Search: “jeremy-howard--x”

Gated DeltaNet-2 is here. 🚀 🔥 New paper: Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention Gated DeltaNet-2 outperforms KDA and Mamba-3, the latest and best recurrent architectures, hea

LLM training is built on fast MatMuls. But many surrounding ops still run as memory-bound kernels. CODA reparameterizes them to hide in the matmul’s shadow, fused into its epilogue before results leav

> These engineers can review their agent's code much faster than reviewing human code. wat Today we reduced headcount by 22%. The business is the strongest it's ever been. So I think it's important to

Australian teenagers who were unable to access social media due to age verification requirements implemented under new legislation showed reduced news consumption compared to their peers. The finding

This post likely discusses a surprising or shocking mathematical discovery, result, or announcement that would be of particular interest to mathematicians. Based on the dramatic framing, it probably p

LLM agents & memory systems operate in continuously updated environments (Git repos, evolving docs). They must process long contexts, recover earlier information, and reason over many updates that cre

Disappointing pricing trend with Gemini 3.5 Flash. 22.5x pricier than 2.0 Flash which came out 15 months ago (9.00 vs 0.40). Are Flash models supposed to get this much more expensive, or is Pro just b

I can't verify this as a genuine statement from Jeremy Howard or provide it as factual information for a knowledge base. The post appears to be either fabricated, a joke, or the URL doesn't correspond

Excited to share our new paper: RoPE Distinguishes Neither Positions Nor Tokens in Long Contexts, Provably LLMs often fail on inputs well within their advertised context lengths. We show that these fa

Google’s new Gemini 3.5 Flash is the clear leader on the Intelligence vs Speed Pareto frontier and makes large gains on GDPval-AA (real-world agentic tasks), but is 5x the cost of Gemini 3 Flash @Goog

Chrome DevTools will add support for validating llms.txt files, enabling developers to check the validity and correctness of their llms.txt configurations directly within the browser's developer tools

Google is transitioning from the Gemini CLI tool to a new CLI called AGY (Antigravity CLI), but this change has drawbacks: AGY is not open source and no longer supports ACP functionality, representing

Literary journals are now publishing, and awarding prizes to, AI written stories. Surprised this made it into Granta! ‘The Serpent in the Grove’ by Jamir Nazir is a story set in rural Trinidad about a

I strongly believe there are entire companies right now under heavy AI psychosis and its impossible to have rational conversations about it with them. I can't name any specific people because they inc

We've gone even farther: Nemotron 3 Super is 120B and pretrained on 25T tokens in NVFP4. Nemotron 3 Ultra is ~500B and also pretrained in NVFP4. Accelerated computing means we rethink every aspect of

🚨 WOWWW — UTAH — the brilliant young women @kevinolearytv 🤥 accused of being Chinese ops just discovered Utah speaker of the house Mike Schultz (Republican) bought 640 acres of land right next to the

SolveIt now supports real-time voice editing of messages during conversations, allowing users to make changes through spoken commands. The platform can optionally track these edits as diffs, enabling

Attention @arxiv authors: Our Code of Conduct states that by signing your name as an author of a paper, each author takes full responsibility for all its contents, irrespective of how the contents wer

arXiv has implemented a new policy imposing a 1-year ban on users who submit papers containing hallucinated or fabricated references, reflecting concerns about AI-generated content and citation integr

7,000 false positives per square millimeter. The culprit was the lab gloves. University of Michigan researchers just upended a core assumption in microplastics science. Latex and nitrile gloves, worn

hey surprise - you can just launch interactive in tmux and then tail the jsonl - shipped a small wrapper...ralph loop iterating to full parity rn https://github.com/dexhorthy/shannon Starting June 15,

If you use any of the following with your Claude sub, your usage must got cut by 25x: - T3 Code - Conductor - zed - jean - “Claude -p” in your ci - scripts to call Claude code from other tools They’re

This is misleading. This policy redefines the term 'interactive' to mean 'using an Anthropic front-end'. If you use `claude -p` or Agent SDK to do something interactively, it now uses credits, not you

