As expected, that was popular. Here is my attempt at consolidating all the answers into a list. - Prenorm: normalization in the residual blocks before the attention operation and the FFN respectively - GQA (Group Query Attention): more Q than (K, V)

Useful summary of transformer architecture improvements via François Fleuret

DSPy's biggest strength is also the reason it can admittedly be hard to wrap your head around it. It's basically say: LLMs & their methods will continue to improve but not equally in every axis, so: - What's the smallest set of fundamental abstractions that allow you to build

DSPy has been, and continues to be, ahead of its time

Looks promising - o3-pro by OpenAI

Time to catch up DSPy

Informative summary of the DSPy philosophy by LLM programmer-in-chief Omar Khattab

Stop counting “multi-agent” stages—start thinking like software engineers. @dspyoss shows how to treat LLM calls as first-class functions, letting you design any control flow (recursion, exceptions, etc.) and then optimize the entire program. Compound AI Systems are here via

Hot take, cold truth by Yuchen Jin

Anthropic’s new multi-agent Research feature shows why a single LLM, no matter how capable, can’t tackle every knowledge-dense question alone. A lead Claude Opus 4 agent now spawns parallel Sonnet 4 sub-agents that fan out across web, internal repos, and other tools, then

We've launched Claude for Financial Services. Claude now integrates with leading data platforms and industry providers for real-time access to comprehensive financial information, verified across internal and industry sources.

1/N I’m excited to share that our latest OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world’s most prestigious math competition—the International Math Olympiad (IMO).

Comprehensive 822-page book on #MachineLearning with #Python for #AlgorithmicTrading (➕useful in many other business & industry #TimeSeries applications): amzn.to/3aCyVwY by Stefan Jansen —— #IoT #IIoT #Edge #AI #DataScience #EdgeComputing #Industry40 #EdgeAI #Forecasting

3rd edition forthcoming! Stefan Jansen

Indeed.

Less than 89 hours ago, Claude Code unlocked sub agents. Minds are blown. And people are already building with their own agentic AI dev team. 10 wild examples:

This is quite useful.