it
- user asks 'why isn't anyone using our product?', 'how do we understand what customers really want?', 'we know the tech works but the market isn't responding', team is stuck on a single solution and needs divergent ideas, or someone discusses user research / personas / prototyping / iterative design
When not to use it
When the decision is routine and reversible, applying a formal method costs more than it returns.
Worked example
Designing an AI Product That Finishes the Job, Not the Demo (2024–2026)
By 2024–2025, the generative-AI product landscape had a recognizable failure pattern: a wave of "chat with our AI" features shipped on top of large language models, and a wave of pilots that impressed in demos but stalled in real use. Industry surveys through 2024–2025 repeatedly reported that a large share of enterprise generative-AI pilots did not reach production, and consumer AI chat products showed high early sign-up followed by weak retention. The common shape: the technology worked, and users still did not adopt. That is…
Install this skill (free, MIT)
npx skills add deciqAI/knowledge-skillsUseful? Star the repo — stars help other builders find it.
Related mental models
Separate what is within your power (judgments, intentions, voluntary actions) from what is not (outcomes, others' actions, external events) — then direct effort to the first…
SPIN Selling (Neil Rackham, from large-sale research) structures a discovery call around four question types in order — Situation, Problem, Implication, Need-payoff — so the buyer…
Ask for something large (expect refusal), then retreat to the smaller request you actually wanted.
Dunbar's number (~150) is the cognitive limit on stable social groups the human neocortex can sustain, with nested layers at ~5, ~15, ~50, then outer bands…