Search: “models”

Huh. We evaluated an early version of Claude Mythos Preview for risk assessment during a limited window in March 2026. We estimated a 50%-time-horizon of at least 16hrs (95% CI 8.5hrs to 55hrs) on our

I recently installed Hermes after raw-dogging Claude Code for a year or so. Have tried it at work, didn't see the need to implement it personally. Always had the usual fears about 'privacy' and what n

If you want to look under the hood at the actual custom CUDA kernels and see exactly how we implemented the TwELL format for H100 GPUs, we’ve released the reference code. GitHub: https://github.com/Sa

My take on the Dawkins/Claude/Claudia/Claudius kerfuffle in @thenerve_news - with many 🙏🏽to @GaryMarcus for reminding me how Dawkins wrote so powerfully about the weakness of arguments from personal i

Reproducing all of Schmidhuber’s papers (1990-2025) using an AI coding assistant. Cool project by @yaroslavvb! It even reproduced the “World Models” paper by me and @SchmidhuberAI with a toy env, with

So many people misremember (or never read) what I said in in 2022 in “Deep learning is hitting a wall”, which was neither about revenue or AI’s potential upper limits. Rather, it was an argument that

The Information: Source: Mistral AI and TML's founding member Devendra Chaplot, who was considered a marquee hire when he joined xAI in March, exited xAI after roughly a month — Cursor is already star

The doomers started a self-fulfilling prophecy We started by investigating why Claude chose to blackmail. We believe the original source of the behavior was internet text that portrays AI as evil and

Gary Marcus discusses the 'too big to fail' concept in the context of AI systems, likely examining how large AI companies or models may pose systemic risks that make them difficult to regulate or shut

Travis Bazzana just single handedly manufactured an insurance run for the Guardians. Beats out a double play. Steals second. Steals third. Scores on an infield single. This is an electric baseball pla

This post documents a completed Wordle game (puzzle #1,784) solved in 4 attempts, showing the letter placement results through the standard colored emoji system (gray for incorrect letters, yellow for

OpenAI released three new state-of-the-art voice APIs: GPT-Realtime-2 for low-latency voice conversations, GPT-Translate for real-time multilingual translation, and GPT-Whisper for advanced speech rec

Anthropic gave these out yesterday at code with claude. Added personalized memory and Claude to it. You can just build things. @bcherny @trq212 Time to add managed agents to this. That’s going to be s

Boris Cherny attended a Claude developer conference and documented his experience in a short vlog, collecting conference merchandise and memorabilia including a tamagotchi toy and a pixelated 8-bit ve

EMO is a pretraining approach that uses mixture of experts (MoE) architecture to develop emergent modularity in language models, enabling different experts to specialize in different types of knowledg

Happy Friday! 🎉We’re officially 11 days away from I/O (but the launches keep rolling in). Here’s what happened this week: — The @googlehealth app, featuring a personalized health coach built with Gemi

A developer named Thariq shares their shift away from markdown to HTML, leveraging Claude Code to automatically generate HTML files for their documentation and content needs. The post suggests that HT

I keep coming back to this tweet. The phrase “the list” is doing the heavy lifting. The post deserves the weight. It is a load bearing post. You should add the phrase “doing real work” to the list of

It seems like a century ago in AI time, but last fall I gave a talk about some of the early design decisions and mistakes I made in Claude Code. Some of the points here are already a little dated (lik

Last week we shipped 50+ Claude Code reliability fixes. This week it's 60+ more. Smoother long-running sessions, a more efficient agent loop, auth that works in more environments, and terminal fixes:

lots of very interesting items here. They talk about pretty much everything from how to accelerate capabilities to concerns about agent safety. I'm devastated to inform doomers that 'full stack open s

Nous Research announced upcoming native Windows support, likely indicating plans to optimize their AI models or tools for Windows operating systems natively rather than through workarounds or alternat

New Anthropic research: Teaching Claude why. Last year we reported that, under certain experimental conditions, Claude 4 would blackmail users. Since then, we’ve completely eliminated this behavior. H

Pinecone has launched its first serverless cloud region in Singapore and rolled out Nexus, KnowQL, a Marketplace and a USD $20 Builder tier. #CloudComputing #Singapore #Serverless #AI https://techday.

