🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

I provide this service for my wife. I don't think she likes it very much...

Costco's worth $500 Trillion? This is a great example of why retrieval quality needs to be rock solid before working on generative AI. LLMs still treat any passed in context as gospel, so your RAG pipeline needs to be super precise.

Bad AI products come from teams who don’t use AI. Good ones come from teams who feel the friction, fight through the flaws, and keep building. You don’t design this stuff. You uncover it by living with it. By using AI daily and figuring out how to make it work right.

Every user interaction needs to make your product better.

Ethan Mollick LLM memory is still a blunt instrument. The real challenge isn’t storing more. It’s knowing what to use when. Same problem as RAG: context selection. Right now, models treat all context, memory or fresh input, as gospel. That’s gotta change.

Another day, another paper shows LLMs seriously struggle at not saying "I don't know" if there's insufficient context (aka abstention). Instead it just hallucinates. And reasoning LLMs are even worse. Better retrieval (RAG) is still the best way to get better quality.

I really like the term “context engineering” over prompt engineering. It describes the core skill better: the art of providing all the context for the task to be plausibly solvable by the LLM.

tobi lutke Hiten Shah This is why RAG isn't dead. Selecting the right context (retrieval) is really the key difference in output quality. LLMs still massively struggle with bad or too much context. Humans still need to provide the right context!