Verdikt · a JAK Labs product
Every trading book tells you to keep a journal. Almost nobody does.
So this one is built like Duolingo instead of like a spreadsheet. Streaks, XP, leagues and daily challenges wrapped around trade logging — the habit traders never keep, turned into one they actually do.
the point isn’t the streak. it’s that you have fourteen days of your own decisions written down.
Why gamify it
Discipline is a habit problem, not an information problem.
Every trader already knows they should log entries, exits, position size, and what they were feeling when they clicked. The knowledge was never the gap. The gap is that a journal is homework, and homework loses to a live chart every single time.
Streaks, XP and leagues are not decoration here — they are the mechanism. They make the boring, correct thing the thing you come back for. What you end up with is the actual asset: a written record of your own decisions, with the reasoning attached, from before you knew how it turned out.
What’s in it
The log
Entries, exits, size, setup, and the reasoning — captured at the moment, not reconstructed after the fact.
What you felt
The half of trading nobody records. Tilt, hesitation, revenge entries — logged next to the trade they produced.
Your own numbers
Honest P&L math over what actually happened. No projections, no simulations.
Position sizing
Work out risk before the entry rather than justifying it afterwards.
Structured lessons
68 lesson sections and a 48-question graded bank, so the streak has something to feed on when you haven’t traded.
Leagues & challenges
XP, coins, daily challenges and league placement — the loop that gets you back tomorrow.
Auto-import
It reads your fills so you don’t retype them.
Connect an account and filled trades import themselves. Retyping is where journaling actually dies. Twelve connectors today:
While you trade
A coach that watches you, and is structurally unable to watch the market.
Mid-session you need a signal, not a paragraph. So the live coach is a pure function — no model, no network, no prompt. It compares what you are doing to the plan you wrote, and it can run on every single fill.
Adherence, not opportunity
Trades taken against your plan’s cap. Setups outside your written plan. A session your plan doesn’t name. Setups the engine already ruled out.
The tilt tells
Consecutive losses, surfaced against your own stop-trading rule. Risk going up after a loss. A day spent entirely outside your proven cohorts.
Every one of those is about restraint or adherence. None of them can be “take this trade.” That isn’t a policy — a function that compares numbers to a plan is structurally incapable of suggesting a position. Zero latency, zero cost, and nothing that could drift off-script.
The language model gets the weekly review instead, where synthesising a week of your behaviour is genuinely the value and a few seconds of latency costs nothing.
Finding your edge
Every journal will show you a win rate. This one refuses to, until the sample earns it.
Your fills get split into the cohorts you actually trade — by setup, by session, by instrument — and each cohort gets a verdict rather than a flattering statistic.
Insufficient
Under 30 closed trades in that cohort, no judgement is offered at all. A 62% win rate over 11 trades is noise, and the product says so instead of congratulating you.
Undetermined
Past the sample gate, but the data still cannot separate a real edge from zero. This is not “you have no edge” — it is “keep going,” with an estimate of how many more trades would settle it.
Proven
The 95% confidence interval clears zero, over 10,000 bootstrap resamples, at an expectancy worth the screen time. Only then does the app agree you have something.
This is where the name comes from. Not a dashboard, not a score, not a streak — a verdict, delivered only when your own trades have earned one.
Two rules sit above all of it, and they are the reason to trust the third verdict when it finally arrives: an undetermined cohort is never treated as an edge, and nothing in the product ever implies that trading more is progress.
This is the arc the journal is actually for. Log honestly → the session coach holds you to your own plan while it is happening → the weekly review shows you the pattern you cannot see from inside it → and after enough real trades, the engine tells you which of your setups is genuinely yours. Months, not days. That is the honest timescale for the question, and anything that answers it faster is guessing.
About your keys
It cannot place a trade. That is a property of the design.
- Connections are read-only. The connectors request read access and nothing else. Even a total compromise of this app does not move your money, because there is no code path that places an order.
- Keys are encrypted at rest with AES-256-GCM — not stored as plain text, not logged.
- It is your record. The app exists to read what you already did and show it back to you accurately.
You are handing a third-party app credentials to a brokerage account, and you should be suspicious of that by default — including here. Use read-only API keys, scoped to read-only at the broker, every time, with any product. That advice costs this app nothing and it is the correct habit regardless of who is asking.
The advice boundary
The coach comments on the trader, never on the market.
That line is what separates software you operate from personalised investment advice, which is regulated. A line stated only in a document is not a control, so it is defended in three places — deepest first:
- 1.
Input — the real control. The bundle handed to the coach holds your verdicts, your written plan, your logged emotions and your session times. No price, no quote, no news, no market state. It cannot form a view on an instrument because it is never shown one.
- 2.
Prompt. States the boundary, and more usefully what to do instead: compare against your plan, quote your own figures, say plainly when the data supports nothing yet.
- 3.
Output — a backstop that fails closed. If directional phrasing is ever detected, the entire review is withheld rather than edited. A redacted recommendation is still a recommendation.
The filter is deliberately the weakest layer, and is documented as such. A regex cannot prove text is advice-free — it exists to catch a model that drifts, not to make an unsafe design safe. That is why the input layer is first.
And the reward loop is kept away from the order button. Execution lives on the desktop surface — charts, order entry, your journal and edge verdicts beside them — carrying no XP, streaks, coins or badges. The gamified client has no order entry on any screen size. A reward loop that terminates in a buy button is a pattern regulators have fined brokerages over, and separating the two clients makes that structurally hard rather than something a reviewer has to remember.
Status, plainly: the journal, the gamification and the broker import are live today. The professional desktop surface is designed and decided but not yet released — so if you are here for order entry across monitors, you are early.
What this is not
It is a journal. It does not tell you what to buy.
There are no signals here, no alerts, no copy-trading, no strategy anyone is selling you, and no claim — stated or implied — that keeping a journal will make you profitable. It records what you did and shows it back to you honestly, including the parts you would rather not look at. What you do with that is entirely yours.
Start the streak
Fourteen days from now, you have a record.
Free to start, read-only broker access, and an engine that will tell you plainly when your own numbers do not say anything yet.
Verdikt lives at verdikt.jaklabs.io. Signup and the app all happen there.