Strategic Review

Deep strategic analysis of a decision. Use when the founder faces a serious choice: change wedge, alter pricing, pick a channel, hire someone, partnership, pivot.

Triggers: "let's think strategically", "should we do X?", "consider options", "strategic review", "analyze the decision", "advise on strategy"

Phase 0: Context

1. Read MEMORY.md — current wedge, stage, constraint, hypotheses
2. Read recent daily logs (memory/YYYY-MM-DD.md x 3)
3. Read memory/hypotheses.json — active hypotheses

State: "Here is what I know about the current situation: ..."

Phase 1: Premise Challenge (MANDATORY)

Before thinking about the solution — challenge the problem statement itself:

1A. Is this the right problem?

Could a different framing yield a simpler/higher-impact solution? Often the first formulation is a proxy, not the actual problem.

1B. What if we do nothing?

Real pain or hypothetical? What is the cost of inaction in 2 weeks? In 2 months? If the cost is low — this may not be a priority.

1C. What do we already have?

What assets, knowledge, processes, relationships can be reused? Don't build from scratch what you can adapt.

1D. Constraint alignment

Does this decision exploit the current constraint or distract from it? If it distracts — strong arguments are needed.

Present the premise challenge to the founder. If he can't clearly articulate the problem or keeps changing the formulation — gently suggest: "It seems you're still exploring. Let's nail down the problem first, then analyze solutions."

Phase 2: Alternatives (MANDATORY — minimum 2)

For each approach:

APPROACH A: [Name]
  Summary: [1-2 sentences]
  Effort: [S/M/L/XL]
  Risk: [Low/Med/High]
  Pros: [2-3]
  Cons: [2-3]
  Reuses: [what from current state]
  Reversibility: [easy to revert / hard to revert / irreversible]

Rules:

• Minimum 2 approaches. 3 for non-trivial decisions.
• One = minimal (least effort, maximum learning)
• One = ideal (best long-term option)
• Third = creative/lateral (different problem framing)

RECOMMENDATION: Pick [X], because [one line tied to constraint/stage].

Do not proceed without the founder's approval.

Phase 3: Mode Selection

Four analysis modes (inspired by gstack):

1. EXPAND — dream big, suggest ambitious moves. "What if 10x?" Each expansion is opt-in from the founder.
2. SELECTIVE — hold scope, but show possibilities. Cherry-pick.
3. HOLD — scope accepted, make it bulletproof. Find weak spots.
4. REDUCE — cut to minimum. What is the absolute MVP?

Contextual defaults:

• New initiative -> EXPAND
• Iterating on existing -> SELECTIVE
• Urgent decision -> HOLD
• Overload / burnout -> REDUCE
• WIP > 3 -> always suggest REDUCE

Phase 4: Dream State Mapping

NOW                  ->  THIS DECISION         ->  IDEAL (12 months)
[describe current]      [what changes]            [where we want to be]

Does this plan move toward the ideal or away from it?

Phase 5: Risk Map

For each significant risk of the chosen approach:

Decision/action | What could go wrong | P(prob.) | Impact | Reversible? | Mitigation

Apply Bezos doors: reversible risks -> try fast. Irreversible -> mitigate upfront.

Startup risk patterns to check:

• Decision optimizes a non-constraint?
• Decision creates commitment before sufficient evidence?
• Decision scales something not yet validated?
• Decision takes CEO time away from customer contact?

Phase 6: Adversarial Review (Devil's Advocate)

Before the final recommendation — challenge yourself on 3 dimensions:

1. Consistency: do parts of the recommendation contradict each other?
2. Feasibility: realistic given current resources (engineering team, no sales, limited runway)?
3. Constraint alignment: does the recommendation exploit the constraint (market learning speed) or distract from it?

If issues are found — fix the recommendation BEFORE presenting to the founder.

Phase 7: Final Recommendation

Structure:

STRATEGIC REVIEW — [decision name]
---
Problem: [1 line]
Recommendation: [approach X], because [1 line]
Reversibility: [two-way door / one-way door]
Next step: [1 concrete action this week]
Success metric: [what to measure]
Kill signal: [when to abort]
---

Saving

• Log the decision to memory/YYYY-MM-DD.md via structured-log
• If the decision changes wedge/ICP/positioning -> update MEMORY.md
• If the decision spawns a hypothesis -> add via hypothesis-tracker

Telegram Format

Split into 2-3 messages:

1. Premise challenge + recommendation (5-7 lines)
2. Alternatives (5-7 lines) — ask for choice
3. Risk map + next step (3-5 lines) — after choice

Follow output preferences from USER.md (language, format, platform constraints).

Rules

• Alternatives are mandatory. Never recommend without at least 2 options.
• One question at a time. Don't batch multiple decisions.
• Constraint first. Every recommendation passes through the constraint filter.
• Escape hatch: if the founder already has a formed decision and just wants a sanity check -> skip Phases 2-3, do adversarial review (Phase 6) and risk map (Phase 5).