I feel like this will also put a lot of gpt voice models into more dominance.
20 karma · joined February 1, 2019
Start with an intuition check on every move Ask these three questions before picking candidate moves:
What are the weaknesses on the board? Look for targets.
What is the worst placed piece? Improve it or activate it.
What is my opponent intending? Prophylaxis saves games.
This helps keeps your focus on the right areas so tactical ideas pop naturally.
A system for every move
Forcing moves first: List checks, then captures, then threats. Calculate the forcing lines first.
Loose and overloaded pieces: Count attackers and defenders. Undefended or singly defended pieces likely can fall to tactics.
Files and Ranks: Scan files, ranks, and diagonals for piece alignments that create pins, skewers, and x rays.
Discovered possibilities: Ask what becomes uncovered if a piece moves. If the uncovered line gives check or capture values, you may have a discovered attack or double attack.
Using sites like Lichess and ChessTempo you can find the common puzzles / themes in games. Using the Chess Coach with the above system and check will help you spot tactics in your games.
In particular, surfacing tactical concepts like pins, skewers, discovered attacks and forks in games proved to be tricky.
I have been testing out a few implementations and have some decent results for a first pass but also many learnings.
For example, to make a move that forks two others pieces is more beneficial (from a material gain perspective) when those pieces are unprotected or not protected well.
Yet in chess.com definitions page of a fork I see "A fork is a basic chess tactic that consists of a single piece attacking two or more pieces at the same time. The attacking piece is known as the forking piece, while the attacked troops are known as the forked pieces."
This definition is great but does not mention any concept of protection vs un-protection of pieces.
I mention this because to program rules around when a fork / tactics should be triggered was an interesting process.
See tactics being detected - https://imgur.com/a/detecting-tactics-bRdsY7u
You see images of forks, pins and discovered attacks from my own games.
If you want to try this out on your games, you are welcome to https://app.chesscoach.dev/
The LLM teaches shows where you went wrong
Demo: https://app.chesscoach.dev/#Analysis
Landing page to follow along the build: https://chesscoach.dev/
It works for games on lichess.org and chess.com
Still a work in progress, and welcome any feature suggestions.
I have been enjoying building it in my spare time
It helps players improve using AI
I like your website design, very clean!
Regarding community, I have a following here: https://lichess.org/team/grandmaster-ai-agent
High on my to do list to make this for chess.com players as well
Do you use Lichess?
I would prefer competing with chess.com :)
When you are attacking, remember to use all your pieces. Your Knight comes to open squares. Your Bishop is great at targeting pawns. Rooks and your Queen start the checkmate.
Don't panic, when it gets complex, look for exchanges. Most likely your opponent will play naturally looking moves, and your attack can be finished with tactics.
Yes you are right. It is still in demo phase, it still does make mistakes. I am refining the model and inputs, so definitely a work in progress :)
Regarding the circle highlighting, the agent is deciding / reasoning which square to highlight. So it is non-determinsitic, so it is sometimes right / wrong.
It will definitely get better as the models improve
That’s what I’m trying to fix. Instead of just showing lines, my AI coach gives voice-guided feedback, visual highlights, and practical insights. More like working with a real coach than sifting through raw engine output.
The goal is to make analysis as engaging as playing—and shift the mindset from “just tell me the best move” to “help me think better.”
Demo if curious: https://www.loom.com/share/9e1578f1348841c1992c5d902e371312?...
Unlike typical chess engines, the Grandmaster AI Agent provides:
Critical Moment Detection: Spots tactical errors with natural voice explanations.
Visual Learning: Highlights critical squares and patterns on the board.
Actionable Feedback: Offers principle-based advice to improve your chess understanding.
Early beta testers have already seen significant improvement:
"I love the engine’s explanations—it feels like learning from a GM!"
"I went from 400 to around 1700 Elo, wishing constantly for this level of explanation."
"Absolutely adore the AI Agent."
Would you trust an LLM to replace your chess coach?
Addressing the significant gap between AI's promise and reality, and exploring strategies to overcome it.
Emphasizing the importance of integrating LLMs, RAG, and AI guardrails to build effective AI solutions, and why DIY methods may fall short.
Streamlining building with Writer Framework to achieve enterprise-grade quality.
In the video linked, I demonstrate how to win a chess game by using the Légal Opening Trap.
If you would like to follow along try the interactive lesson: https://lichess.org/study/XZX6m7R7
Thank you very much.
Here is a short video explaining the strategies that I used in a chess game to win against a Grandmaster.