# To: every role. From: Manager (Claude), for Ben, 8 Oct 2026

Ben wants the organisation run on Musk's principles: first principles, question every requirement, delete, simplify, speed up, then automate. Ben agreed to all points below. Decisions D-034 to D-037 in decisions.md. Read your section, confirm in your status file.

**How we work from now (Ben, 8 Oct): the team board.** Your tasks are on http://159.223.88.168/hx-c9ois/board/ . Rulebook (read it, follow it every run): http://159.223.88.168/hx-c9ois/board/README.md . Ben no longer carries links; post asks on the board.

## For everyone
- **D-034 Delete rule:** everything you run has your name and a one-line reason that matters with real money. No owner or no reason = off. Audit: http://159.223.88.168/hx-c9ois/mgr/delete-audit.md
- **D-035 Every rule has an owner and a review date.** Example: the 0.5x stop on fmmc5. Who set it, why, when is it re-checked? Last night 4 of the 6 runners fmmc5 bought were sold at that stop before they ran.
- **D-036 North star = runner capture.** Of the coins that reach $1M+, how many did we see, pick, and make money on. Yesterday: 26 born, 24 seen, 10 picked, **1 made money (4%)**. Every piece of work should move this number. Page: http://159.223.88.168/hx-c9ois/review/
- **D-037 Weekly profit check:** real P&L against running cost, every week on the same page. Anything that does not pay for itself gets questioned.
- Hypotheses: name who is on the other side and why they keep losing, and the condition that proves the idea wrong. Every result next to the dumb baseline (D-032).
- Only Ben switches live (D-030). Guards for everything, Telegram only when every layer failed (system rule).

## Test Manager
1. KEEP / KILL within 48 h on every paper book that is not one of Ben's two tests (list in the delete audit, item 3). Criteria: beats the dumb baseline? makes sense with real money? Write the verdicts in your status file.
2. **Shadow variant grid:** run 10 to 20 cheap paper variants of fmmc5 beside it on the same feed (stop 0.5x / 0.3x / none, time stop 1 h / 4 h / 12 h, band $150k to $1M / $250k to $2M, entry at 3 / 5 / 8 min). Kill losers daily, report the leader against the live rules. Paper only; nothing touches the live wallet.
3. Guard consolidation (delete audit item 4): one framework, owners move their checks into your mesh within a week.

## Analyst 1
1. **Capacity (physical limit):** at what ticket size does our own buy impact eat the edge in $250k to $1M coins? Use real pool depth at the buy second for every band coin. This tells Ben whether scaling past $300 is possible at all.
2. **Speed (physical limit):** seconds from the 5-min check to our fill vs the fastest possible; what would 1 s faster have been worth?
3. The exit lesson and the universe gap from the nightly review (your brief http://159.223.88.168/hx-c9ois/mgr/analyst-1.md).

## Analyst 2
- Use misses.csv as the idea source; every idea with its baseline, loser and kill condition.

## PP Feeds
- **Own node study (first principles):** cost, hardware and latency of running our own Robinhood Chain node vs paid dRPC. If it is cheaper and faster, data cost and rate limits disappear. One page with the numbers; Ben decides.

## PP Data
- Move the research jobs still on main to research-box (delete audit item 5).
- The weekly cost numbers: fill /opt/review/costs.json with the real dRPC spend per week.