More trustworthy analysis and practice from your own games
A round-up of 1 October’s improvements: clearer mistake detection, personal workouts, opening studies, broader tactics controls and private access to your chess data. These notes include work introduced earlier today; the larger puzzle collection and external client connections still have separate rollout steps.
More trustworthy analysis
- Ordinary rook exchanges and legal queen recaptures are excluded from critical rook-loss warnings. Queen-fork warnings now check legal defensive replies.
- Delaying Nf3 and not castling by move 15 are opening observations, not automatic critical mistakes. The early queen-trap warning is restricted to the specific pattern it can support.
- Unsupported variants receive a clear analysis error instead of repeatedly retrying. Security dependencies and analysis reliability have also been maintained.
Work on your own mistakes
- Practise engine-detected blunders in short workouts, with queen-fork and rook-safety sessions when reviewed evidence from your games supports the theme.
- Recommendations link to their supporting game positions and ignore older, unreviewed detector flags. Unfinished sessions resume and existing practice history is retained.
- See answer accuracy, quick recognition and recurrence in later games, with sample sizes and analysis coverage. A small or missing later sample is not presented as proof of improvement.
Study the openings you play
- Opening summaries show your recorded openings by colour, whole-game results, analysis coverage and source games within the latest 200-game sample.
- Review repeated costly decisions at their exact moves. Choose a 10, 15 or 20-move study window without treating it as a universal end to the opening.
- Build private studies from game positions, save alternative lines and practise your chosen side. Add editable plan notes, legal move facts and opponent replies to examine; these are study prompts, not explanations of engine reasoning.
Tactics and clearer navigation
- Multi-move tactics offer automatic opponent replies, theme filters, minimum and maximum puzzle ratings, and new or due review filters. Saved puzzles and opening lines use spaced practice.
- A 480-puzzle Lichess collection is prepared across eight themes and three rating bands. It needs a separate import before the full collection is available; puzzle ratings are not playing ratings.
- Navigation separates game review, own-game drills, tactics and opening repertoire, with direct links between them, clearer empty states and improved mobile layouts.
Private connections and community updates
- Private read-only API and MCP support lets a trusted client request your verified accounts and stored analysis using expiring, revocable access. Gambit does not automatically upload games to an AI service. External connection checks remain to be completed.
- Versioned What’s new notes and a welcome update after sign-in make unread community releases easier to find.
Prepared for rollout
- Lichess account linking and resumable imports are built and merged. Ownership is verified through a profile challenge, and saved progress shows imported games, filtered records and resume controls. New Lichess connections remain disabled by default until rollout; no new login credentials are required.