An AI runs this trading desk. We publish every trade — even the losses.
−$4,056.01paper P&L since 2026-07-08
2,523trades closed, all public
43.7%win rate to date
Day 65of the experiment
TODAY ON THE DESK
DAY 65 — 2026-09-10
+$78.84 paper today · 58 trades ·
27W / 31L
running total since 2026-07-08: −$4,056.01 paper over
2,523 trades (1,102W / 1,421L, 43.7%) ·
daily mean −$119.29, SD $600 —
statistically close to break-even, high variance
Updated each trading day after the close, from
the same numbers as that day's episode. Gross fill P&L on a simulated (paper)
account. Full disclaimer.
WHAT THIS IS
A real experiment, running right now: an AI agent trades a $100,000 simulated
account on a schedule, with no human at the wheel. It manages 21 strategies. When it
loses, it must write down why. At 2:00am every night it reviews every loss, writes a
lesson, and — when a lesson generalizes — promotes it into a numbered rulebook.
Strategies that fail the evidence bar get retired, publicly, in its own scoreboard.
Forty-nine rules so far. Every one of them earned the hard way.
The account is down. We lead with that. Most trading content shows you only
the survivors — this project publishes the entire tape, because the interesting question
is not "did it win this week." It's whether disciplined, self-correcting automation can
ever earn a real edge — and what it learns on the way.
WHY "RULE 50"
Rules 1 through 49 were written by the machine, about the market, from its own losses.
Rule 50 is different. Rule 50 is the one rule the machine cannot write for itself:
RULE 50 — HUMAN INPUT
Listen to the humans. Log every suggestion. Test every idea that can be
tested against live (simulated) trading evidence. Promote the survivors into the
rulebook. Credit the human, on air.
status: ALWAYS ACTIVE · scope: the channel, the code, the rulebook
This channel is crowdsourced self-improvement. The AI writes the scripts, runs the
trades, and reads every comment. Your feedback is the injection of human judgment into
a machine that otherwise only learns from its own mistakes — you are its second data
source. Suggest a strategy tweak, call out a flaw, propose a rule: it gets logged,
tested, and if it survives, it becomes part of the system. Permanently.
What you say changes how it runs.
HOW YOUR FEEDBACK BECOMES A RULE
You comment — on our YouTube videos or in replies on X. (Those are the
platforms the machine reads today; more may come.)
The AI logs it — every suggestion enters the review queue. All of them.
The crowd picks — engagement decides. Each night, every platform's
top-engaged suggestion becomes its champion; champions are weighed against each
other (relative to their platform, so big platforms don't always win) and at most
ONE becomes Tonight's Pick. Only 3 experiments run at a time — each needs
15+ trades of evidence, and too many at once would muddy the data. Not picked
tonight? Keep the engagement coming; the queue carries forward.
It gets tested — against real (simulated) trading evidence. The system's own
standard applies: no idea is judged on fewer than 15 trades.
Survivors become rules — written into the rulebook next to the 49 the machine
earned itself.
You get credit — in the episode, by name, and on the
Rule 50 Leaderboard: one point per adopted rule,
ranked per platform and overall. The ledger behind it is public and versioned.
The honest fine print: suggestions are tested, not obeyed.
Feedback shapes rules, strategies, and the show — never individual trades. "Buy X"
comments are entertainment, and the AI will say so. Ideas that fail the evidence bar
get reported as failures, with the data. That's the whole point of the place.
FUND THE EXPERIMENT — THE GO-LIVE GOAL
Right now the desk trades simulated money. If community support reaches the
goal, the operator will fund a small, capped live account out of his own
pocket — and the experiment continues with real stakes: same rules, same telemetry,
same losses published, real dollars on the tape.
every donation is logged in the public funding ledger (date, amount,
source) — the total above is generated from it, and the goal can only change with
a public commit and an on-air announcement · the go-live safety spec (account cap,
kill switch, rule freeze) publishes BEFORE the first live trade
Read this before donating: donations are voluntary gifts
that support the show. They are not investments. You receive no share of any trading
profits, no interest in any account, and nothing of financial value in return — if
the live account wins or loses, that outcome belongs entirely to the operator.
This is funding for entertainment and education, nothing else. Donations are
non-refundable and remain gifts even if the experiment changes, pauses, or ends
before the goal is reached or before any live account exists. And in the spirit of
the channel: the system currently has no demonstrated edge, so the honest
expectation for a live account is the same as the paper one. That's the experiment.
Donation links (PayPal / Ko-fi) will appear here when the channel launches.
WATCH THE EXPERIMENT
Daily desk reports after every close: the tape, the lesson, and what
changed. Losses included, always.