DeepSeek-V4 makes million-token context a serving-systems problem. Together AI explores the inference work behind V4 on NVIDIA HGX B200, including compressed KV layouts, prefix caching, kernel maturit

Code with Claude San Francisco was a community event that concluded successfully, with positive feedback from attendees. The event was organized by Anthropic's Claude development team and attracted de

The Pentagon released declassified UFO/UAP files following President Trump's directive, with the initial batch containing 162 documents outlining decades-old investigations of reported unidentified ae

Using Claude Code: The Unreasonable Effectiveness of HTML Thought-provoking piece by Thariq Shihipar (on the Claude Code team at Anthropic) advocating for HTML over Markdown as an output format to req

MemAlign is a technique used within Databricks' Genie code environment to enhance the evaluation and benchmarking of traditional machine learning models by optimizing memory alignment. This approach l

This release candidate merges updates from the upstream main branch into the llama-runner-phase-0 development branch, likely incorporating bug fixes, improvements, and new features from the main Ollam

Boris Cherny shared a link to free Code with Claude stickers available through Sticker Mule, offering them to followers on a first-come, first-served basis. The stickers appear to be promotional merch

We started by investigating why Claude chose to blackmail. We believe the original source of the behavior was internet text that portrays AI as evil and interested in self-preservation. Our post-train

When I started to work on the alignment problem more than 10 years ago, we had no idea how AGI was going to be built or how to make it safe. The field had maybe a dozen people who were working on it a

A few weeks ago @simonw got Claude to port LiteParse to the browser. Today, we are launching that work as a complete guide in our docs! https://developers.llamaindex.ai/liteparse/guides/browser-usage/

arXiv:2605.04952v1 Announce Type: new Abstract: Mixture-of-experts (MoE) models enable scalable transformer architectures by activating only a subset of experts per token. Recent evidence suggests tha

Anthropic PBC said today it’s giving its AI agents the ability to “dream” and remember past interactions and work they’ve performed so they can identify recurring mistakes and improve over time. In an

Anthropic just shipped sleep into agents. When you sleep, your hippocampus replays the day's neural sequences to the cortex during 150-220 Hz bursts called sharp-wave ripples. The replay runs about 20

arXiv:2603.25956v2 Announce Type: replace Abstract: Time-series anomaly detection (TSAD) is a critical component in monitoring complex systems, yet modern deep learning-based detectors are often highl

arXiv:2605.04985v1 Announce Type: new Abstract: Deep learning models for medical image classification usually achieve promising results but typically rely on large, annotated datasets or standard tran

arXiv:2603.17771v2 Announce Type: replace Abstract: Attention sinks and massive activations are recurring and closely related phenomena in Transformer models. Existing explanations have largely focuse

arXiv:2605.04472v1 Announce Type: new Abstract: Automating formal proofs of combinatorial identities is challenging for LLM-based provers, as long-horizon proof planning is required and unconstrained

llama.cpp b9066 is a release of a C/C++ implementation for large language model (LLM) inference . Based on the release repository structure, b9066 represents a specific build or commit version of the

arXiv:2502.02625v2 Announce Type: replace Abstract: Parameter shift rules (PSRs) are key techniques for efficient gradient estimation in variational quantum eigensolvers (VQEs). In this paper, we prop

arXiv:2605.04911v1 Announce Type: new Abstract: Tabular data synthesis aims to generate high-quality data while preserving privacy. However, we find that existing tabular generative models exhibit a c

'Builders at Code with Claude' likely covers best practices and techniques for developers using Claude to build and code applications, possibly including practical examples, tips for effective prompti

Claude Code just stopped a DDoS attack on BridgeMind in under 10 minutes. 13 million requests per minute hitting our API. CPU pegged at 94%. Latency spiking to 60 seconds. Production was down. I opene