This is actually a huge change, and we're not just using hyperbole. We fundamentally reworked how the chat works so that it actually... works. This might get a little technical, so bear with us. Let's break this down.

How did the chat work before?

AI models are notoriously bad a chess, and we want to solve this problem. Before, we would provide the AI model with information from Stockfish, and the model would use that information to answer the user's question. The problem with this, was that the AI models were still not able to conceptualize a position, tactics, moves, etc. When using the chat, it would give you some general chess advice and could tell you the best lines, but it wasn't very good at understanding the position. It could tell you what the best moves were, but not why, and if we ever asked about hypothetical moves, it wouldn't have that information and begin to hallucinate.

How does the chat work now?

Instead of dumping a bunch of chess engine data into the AI and hoping for the best, the chat now works more like a conversation between the AI and the chess engine. When you ask a question, the AI can actively query Stockfish on its own to get the specific information it needs to answer you.

Think of it like this: before, we handed the AI a textbook and said "answer questions from this." Now, the AI has a chess grandmaster sitting next to it that it can ask questions to in real-time.

The key difference: tools

The AI model now has access to a set of chess-specific tools it can use while thinking through your question. It can:

• Analyze the current position to understand what's going on and what the best moves are
• Evaluate a specific move to tell you whether that knight sacrifice is brilliant or a blunder
• Play out hypothetical lines so when you ask "what if I play Bxf7+?", it can actually look at what happens next instead of guessing
• Check which moves are legal so it never suggests an impossible move
• Look at earlier moves in your game to review where things went right or wrong

The AI decides which of these tools to use based on your question. Ask "what should I play?" and it'll grab a quick position analysis. Ask "was my 15th move a mistake?" and it'll go back in the game, pull up that position, and run a deep comparison.

Context-aware responses

The chat also adapts to what you're doing. Solving a puzzle? The AI acts more like a coach, nudging you toward the answer with hints rather than spoiling the solution. Reviewing a game? It gives you deeper analysis and points out critical moments. Playing a live game? It keeps things short and actionable.

The result

The chat now actually understands chess positions. It can reason about tactics, explain why a move is good or bad, explore "what if" scenarios, and give you advice that's grounded in real engine analysis rather than pattern-matched guesses. It doesn't hallucinate moves, it doesn't give vague advice, and it adapts to your specific situation.

We genuinely can't believe we shipped the old version for as long as we did. This is what the chat should have been from the start, and we're so excited to finally start using it ourselves!

What's next?

We need to test the new chat functionality with you, our users. We really want to know how you use the chat, and what we can do to improve it! If you find any issues, hallucinations, or just have any feedback, please let us know by emailing us at sebastian@soleinnovations.com.