Who does what
Gudmundur · owner
Decides scope and priorities, approves every change to live trading, owns the servers, data and broker accounts, and signs off each phase.
Matt · co-owner and trader
The human voice: weekly live review of learners' simulator trades, his own real trades in the honest ledger, the trader's view on every lesson. Teaches; never calls trades.
Katherina · first learner and builder
Learner number one: takes every path first and reports what confused her. Helps with lesson text, the site, videos and cohort support, under the owner's supervision. Never holds broker credentials.
Claude · project manager and builder
Writes the tasks and definitions of done, builds with Claude Code behind feature flags, runs the tests, keeps this plan current, and reports honestly, failures included. Fable reviews the big decisions.
Four phases and a decision
Phase 1 · The honest core
- Ask KAT v1 for Gudmundur and Matt: a question in plain language becomes a test on 15 years of data and a short charted answer.
- Decision explainer: every engine signal shown with what it saw, the rule that allowed or blocked it, and the outcome.
- Track Record page and the failure library, starting with the 36 trials already in the ledger.
- Success gates written with Matt, dated, with a 12-month clock.
Done when: Gudmundur and Matt use Ask KAT every week and the Track Record page updates itself nightly.
Phase 2 · Learning paths and simulators
- Futures path first (the engine's home), then options (spreads, risk graphs) and stocks (trend and early growth).
- Simulator class mode: learner accounts, a task at the end of each lesson, a trade journal, the coach after every trade.
- Training-wheel accounts with the risk rules switched on.
- First cohort: Katherina and up to ten students, tested before they start.
Done when: the first cohort has finished the futures path and the coach has reviewed at least 300 practice trades.
Phase 3 · The daily companion
- Plain-language morning brief and evening review for learners.
- Daily simulator challenges.
- Matt's weekly live review of learners' trades.
- Cohort community and a strategy lab where learners test their own ideas through Ask KAT.
Done when: 60 % of the first cohort is still active after three months.
Phase 4 · Prove it
- Second cohort; 20+ learners in total, tested before and after.
- Engines run forward under frozen rules.
- A public page with the track record, the failures and a waiting list. No prices.
Done when: the success-gate report is written, pass or fail.
Decision
If the gates are passed and learners are demonstrably helped, we design the offer: a free tier that stays free and a fair price of around €25 a month, with the legal set-up for every country it is offered in. If not, it stays our own tool, and the result is published like every other test.
Task board, Phase 1
Every task has an owner, a due date and a definition of done. Nothing touches live trading without the owner's approval; new features ship behind a switch that is off by default.
| ID | Task | Owner | Due | Done when | Status |
|---|---|---|---|---|---|
| U-01 | Write the success gates with Matt (engines and teaching), dated | Gudmundur, Matt, Claude | 1 Nov 2026 | Committed document both owners have read and agreed | To do |
| U-02 | Ask KAT v1: question → pre-set test on 15-year data → charted answer page | Claude | 22 Nov 2026 | Five real questions answered end to end, each in under ten minutes | To do |
| U-03 | Failure library: each of the 36 trials as a two-minute lesson with its chart | Claude, Katherina | 29 Nov 2026 | Katherina understands every lesson without help | To do |
| U-04 | Decision explainer: every engine signal with its reasons and outcome | Claude | 6 Dec 2026 | A full trading day explained, every skip and every trade | To do |
| U-05 | Track Record page, refreshed nightly, failures visible | Claude | 13 Dec 2026 | Every engine forward after costs, expectation beside result | To do |
| U-06 | Pre- and post-test for learners (knowledge and simulator discipline) | Claude, Matt | 20 Dec 2026 | Katherina has taken the pre-test; scoring is automatic | To do |
| U-07 | Data-licence check: what may be shown to learners from each vendor | Gudmundur | 20 Dec 2026 | Written answer per vendor on file | To do |
Backlog for Phases 2 and 3: what beginners value most
The practices that keep people in the best-loved trading communities, built in our own form, each with a measured result.