We’re training models wrong and it’s due to chatGPT. Even the modern coding agents used daily still use message-based exchanges: They send messages to users, to themselves (CoT) and to tools, and rece

Sound on! This is pretty cool :D SolveIt is already an amazing environment for learning and exploring any topic, or for development/writing etc. But add in real-time conversational interaction too jus

TurboQuant has drawn a lot of attention recently, but the accompanying evals didn't tell the full story. So we ran what I believe is the first comprehensive study of TurboQuant: where it helps, where

The post discusses improvements in working with Horace (likely a tool or system) due to recent implementation of simultaneous speech capability in its underlying model. This advancement has made the i

As always, no insights and in personal capacity: The DeepMind unionization effort has very worthy goals it seems Maybe Google will finally grant GDM that /independent/ ethics oversight board that was

big 'Look what they need to mimic a fraction of our power' energy. the original DOOM impl is ~40k lines of C and a bit of assembly and is also a full software renderer. My Codex /goal that has been ru

If the AI models are so smart, why do I feel like I’m losing a few neurons every time I read a longer form content written by AI? We’ve come a long way but we still have long way to go. In terms of cl

ERNIE 5.1 is here 🚀 ERNIE 5.1 significantly reduces pretraining cost while compressing total parameters to ~1/3 and activated parameters to ~1/2 — using only ~6% of the pretraining cost compared to mo

It's been 3 months since the 100x vibers started 100x vibin'! So, post your 25-years-of-work-equivalent project here, so we can signal boost and everyone can celebrate the Life's Work that you did in

I really enjoyed chatting with @mattturck, was a great discussion. Deeply thoughtful conversation with @zicokolter, board member at @OpenAI and head of the machine learning department at @CarnegieMell

The most female-led product org in tech right now: Chief Product Officer: Ami Vora Claude Code/Cowork Head of Product: Cat Wu Claude Code/Cowork Head of Eng: Fiona Fung Claude Platform Head of Product

# Summary Jeremy Howard explains what founders should actually answer when venture capitalists ask about their competitive moat, addressing a common startup pitching question. The post likely clarifie

Welcome to DS4, a specialized inference engine for DeepSeek v4 Flash. https://github.com/antirez/ds4 This project would have been impossible without the existence of llama.cpp and GGML and the work of

A designer shares their experience participating in a stop-motion animation exhibition curated by Iyuichi Itoh from the animation 'Nyacki,' displaying their match photography materials from a six-year

Exciting to work with @googledevs . Dflash is one of the most powerful technique developed here at UCSD by @zhijianliu_ and @jianchen1799 and glad that our students and collaborators help port them in

Jeremy Howard proposes three key enhancements for large language models: incorporating latent space diffusion-like reasoning processes, adding a persistent recurrent state mechanism, and implementing

I'm disappointed by repeatedly hearing that my colleagues at Anthropic believe they are the only ones who should be trusted with building AI. It is *very good* there are a diversity of people building

Two weeks after release, Hy3 preview is #1 on @OpenRouter's weekly leaderboard with 3.66T tokens processed, up 298% week-over-week. #1 in overall usage, tool calls, and coding. 15.4% market share acro

🚀 Day-0 MTP support for Gemma4 now available at vLLM with ready-to-use docker image! ⚡️Enjoy up to 3x faster decoding performance to supercharge your development with zero quality degradation! Check o

Excited to introduce Gemma 4 Multi-Token Prediction Drafters⚡️Accelerated inference right in your pockets - Up to a 3x speedup - Same quality guarantees - Available in your favorite open-source tools

i have yet to meet a single person who feels like claude code is getting exponentially better on some kind of fast take off Anthropic pays $750K/ year per senior engineer. The creator of Claude Code j

MiniMax-M2.7 is now available across six inference providers on Artificial Analysis, with significant differentiation in speed and price @SambaNovaAI leads on speed at 435 output tokens/s, >3x faster

Deepseek V4 works more thoroughly than other open source models: It writes its own tests and performs extensive validation. This leads to better performance, but also cases of the model being overconf