Were you right, or were you lucky?

Log your thesis before you buy. Conviction extracts your claims and checks them against reality — so your process gets graded separately from your P&L.

Start your journal See an example entry

Free while in beta · Private by default · Never investment advice

NVDA · longconviction 7/10 · felt calm

On record

Mar 02 · 14:11

May 18 · 09:47

Versions append. Nothing is overwritten.

The market prices this as a hardware cyclical while recurring software revenue quietly crosses 40% of mix…

expects 120–165 by 2027-03-31

Would prove this wrong

Software share stalls below 40% in Q2

Price at or below 95 · checked daily

NVDA — price has left your expected range. Still hold?

Hold with reasonsReviseClose

Enforced falsifiersImmutable versionsDaily reality checksReturns-blind scoringTimestamped exports

What would prove you wrong?

A thesis that can't be wrong isn't a thesis. Entries refuse to save without a falsifier — that friction is the product, not a bug.

Reality checks you didn't ask for

Your stated range and falsifier thresholds are checked against price every day. When reality disagrees, you're confronted: still hold? Hold takes written reasons.

A score P&L can't fake

The Process Score grades falsifiability, completeness, pre-commitment, and follow-through. It never looks at returns — a lucky win can't buy a good grade.

Why not a spreadsheet — or a chat with an AI?

Spreadsheets are silently editable and never talk back. A chat only confronts you if you ask — and hindsight-you never asks.

Conviction keeps an immutable, timestamped record and does the confronting unprompted. Trade journals grade your execution. This grades your reasoning.

Your next decision deserves a record.

Log it before the outcome can distort it. The first entry takes five minutes.

Start your journal