Ask AI what already contradicts your trade.

Two weeks ago, AI argued the other side of your trade. Last week, it surfaced the assumptions underneath it. This week, the third of the five questions:

What evidence currently contradicts my thesis?

The word doing the work is currently. The first question asked what could go wrong. This one asks what is already wrong: the facts available today that don't fit the story you've built.

"Find me evidence that supports this trade."

Few people type it that bluntly, but most research prompts lean that way. You go looking for reasons, and you find them, because there are always reasons. An AI coach asked to help will find them faster than you would, and present them more neatly.

You've already filtered the evidence

By the time a trade idea feels ready, you've usually stopped looking. The information that fitted made it into your reasoning. The information that didn't was put down to noise, or never noticed.

Take a hypothetical long on a large-cap share after a strong earnings report. The headline figures beat expectations, so the thesis writes itself. But the same report might carry softer guidance, a narrowing margin, or a line about weaker demand in one region. The share may have barely moved on the day. Each of those is evidence. None of them made it into the two sentences you'd use to explain the trade.

The aim isn't to find a reason not to trade. It's to put back what you left out, and then decide.

A prompt that does the job

"Here is my trade idea: [instrument, direction, entry, stop, and timeframe]. My thesis: [two or three sentences]. Here is the information I'm working from: [paste or summarise the report, chart notes, or news]. Using only this information, list every piece of evidence that currently contradicts or weakens my thesis, and point to the specific item for each. Don't include hypothetical risks or things that might happen later. If the evidence is mixed, say so plainly rather than balancing it out."

Three things make it work:

1. It hands over your source material and limits the AI to it. That keeps it from inventing facts, and makes every point something you can check.
2. It asks for evidence, not risks. "The trend could reverse" doesn't count. "Margins narrowed in the latest report" does.
3. It tells the AI not to soften the answer. Left to its defaults, an AI coach tends to pair every awkward point with a reassuring one.

What to do with the answer

Check each item against the source first. If the AI has misread something, discard it. Then sort what's left:

1. Explained. You can say why it doesn't change the thesis, and the reason still holds once it's written down. Keep the note.
2. Unexplained, but minor. It doesn't break the thesis, but it shouldn't be ignored. Make it something you watch, and decide now what change in it would take you out.
3. Unexplained, and central. It contradicts the main reason for the trade. Your thesis is weaker than it felt an hour ago. Wait, reduce size, or pass.

Then record the contradicting evidence in your journal, next to the thesis rather than underneath it. When you review the trade, you'll see what you knew at the time and how you chose to weigh it. That is the part you can learn from. The outcome on its own tells you far less.

A thesis that has survived its own contradicting evidence is worth more than one that has simply never met it.

Next week: what am I potentially overlooking?