Things to think with.

Your work

Everything runs in this browser. Saves belong to this browser and this address; another browser or the local preview has separate storage. Clearing browser data clears local work. Export useful work for a durable copy. Scaffold JSON exports can be imported as new workspaces; model exports capture settings and results for inspection.

Examples are fictional. There are no external AI calls, data feeds or runtime libraries. Each model exposes its mechanism and assumptions; the numbers describe that mechanism and are not calibrated forecasts.

M01 · The commitment spiral

A team model connects accepted work, delivery effort, delayed rework, fatigue and trust. Replay the same demand with different policies. These values are illustrative, not calibrated organisational forecasts.

M02 · How teams fit the work

Six capabilities sit within three team boundaries. Work waits for capacity and incurs boundary delays; wider teams add adjustable coordination costs. Different workloads can favour different arrangements.

M03 · Exploration versus delivery · Archived

Allocate ten weekly effort points over twenty weeks. Noisy discovery, delayed delivery feedback and learning through use interact. Replays share the same underlying situation; useful output is a fictional proxy, not business value.

M04 · Where flexibility gets trapped

A fictional battery day connects state of charge, power limits and commitments. Offers arrive progressively, and comparisons share the same scenario and available information. It is an optionality experiment, not trading or asset valuation.

M05 · Steering through delay

A reservoir separates action, physical response and delayed observation. Compare feedback rules on the same withdrawals. Replaying a fixed pattern is informed practice, not a blind skill test.

M06 · How exceptions accumulate

Six quarters reveal requests progressively. Setup, ageing upkeep, shared-surface friction, reuse and retirement create consequences. Fictional costs are exposed; a useful exception can remain a good bargain.

M07 · Specialists and shared knowledge

Coaching takes time from mentor and learner; skill improvements arrive at week end. Work consumes a fixed weekly capacity and learns slowly through doing. Divisible work, fixed demand and skill scores omit certification and staffing realities.

M08 · Local wins, collective losses · Archived

Four teams choose among three actions with visible cross-team payoffs. Rewards mix own and shared outcome; simultaneous myopic responses may settle or cycle. The payoffs are teaching assumptions, not measurements of actual incentives.

M09 · When predictions change behaviour

Price-taking responses collectively change the realised price, which updates the next forecast. Common and private signals use matched update speeds, noise amplitudes and starting estimates. Repeated fictional clearing rounds omit inventory, state of charge, bids and market calibration.

M10 · Accurate enough for which decision?

Swap a fixed error budget between eight threshold decisions. MAE and RMSE stay fixed while asymmetric losses can change. Reallocating known errors is a hindsight experiment, not a trainable forecasting method or valuation.

M11 · Almost reliable parts

All-required and either-sufficient gates combine independent component states exactly. Reusing a node expresses a shared dependency. Hidden correlation, recovery time, load and repair are outside the model.

M12 · Crossing the adoption gap

Twelve people compare old and new utility using their partners’ current choices and switching friction. Pilot commitment, bridges and compatibility are explicit. Fixed preferences and simultaneous decisions omit persuasion, uncertainty and learning.

S01 · Reframing workbench

Develop several descriptions of a situation, state what each hides, then choose an intervention to test. Example text can be replaced. No AI evaluates the problem or decides which frame is correct.

S02 · Possibility mixer

Generate and map the same concepts using any two dimensions. Placements are hypotheses about fit; captured ingredients and gap context stay intact. Counts show coverage, not usefulness. Copy earlier Territory work without changing its original.

S03 · Constraint playground

Move constraints into imagined remove, reverse or exaggerate experiments. Restore the real condition and develop an adaptation. A thought experiment does not change physical limits, agreements or permissions.

S04 · Analogy workshop

Map roles and relationships, name where the match breaks, and develop an adaptation. Examples provide mechanisms to investigate, not evidence that transfer will work.

S05 · Intervention workbench

Connect explicit causal hypotheses, state conditions and rival explanations, then compare interventions by predicted observations and bounded tests. Branches preserve their sources. Moving or reviewing a claim does not establish causation.

S06 · Three genuinely different answers · Combined into Alternatives

Open the Alternatives workbench to copy earlier work into the shared collection.

Develop three distinct answers, retain their weaknesses, then compare against your criteria. Borrowing keeps a source snapshot; it does not automatically produce a coherent combination or evaluate quality.

S07 · Alternatives workbench

Generate options, respond to concerns or reconcile proposals in one collection. Retain parent and borrowed-benefit snapshots; compare mechanisms, benefits, costs and assumptions against your own criteria. Connections record intentions, not proof that a benefit survives. Tests and evidence remain your judgement.

S08 · Unexplored territory · Combined into Mixer

Open Possibility mixer to copy earlier maps into the shared concept collection.

Editable axes locate your ideas and expose gaps and concentrations. Empty space might reflect an overlooked assumption or a good reason to stay out; a gap is not evidence of an opportunity.

S09 · Scene-first invention

Specify actor, trigger, knowledge, uncertainty, available actions and stakes. Place interventions in the sequence, then vary the scene without losing the original. This is a scenario rehearsal, not evidence of user behaviour.

S10 · Idea family tree · Shared capability

Open Alternatives to copy earlier Family Tree work into the workbench. Mixer also supports branching, combinations and a generation view; Reframing retains reasons and source snapshots.

Fork and combine concepts while keeping parent snapshots and reasons for change. Park ancestors without deleting their history. Lineage supports comparison; it does not score originality or promise.

S11 · Questions before answers · Combined into Reframing

Open the combined workbench to connect questions to frames and copy earlier Questions work into a new problem.

Branch questions across lenses before developing selected ones. Separate provisional answers, evidence and possible approaches. Answer status is your judgment, not an automated verification.

S12 · Productive disagreement · Combined into Alternatives

Open the Alternatives workbench to copy earlier work into the shared collection.

Make each proposal’s benefit, context, costs and disconfirming evidence explicit. Explore contexts, sequences or a new mechanism. Marked benefits express an intention; a test must establish whether they are protected.