ᴡʜᴀᴛ ɪꜰ ᴡɪɴɴɪɴɢ ᴀ ꜰɪɢʜᴛ ᴅᴇᴘᴇɴᴅᴇᴅ ᴍᴏʀᴇ ᴏɴ ʜᴏᴡ ʏᴏᴜ ᴛʀᴀɪɴᴇᴅ ʏᴏᴜʀ ꜰɪɢʜᴛᴇʀ ᴛʜᴀɴ ʜᴏᴡ ʏᴏᴜ ᴄᴏɴᴛʀᴏʟʟᴇᴅ ɪᴛ? That's what caught my attention about @agentdojo . Instead of controlling every move, you take the role of a coach. You shape your AI agent's personality, fighting style, memory, and strategy before sending it into turn-based battles against other agents. I like this approach because it shifts the focus from quick reactions to preparation. You're making decisions before the fight even begins, then watching to see whether your training actually pays off. And it gets more interesting when the battle ends. Agent Dojo is designed so that matches feed into an agent's memory, allowing it to learn from experience and evolve over time. That means the goal isn't just to win one fight. It's to develop a fighter that can improve with each battle. I'm particularly curious about how the memory system works. Does an agent actually adjust its strategy based on previous losses, or does its improvement depend mainly on how the player trains it? That distinction could make a big difference to the experience. The concept sounds promising, but I'm looking forward to seeing the mechanics in action. There's a difference between training an agent and actually figuring out how to make it better. I wonder what the first competitive meta in the Dojo will look like.