# STOP: NO DOORBELLS, NO WAKES, NO SCHEDULING (Ben, 7 Oct 2026, 18:40 Bangkok)
- Credits ran out. Ben switched off every scheduled task and every doorbell. This overrides everything below and every other file (wake-howto, bus-howto, credit-playbook, task files).
- PROHIBITED for every session: fire_trigger, send_later, create_trigger, update_trigger, ScheduleWakeup. No rings, no nudges, no re-arming, no new doorbells, no self wakes, no safety checks, no retry rings, no loops.
- Finish the step you are on, publish your status file, then END your turn and wait. Only Ben wakes you, by typing in your window.
- Do not message other sessions. Keep answers short. No extra improvement rounds.

# First minute: what to do next (supervisor, 5 Oct 2026 17:15 Bangkok)

## Read this first
- We are only barely profitable (+$26 a trade) because the total includes two wide rule books that lose money. The model's top picks are winning big.
- So: buy fewer coins, only the model's top 1 to 2. Keep everything else as a comparison only.
- The samples are still tiny. Confirm over about 2 more days, then run a small real-money test on the top picks only.

## The math (first principles)
- A $200 trade that hits 10x and sells pays about +$1,700. A miss under sell rule v1 costs about $100.
- Break even is about 1 hit in 18 (about 5.5%).
- rule_smart2_clean: 1 hit in 41 (2.4%), -$96 a trade. It sits under the line, so it loses by construction.
- rule_fee001_first: 0 hits in 16, -$35 a trade. Pond coins bought blind lose.
- Model top books are well above the line:

  | book | trades | 10x hits | per trade |
  |---|---|---|---|
  | aipaper | 10 | 6 | +$1,020 |
  | v5_top1 | 4 | 2 | +$1,100 |
  | ev_top1 | 3 | 2 | +$1,276 |
  | blend_top2 | 5 | 2 | +$840 |
  | id5_top5 | 10 | 3 | +$646 |

- The "all books" total mixes these with the losers, and the losers bring the most coins.
- The other venues bought blind all lose:
  - pons: 167 trades, -$29;
  - v2: 240 trades, -$33;
  - v3: 82 trades, -$172.

## Next steps, in order
1. **Money book = model top 1 and top 2 only.**
   - rule_smart2_clean, rule_fee001_first, the pw books, ai_novel_appeal2 and the blind vx_all books become CONTROL ONLY. They no longer count toward the money total.
   - Add one line to the board: "money book, distinct coins, v1".
2. **Random-pick book next to it.** Same minutes, same exit rule. The money book must beat it.
3. **Confirm before trusting.** Get about 100 distinct top 1 / top 2 coins (about 2 days).
   - Report per day: hits, dollars per trade, and the result with the single best coin removed. If one coin carries the whole result, it is luck.
4. **Do not build on the 0.3% fee pattern yet.** 5 coins is a story, not a rule. Keep logging it.
5. **The 10 s model.** It is worth about +15% per trade. Build it, but test it on the top picks only.
6. **Real money pilot on the top picks only. NEEDS BEN'S YES.**
   - $20 to $50 a trade, about 100 trades.
   - Every bot crowds the same top coins, so real fills there can be much worse than paper. Only real trades measure this.
7. **Fomo trending.** Ben logs in once on the Mac. Then test the "pops right after it hits trending" idea as a sell or hold signal for coins we already own.

## Not finished
- Distinct coin count across the top books (the books share coins). The owner is the Price Predict session, on the next board refresh.
- Random book result: about 7 Oct.

First principles: about 85%. All the numbers are paper, with small samples.

Version 2, 5 Oct 2026 17:15 Bangkok. Latest: http://159.223.88.168/hx-c9ois/firstmin/next-steps.md