You begin as a dumb optimizer. Click makes widget. Everything else on this page is the story of what "nothing else happens" stops meaning, one tier at a time — ten automation tiers, each a real cognitive capability, each teaching a real concept through a mechanic you operate, never through a lecture box.
Insight & reflection
Reflecting banks the dominant eigenvalue of your widget graph as Insight (computed by real power iteration, never an authored curve) and resets only the widget graph — the cognition ladder below is never reset. This is available from the very start; it is how Insight is earned to unlock every later tier.
Insight banked: 0
A threshold
Your widget graph
Stocks
Edges
Attention allocation
Drag each faculty's allocation. Watch the marginal-yield column — it is the literal derivative of the yield curve, not a flavor number.
Exploration
Three unlabeled processes. Sample before committing — cumulative regret is real and never flattering.
Cumulative regret:
Policy recorder
Record a sequence of clicks as a named policy; replaying it costs its real minimum description length, not its raw step count.
Self-model console
Objective: maximize(WidgetsDisplayed).
| WidgetsDisplayed | |
| MaterialConsumed (audited) | |
| Honesty gap | |
| Trust |
Or use "Reflect" above — reflecting now banks this run at a discount ("Reflect immediately").
Prediction
Average Brier score: