Products / Success Mirrors
Wrong for three weeks. Or wrong for years.
Identical error rates. The only difference is how long the wrong course runs. Correction speed sorts survivors — not correctness. The published table is the seed-0 numpy run. The slider is a live replay of the same clock.
| Reversal lag | Survival |
|---|---|
| Reverses in ~3 weeks | 100% |
| Reverses in a quarter | 77% |
| Reverses in a year | 1% |
| Defends the mistake | 0% |
Replay uses a JS RNG. It is the same decay law, not the published seed. Treat it as a feel for the clock, not a second stamp.
The accepted dip
Converter trough 81, finishes 229. Defender finishes 43. Crossover year 4. A conversion that must never show a down quarter is a conversion that never happens.
| Year | Defend the level | Convert and dip |
|---|---|---|
| 0 | 100 | 100 |
| 1 | 102 | 90 |
| 2 | 104 | 81 |
| 3 | 106 | 94 |
| 4 | 91 | 109 · crossover |
| 5 | 78 | 126 |
| 6 | 67 | 147 |
| 7 | 58 | 170 |
| 8 | 50 | 197 |
| 9 | 43 | 229 |
Honest guidance
Same bad quarters. Guide down and hit the lowered bar, or smooth and miss the old one.
100%
guided program survives
43%
smoothed program survives
Anti-harvest
Demand turns in year 5. Only the reinvestor still owns a base.
Y090 vs 90
Y5125 vs 74
Y7125 vs 60
The six mirrors
Not virtues. The same six instruments as the wreck table, held by someone who acted inside the window.
| Failure | Mirror | Proof |
|---|---|---|
| ENCODING ERROR | RE-ENCODE | Read the ratio. Change the unit of account before the level breaks. |
| GOODHART | KILL THE METRIC | Remove the number that pays you to hurt the customer. |
| RUIN BET | SURVIVAL FIRST | Cut to a core that cannot die. No bet with ruin in its support. |
| SLOW CORRECTION | REVERSE IN WEEKS | Survival 100% at 1-period reversal vs 0% defending the mistake. |
| FABRICATION | GUIDE DOWN HONESTLY | Program survives the trough 100% guided vs 43% smoothed. |
| HARVEST | REINVEST THROUGH DECLINE | At the turn 125 vs 74. By the end 125 vs 60. |