Everyone talks about perfect prompts, but the real problem is memory - say hello to thredly!

I’ve noticed something strange when working with AI like ChatGPT and Gemini. You can craft the most elegant prompt in the world, but once the conversation runs long, the model quietly forgets what was said earlier. It starts bluffing, filling gaps with confidence, like someone trying to recall a story they only half remember.

That made me rethink what prompt engineering even is. Maybe it’s not just about how you start a conversation, but how you keep it coherent once the context window starts collapsing.

I began testing ways to summarise old messages mid-conversation, compressing them just enough to preserve meaning. When I fed those summaries back in, the model continued as if it had never forgotten a thing.

That experiment eventually became thredly, a small tool I launched here recently that automates that process, turning long chats into structured memory hand-offs you can reuse anytime.

It turns out, memory might be the most underrated part of prompt design. The best prompt isn’t always the one that gets the smartest answer, it’s the one that helps the AI remember what it’s already learned.

Has anyone else tried building their own memory systems or prompt loops to maintain long-term context?