ComingUp
About
https://reddit.com/link/1u9uvf9/video/rp5u7cpwt68h1/player Been going down a rabbit hole on why AI assistants feel weirdly dumb after a few sessions. Turns out the problem isn't the LLM. It's that almost every AI product treats memory as a flat text buffer or skips it entirely. The real issue: RAG retrieves. It doesn't remember. "I prefer async communication" and "I had pizza for lunch" get stored with identical weight New facts don't override old ones they ju...
Comments (4)
so what makes the memory actually stick though
So the model was fine all along, just needed better memory.
spacing and type choices make this super easy to parse
context rot is the real villain here
Related Products
Typist
Transcribe audio to text in seconds.
rapym
TKCORE AI
All-in-one AI platform featuring TkCore-V5.5-Pro & top LLMs
Textsight.ai
TextSight.ai detects AI-like writing, gives a 0–100 Authenticity / Humanization
ai image prompts
Copy-paste AI image prompts that actually work.
ChatFlow
AI chatbot for your website — live in minutes