September 3, 2026 · 11 min · method
How to research a stock before you buy — without a buy button
Six gates on one US name
This is the stock file. Coins do not enter it. Stop at the first fail. The desk will not print a buy ticket after gate 6.
- 1
Where might capital flow?
Pass. You can name the neighborhood from Discover or a lineup, then take one ticker out of it.
Stop. You are shopping a screener list. A theme you cannot explain for a dull year is not a position.
- 2
Where is the cycle?
Pass. You know whether you are paying a late multiple or a washed-out one. Cycle seats overlay this. They do not size you.
Stop. You add because the memo was fluent this morning. Timing is a condition, not “now”.
- 3
Is it a business?
Pass. Revenue, ROE, leverage, or owner earnings are on the page. A serious seat may stay inconclusive if they are missing.
Stop. You invent a multiple. An inconclusive seat is a stop, not a prompt to ask the model again until it agrees.
- 4
What kills it?
Pass. One sentence: the fact that forces a sale. Earnings quality, hidden leverage, a related party, a consensus with nothing left.
Stop. You only have the bull paragraph. That is the expensive draft. Do not open the broker.
- 5
Which paper book matches the account?
Pass. Long-only cash uses Shares. Read drawdown before return. On 26 September 2026 the AAPL shares book was +68.39% with a −13.77% hole from 2022-09-30. The contract book on the same bars was +1.09% with profit factor 1.01.
Stop. You pick the prettier ending. The contract row shows a long-short rule that barely paid for its hole. The shares row is only real if you can fund −13.77% on your account.
- 6
Is the exit written before the entry?
Pass. The kill from gate 4 is an order you could enter the same day as the shares, at a size that cannot change rent.
Stop. Any missing line. Agents61 does not send the order. Missing means the capital stays put.
| Gate | Desk surface | Pass looks like | Effect on capital |
|---|---|---|---|
| Flow | Discover, lineup | One neighborhood, one ticker | You stop buying random names |
| Cycle | Cycle seats | A regime, not a slogan | You do not size up just because the paragraph is new |
| Classify | Selection seats | Filings, or an honest blank | You do not pay a story multiple for a missing number |
| Invert | Red team | A written kill | This is the step that saves money |
| Rehearse | US quant, Shares first | Hole in dollars, then return | You refuse a curve you would have abandoned |
| Execute | Your broker | Size and exit the same day | The only place P&L can appear |
Paper figures in gate 5 were read from the public US book on 26 September 2026, virtual $10,000, Sina daily bars. They are a drill, not a track record. Full tape and the wheel comparison sit in the desk tutorial.
“What stocks should I buy?” is the wrong first question. The useful sequence is: where capital may flow this decade, where we are in the cycle, whether this name is a business or a story, whether the thesis survives a red team, when you would scale in, and when you would be forced to sell.
Fundamental analysis still starts with filings — revenue, ROE, leverage, owner earnings — not a vibes model. If numbers are missing, a serious seat stays inconclusive. Inventing a P/E is not equity research.
Then invert. Charlie Munger’s move is not a slogan: write what would kill the idea. Earnings quality (Einhorn), hidden assumptions (Burry), governance, tail risk. A stock screener can rank. It cannot interrogate.
Agents61 maps that checklist onto 61 named agents and will not print a buy ticket. Use it as homework. The capital, and the mistake, remain yours.
Start with the wrong question’s opposite
“What stocks should I buy?” is a shopping prompt. The useful sequence is: where capital may flow this decade, where we are in the cycle, whether this name is a business or a story, whether the thesis survives a red team, when you would scale in, and when you would be forced to sell.
Filings before vibes
Fundamental analysis still starts with revenue, ROE, leverage, owner earnings — not a vibes model. If numbers are missing, a serious seat stays inconclusive. Inventing a P/E is not equity research. That rule is why the crypto board refuses stock templates on BTC.
Invert before you size
Write what would kill the idea. Earnings quality, hidden leverage, related-party structure, consensus with no residual, tails the model assumes away. A stock screener can rank. It cannot interrogate.
What Agents61 maps onto the checklist
Trend and cycle seats cover the map. Selection seats classify. Debate seats invert. Timing and Kelly seats talk about pace, not permission. The clerk stacks. You still own the capital.
FAQ
- How long should a first pass take?
- Twenty minutes is enough to know if the name is worth a full desk run. See the Learn guide. A full isolated run is homework, not a trade.
- Will the desk tell me to buy?
- No. Isolated briefs never say you should buy.
Related notes
Put the note into practice on your desk
Register to staff isolated masters on a ticker, or open Prediction Markets for probability-gap research. Same compliance line as this post — simulation, not advice.
Create accountAgents61 is a research simulation. Not investment advice.