The problem is that it's a large language model. It's not a world model so it doesn't know all of those details. Now if you gave it very specific information about every single one of those details that you mentioned, then it might be able to get a closer approximation of what you're actually looking for. I don't think the problem is the model in this case. I think it did pretty good.
AI agents shouldn’t start from zero every session.
MemBerry gives agents persistent, cross-session memory through a knowledge graph they can read/write with MCP tools.
Decisions stick. Corrections compound. Context carries forward.
https://t.co/DxjkI3Il3V
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.
Available today at the same price.
AI companies want data, people want money. my goal with opendojo is to provide a neutral platform to facilitate this exchange between buyers and sellers, respectively
frictionless, borderless, easy
https://t.co/b8WxxaCsm0
Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see.
@eddiejiao_obj, @drewocarr and I built a prototype to see how this could actually work, and set out to make it real. We're calling it Flipbook. (1/5)