Where things stand.

784 pre-solved flop situations: 16 pre-flop lines × 49 representative flops. Internal solver target: 0.8% of the pot or better, within the configured game tree. All library entries loaded and passed integrity checks; one flop was inspected. This target is not an external accuracy certification.

Reference-based questions, expected-value feedback and spaced review. Two questions were answered and graded; the practice page opened. Solo practice uses free play chips.

Bet maths, range equity, tournament chip-value calculations, push/fold equilibria and variance tools. Sample values were computed; the variance page appears in the recorded demo. Calculations depend on their assumptions.

Strategy and EV across verified flops, grouped by board texture. The page opened and listed library lines.

599 neutral demo hands were imported. A hand without a hero was refused, automatic folder import was disabled, and a recent hand was blocked by the 15-minute analysis cooldown. Only the user's own exported hands are accepted.

Raw, all-in-adjusted and AIVAT-adjusted win rates, with intervals and coverage counts. Run on the 599 imported demo hands. Adjustments do not remove all uncertainty.

Re-solves an assumed tendency to explore a hypothetical response. The page opened in the overview check; a full re-solve was not repeated. Earlier internal runs exist. Hypothetical results do not feed training or the library.

Estimates a lower bound on best-response gains against frequencies from the user's own past hands. The page opened; a full run was not possible with the demo's small coverage. Earlier internal runs exist.

Records the user's own entered or imported session results. Source reviewed; the setup page opened, but no session was recorded in the overview check. No bank, poker-site or payment connection; no funds are held or transferred.

Both generated from public demo hands. The checkup correctly returned insufficient data. Reports state their coverage and sample limits. Broader reference coverage remains a blocker; these formats are not available as a paid service.

281 claims as of 25 October 2026: V1 88, V2 102, V3 2, and 89 without a V-tier. Tiers describe evidence within stated conditions. Refuted claims stay on record.

Pre-registered predictions, held-out tests and wider situation coverage. An internal historical test on 20 October 2026 found an exact reference for about 1.4% of hands in one public research sample. That is a sample-specific coverage result, not a current or general accuracy rate.

Tournament ICM error detection is being developed. Internal course volumes 1–4 are typeset; volume 5 has an author draft. Typesetting is not publication, external review or public course availability.

Two-person report review, a tournament-preparation pack and a limited client pilot, subject to reviewed contracts and data agreements. Scope and dates are not commitments.

A research-led study page and AI-assisted explanations of solver-sourced numbers; no such explanation feature is publicly available. Windows hardware verification, a hosted web version and app distribution remain plans.