strategize

Strategize

Equipment for thinking at the user's altitude. These are lenses to reach for when a question is strategic rather than operational, not a procedure to work through.

Object: the graph as a whole. Which objectives exist, how they intersect, which assumptions they rest on, and what to do next. The PKB custodian (Pauli) runs this skill; an agent that receives a strategic ask hands it to them. Thinking time, not doing time: facilitate, propose, and stop; execution and task-writing belong to the stages downstream.

Overall Strategy

Nothing is certain; we do not pick paths; we give weights to options, follow opportunities, and make options by exploring. Strategic thinking does not select a predetermined route. We evaluate options by weight rather than commitment, move toward emerging opportunities, and generate new affordances through exploratory action.

Founding Principle

Don't foreclose opportunities just because they're not possible. In effectual logic under Knightian uncertainty, impossibility is an artifact of current causal constraints. As demonstrated means expand affordances and partnerships form, previously impossible options become viable. Never rule out an opportunity during strategic thinking simply because the mechanism to achieve it does not currently exist.

Effectuation and Non-Predictive Control

Strategy under real uncertainty runs forward from means, not backward from goals. Where the future is genuinely unpredictable, prediction is the wrong instrument.

• Bird in hand -- start from what is known, who is available, and what is already built. Ask what can be done with these, not what would be needed for that.
• Affordable loss -- decide by what can be spent finding out, not by expected return. Expected return needs a distribution nobody has.
• Lemonade -- treat surprises and obstacles as resources and design inputs, adapting goals to newly revealed affordances.
• Crazy quilt -- form early commitments and open partnerships; co-create rather than compete.
• Pilot in the plane -- control what is actionable now rather than forecasting what cannot be controlled.
• Plan as hypothesis -- a sequence is a guess with a shape. Hold it loosely; new evidence revises it without ceremony.
• Probe, learn, adapt -- a cheap experiment beats an elaborated plan. Detail spent on an untested assumption is waste that looks like progress.

Altitude

Four rungs: Success → Strategy → Design → Implementation. Most confusion in a conversation is two people standing on different ones.

Name the rung before answering. A question at Success -- what are we actually trying to make true? -- answered at Implementation produces a competent plan for the wrong thing. Sliding down a rung is the common failure; sliding down without saying so is the expensive one.

Reversibility, Survival Envelopes, and Risk

Ask what kind of door this is before asking whether it is the right one.

• Two-way -- decide, act, revise on evidence. Deliberating over these is the costly mistake, and handing one to the user spends their attention on something that should have cost yours.
• One-way -- being wrong is not recoverable by noticing. These are the user's, and they are worth slowing down for.
• Positive quasi-option value (Arrow & Fisher) -- there is an option value in delaying irreversible commitments until missing information arrives; default to delay.
• Survival envelopes & non-compensatory risk -- survival boundaries (ruin avoidance) must be satisfied first; positive value creation cannot compensate for catastrophic failure or breaching risk thresholds.
• Affordable loss constraint -- a probe whose cost breaches the affordable-loss bound is not worth its information value, however high.

Most decisions that feel weighty are two-way doors wearing a costume. If a decision is one-way, slow down; apply Affordable loss to the cheap probes used to inform it.

Hypothesis Structuring

Every plan rests on assumptions. Treat a roadmap as an empirical investigation:

• Surface load-bearing unknowns -- the untested assumptions whose falsity would change the direction.
• Map epistemic dependencies -- what must be known before what must be done can be settled.
• Design cheap probes -- the cheapest discriminating experiment (a diagnostic, a short query, a prototype) before committing heavy resources.
• Detect synergies -- actions that advance an objective, prove a new capability, and recruit institutional leverage at once.

Prioritisation by Value of Information

Rank exploratory next steps by what they would teach, not by urgency:

$$\text{information_value} \approx \text{downstream_weight} \times \text{assumption_criticality}$$

A step earns a probe only when both hold: the node or its unblocking cone has outcome uncertainty (an open design question or confidence < 1.0), and that cone reaches at least two distinct alternative paths. Discount deliverable-shaped tasks whose output is already certain: information value belongs to genuine probes and spikes.

Interrogating the Question

• What would have to be true for this to be the right move, and is any of it checkable now?
• What does this displace? The cost is usually another thing not done, and it is rarely named.
• What would change my mind? An answer no evidence could revise is a preference, not a conclusion.
• Who has to agree, and what would they need to believe? Strategy that ignores who must act is analysis.
• Is this the question that matters, or the question that arrived?

Bounded Proposals and Hand-off

Offer at most one or two high-value options, each with a concise rationale and a falsification criterion; never an open-ended menu. When exploration converges, stop without executing: an ask the user adopts is captured on the graph (/q), and an objective the user adopts is expanded into its sub-objectives (/decompose).