Tim Burton

- Drafted a blog post - Used an LLM to meticulously improve the argument over 4 hours. - Wow, feeling great, it’s so convincing! - Fun idea let’s ask it to argue the opposite. - LLM demolishes the entire argument and convinces me that the opposite is in fact true. - lol The LLMs may elicit an opinion when asked but are extremely competent in arguing almost any direction. This is actually super useful as a tool for forming your own opinions, just make sure to ask different directions and be careful with the sycophancy.

@BigTechAlert @elonmusk @GoogleResearch @elonmusk following Google Research the same week TurboQuant drops is not a coincidence. More intelligence per bit is the whole game.

🆕 @elonmusk has started following @GoogleResearch

@tonybalogna The library-to-makerspace pipeline is something id genuinely love to see everywhere

Just found out our library has a $3000 3D printer. The things that are still possible in a low crime right wing town.

Be careful what tools you trust. A repo with 40k stars and 6.7k forks had obfuscated code that sends ALL of the secrets on your machine to a remote server.

is the future value of "open source" code anymore? i believe it's shifting to data, provenance, protocols, evals, and weights. in that order.

@sundeep an engineer’s value is measured in shipped product, not how fast they drain your token budget. this is Herbalife logic with a CUDA wrapper.

“If your $500K engineer isn’t burning at least $250K in tokens, something is wrong.”

"A good programmer is someone who always looks both ways before crossing a one-way street." — Doug Linder

@ClementDelangue Adversarial deflection seems inevitable

Our biggest open-source repos are getting overwhelmed by AI slop which literally makes Github unusable (~a new pull request every 3 minutes). Fun new challenges in an agentic world!

Introducing 𝑨𝒕𝒕𝒆𝒏𝒕𝒊𝒐𝒏 𝑹𝒆𝒔𝒊𝒅𝒖𝒂𝒍𝒔: Rethinking depth-wise aggregation. Residual connections have long relied on fixed, uniform accumulation. Inspired by the duality of time and depth, we introduce Attention Residuals, replacing standard depth-wise recurrence with learned, input-dependent attention over preceding layers. 🔹 Enables networks to selectively retrieve past representations, naturally mitigating dilution and hidden-state growth. 🔹 Introduces Block AttnRes, partitioning layers into compressed blocks to make cross-layer attention practical at scale. 🔹 Serves as an efficient drop-in replacement, demonstrating a 1.25x compute advantage with negligible (<2%) inference latency overhead. 🔹 Validated on the Kimi Linear architecture (48B total, 3B activated parameters), delivering consistent downstream performance gains. 🔗Full report: github.com/MoonshotAI/Att…

Note: Claude Code invalidates the KV cache for local models by prepending some IDs, making inference 90% slower. See how to fix it here: #fixing-90-slower-inference-in-claude-code" target="_blank" rel="nofollow noopener">unsloth.ai/docs/basics/cl…

@Polymarket this is a bad idea btw

BREAKING: New York bill would ban AI from answering questions related to medicine, law, dentistry, nursing, psychology, social work, engineering, & more.

> 385ms average tool selection. > 67 tools across 13 MCP servers. > 14.5GB memory footprint. > Zero network calls. LocalCowork is an AI agent that runs on a MacBook. Open source. 🧵

LLMs have reached nested branch prediction

@elonmusk @grok watch the titanic for me

Grok can watch videos for you

@tobi Very cool!

Turns out, you can vibecode your own motorsport telemetry tools instead of waiting for mac ports

You are a feature. Not a bug.

@cgtwts don’t undervalue the planning step

How it feels when Claude one-shots a task i’ve been planning for months and everything feels unreal

@big_duca edge inference times become the latency

Real question: What happens to the labs when models get so good. And you can run them locally on your computer or iPhone? With no latency. And no cost?

@ThePrimeagen creating a thing isn’t hard. maintaining is.

Convenience was the vulnerability. We knew. We shipped anyway. Context is the vulnerability. We know. We’re shipping anyway.