| ID | Item | Phase | Done when |
|---|---|---|---|
| U-08 | "Today's plan" before the open: levels, bias, main risk and what would change the view, in plain language | 2 | Published every trading day for a month, scored nightly |
| U-09 | Trade-management narration for every simulated trade (stop moves, partials, exits) with the measured cost of each choice | 2 | A full day of trades narrated and understood by Katherina |
| U-10 | Prop-firm path: what an evaluation costs, the rules that fail most people, practising them in the simulator | 2 | A learner passes a simulated evaluation under real prop-firm rules |
| U-11 | A checklist for every engine strategy, showing how often each condition really mattered | 2 | Checklists for all live-eligible strategies |
| U-12 | Practice rooms by style: futures day trading, stocks, option swings, option selling | 2–3 | All four rooms open to the first cohort |
| U-13 | Learners' trade feed with lessons learned, ranked on discipline | 3 | Used weekly by most of the cohort |
| U-14 | Weekly research note: one Ask KAT answer published with its chart and the failures | 3 | Twelve weeks in a row |
| U-15 | The engine thinking aloud through the first hour, live on the site | 3 | Every trading day for a month without gaps |
| U-16 | Broker and platform guide for beginners, by country | 3 | Reviewed by Matt and two learners |
| U-17 | Measure first: privately score a room's breakout and flow alerts for 3 months against matched random picks (Signal Desk) | 1 | Private report: do the alerts beat a random pick? |
| U-18 | KAT breakout bot on our own data, with a plan card and context warnings | 2 | Runs every trading day for a month; every alert stored |
| U-19 | Alert scoreboard: every alert's outcome, refreshed nightly | 2 | 200+ scored alerts, failures visible |
| U-20 | Lesson, simulator link and weekly post-mortem for each alert | 2–3 | Katherina uses it for a month |
| U-21 | Flow and dark-pool alerts, only with a data licence and a positive U-17 | 3 | Owner decision with the U-17 numbers |
Success gates
Written down in advance and judged once, like every other KAT test. Moving a gate after the results is not allowed.
Teaching
- 20+ learners finish a learning path
- Knowledge test scores rise from before to after
- Simulator discipline improves: a stop on every trade, size within limits, fewer rule breaks
- 60 % still active after three months
- At least half say they would pay €20–30 a month
Each engine shown as proof
- 12+ months forward on paper with real broker fills: 300+ intraday trades or 250 trading days
- Then 6+ months live at the smallest size
- Positive after commissions and slippage, t ≥ 2, ahead of a pre-set benchmark
- Worst drawdown within twice the expectation; no day worse than −3 %
- Checked by someone outside the team. Any rule change restarts the clock.
What we measure every month
| Measure | Why | Target by September 2027 |
|---|---|---|
| Ask KAT questions answered | Is the core feature used? | Weekly use by every active learner |
| Knowledge test change | Do people actually learn? | Clear rise for most learners |
| Practice trades with a stop | Discipline, the first habit | Over 95 % by week 12 |
| Rule breaks per 100 practice trades | Safety | Falling month on month |
| Three-month retention | Is it valuable enough to stay? | 60 % or more |
| Engine forward results after costs | Honesty of the track record | Published nightly, whatever they show |
Risks and what we do about them
| Risk | Answer |
|---|---|
| No audience against a room with about 25,000 members and 300,000 followers | Build in public: the failure library and Ask KAT answers are shareable on their own. Count waiting-list names without paid ads. |
| Learners do not improve | Measured from the start; if the first cohort does not improve, change the paths before adding people, and publish the result. |
| No engine earns a record | The teaching value does not depend on one. Signals stay off until an engine passes its gate. |
| Rules for selling trading content | Rules follow each customer's country, wherever the company is registered. Settled with a lawyer before any offer in October 2027. |
| Data licences | Task U-07: show learners only what each vendor permits. |
| Time | Claude builds; owners review. Each phase has one definition of done and nothing else is added. |