LLMs have a “lost in the middle” problem – they focus on the start and end of documents but miss key info in between. (Adam Zewe, MIT News)
Get Technical: Deep Dives
What makes workflows different from agents? A good introduction and explanation from Anthropic, and a case for keeping things simple.
This research paper argues against “reasoning/thinking” hype: intermediate tokens often lack substance, despite appearances.
Video: AI prompt engineering deep dive (Anthropic)
Seamless MCP-powered integrations sound appealing, but raise performance and security concerns. (Shrivu’s Substack)
Deep-dive into RAG and evaluation: How Süddeutsche built their election chatbot. (Medium)
Anthropic looks under the hood of Claude 3.5 Haiku, using circuit tracing to see how it works across different kinds of tasks.
Teach AI your face in a few clicks: A beginner’s guide to finetuning FLUX
Resources (from the author of AI Engineering) (Chip Huyen, GitHub)
Building LLMs is probably not going be a brilliant business (Cal Peterson, calpaterson.com)
OK, I can partly explain the LLM chess weirdness now (dynomight, dynomight)
Are you curious about AI agents and how they work, but don’t know how to START? (Andreas Horn, LinkedIn)
You Exist In The Long Context (Steven Johnson, thelongcontext.com)
How a stubborn computer scientist accidentally launched the deep learning boom (Timothy B. Lee, Ars)
