@agupta: some ideas are much clearer when you can use coding agents to show a proof of concept. eg I hadn’t really understood ho…
Summary
A tweet highlights how coding agents can clarify complex ideas, using GPU vs NPU memory competition on devices as an example demonstrated through code.
View Cached Full Text
Cached at: 04/21/26, 09:12 AM
some ideas are much clearer when you can use coding agents to show a proof of concept. eg I hadn’t really understood how GPUs vs NPUs compete for memory on devices until this writeup (w/ its accompanying code)
Similar Articles
@levidiamode: 163/365 of GPU Programming Looking at a few different agentic GPU kernel optimization systems today. The two I'm most i…
A tweet discussing two agentic GPU kernel optimization systems: Auto GPU Kernel by @dogacel0 and Kernel Design Agents from @songhan_mit's lab, both winners at the MLSys Sparse Attention FlashInfer competition. The thread highlights different approaches using subagents and Claude skills for GPU programming.
@karpathy: Wow, this tweet went very viral! I wanted to share a possibly slightly improved version of the tweet in an "idea file".
Andrej Karpathy shares thoughts on sharing ideas rather than code in the era of LLM agents, proposing an 'idea file' format where the concept itself is more valuable than specific implementations.
@garrytan: My friend @finbarr says: It's like code as memory. You work with your agent in a non deterministic way to figure out ho…
Garry Tan discusses the concept of "code as memory" for AI agents, suggesting they generate executable scripts for new tasks and reuse them for efficiency.
@Thom_Wolf: watching a team of agents tackling a hard theoretical physics problem is quite mesmerizing - self-correcting, deriving …
A tweet observes AI agents collaboratively solving a difficult theoretical physics problem, demonstrating self-correction and equation derivation.
@yoheinakajima: and here's what a coding agent on activegraph looks like basically you always get a trace and a graph automatically
Yohei Nakajima shares a demo of a coding agent on activegraph that automatically produces a trace and graph.