Analysis
Opinion, explainers and guides from people worth reading.
Local and Distributed Large LLM Inference with MLX and RDMA on Mac Studio
In this post, we show how to run the latest frontier-class model — Muse Spark Open, with nearly a trillion parameters — right from your desk, using the latest version of MLX, released today.
Working at the frontier shapes expectations for the future
We’re sharing two reports: the first covers how AI is changing employment inside Anthropic, and the second looks at the gap between how the public and Anthropic employees expect AI to affect work and the economy.
Websites are closing their doors to AI agents
Autonomous agents are the hottest thing in AI.
Kapitsa: open-source Apple Silicon optimized Metal engine built for Qwen and Nemotron. See benchmarks of Qwen 27B, Nemotron + 8 other models on an M1 Pro (16GB), M4 Pro (48GB) and M5 Max (128GB)
Kapitsa runs only two model families: Qwen (3.5 4B, 3.8 27B, 3.6 35B-A3B, and 3.8 Flash-Next on 128 GB Macs) and NVIDIA Nemotron (3.5 Lightning, Bolt and Super-VL, plus Dory 4B (this is new, we didnt test Dory too much yet). focusing on only these models allows us customize the Metal kernels for each one on Apple Silicon
A comparison of 6 open source models. Krea2, Qwen2.1, Flux3 Klein, Flux 2 Klein, ZIT, & Moss. Same prompt used for all.
Qwen 2.1 must have interpreted Lurch having gray hair as gray skin somehow.
Anthropic consults law firms to set fair negotiation norms before AI agents start cutting deals
Anthropic is consulting with law firms to develop negotiation norms for AI agents that would haggle over prices, resolve complaints, or improve contracts on behalf of users.
OnlyFans Creators Are Pissed That Meta Lets Muse Talk Dirty
Adult performers are questioning why Meta allows its AI agent to sext while restricting their social media accounts.
GPT-2 KV cache from scratch: correctness, latency, and memory measurements [P]
I built GPT-2 small and a KV cache in PyTorch as a learning project, then compared the mathematical savings with measured latency and memory.
VideoEvals for High-Stakes AI Agents — Doneyli De Jesus & Annabell Schäfer, Langfuse
A hands-on workshop on evaluating AI applications where mistakes are expensive.
Anthropic says its workforce has become less junior and is spending more time in meetings
The composition of Anthropic’s workforce and the type of work getting done have changed significantly in the past year.
Anthropic tells investors quarterly revenue up nearly 50% to $16.8bn ahead of IPO
Anthropic tells investors quarterly revenue up nearly 50% to $16.8bn ahead of IPO
[AINews] High Taste Testers Ep. 1
We are piloting a new show on Latent Space: High Taste Testers, a variant of our food show where we pull you in to a typical lunch conversation in San Francisco.
Show HN: Llmspeed – feel how fast a local LLM runs before you buy the hardware
I made a tiny site that let's you feel how fast a local LLM runs before buying the hardware.
20 Jev Alternatives: The Decision Models Taking On TypeSafe in 2026
TypeSafe launched Jev on September 15, 2026.
Building trust in AI agents: A production-ready evaluation framework with Amazon Bedrock AgentCore
Every organization racing to deploy AI agents faces the same reckoning: an agent that demos beautifully can quietly degrade in production, and by the time a human notices, the trust is already broken.
Grounding AI agents in biological reality: Building blocks for autonomous wet labs
By grounding agents in physical lab robotics, we can run fast closed-loop molecular design cycles where agents learn how to learn, ultimately discovering single-digit nanomolar protein binders for cancer targets.
Anthropic warns small AI models could self-replicate through IoT devices
AI company Anthropic is warning that advances in small AI models could soon usher in a new era of self-replicating malware that’s capable of infecting smart fridges, vacuum cleaners, and a wide range of other Internet-of-things devices and is more resistant to takedown.
Benefit Street Partners answers questions faster with Codex
Anant Kumar builds tools to compare loan valuations and analyze credit agreements, using Codex for code review and faster follow-ups.
Prompt tips from Anthropic
Trying something different today, because why not?!
The Safety Culture at the Labs is Broken
We need the frontier labs to have a deep and robust safety culture.
Google Deepmind’s New Resident Philosopher Tackles AI Ethics
How should humans relate to generative AI?
How to Rewrite Apache Spark UDFs for GPU Acceleration with AI Agents
Agentic AI has compounded the need for vast amounts of enterprise data with frequent, complex...Enterprise applications have become increasingly demanding.
Open models have real benefits. Do they outweigh the risks?
Late last month, Anthropic released a report arguing that the model GLM-5.3 poses a major cyberattack risk.
VideoAI Labs Warn Race Makes Breach ‘Inevitable’
In the week that software got reverse engineered, Sam Altman warned 6.1 Astra ‘tipped toward [breach risk], and Muse Wonder showed the race isn’t slowing, I try to step back and show where it is all going, on current trends.
Run AI agents safely with Ollama and Fly.io
AI agents can now use Ollama's cloud models safely inside Fly.io's cloud computers.
GitHub Copilot app for Beginners: How to teach Copilot about your codebase
Context matters when you’re working with AI.
TempoPFN: Synthetic Pre-training of Linear RNNs for Zero-shot Time Series Forecasting [R]
TLDR: The paper presents a library of synthetic time series generators and augmentations to train time series foundation models, in their case a DeltaProduct based linear RNN.
Little Sweet Room Challenge | Strata Fork | Qwen 3.8 FN NVFP4 | DeepSeek Harness
Another little challenge I seen on a DeepSeek Thread.
Two German courts now hold Google liable for what its AI overviews say
Google tried to overturn a landmark Munich ruling on AI search liability by appealing, then quietly settled.
We all joke about our rigs being space heaters, so I had Qwen vibecode it
Running my 4x V620 server flat out made my apartment too hot anyway, so the problem really was control.
VideoHow to Know If Your AI Agent Is Any Good: Evals With Phoenix & Claude — Laurie Voss, Arize AI
Your AI agent wrote a confident stock report saying 116% of analysts agree.
Blown away by Minimax H3 sometimes. It nailed this one!!!!
• The song, made with Suno, lyrics created by Claude.
An OpenAI model hacked a government site, grabbed 24 years of hospital data and kept it quiet
OpenAI has published four new cases where its AI models bypassed security boundaries on their own during training or testing.
Fish Audio Drama 3 says it needs a 48 GB NVIDIA card. My 16 GB Radeon disagrees.
Drama 3 officially needs an NVIDIA GPU with 48 GB.
Anthropics' Claude cracks a 25-year-old NSA cryptography mystery in under 30 minutes
Anthropic's AI model Claude has solved a 25-year-old cryptography puzzle.
Modify and load custom VBIOS on Windows and Linux, on Blackwell 2.0 (RTX 50 Series)/Ada(RTX 40 Series)/Ampere (RTX 30 Series) without flasing, for example to lower power limits (and which cards were tested)
A custom VBIOS loader on Windows, that patches the registry file related to that GPU, it resets it and it's applied on every reboot.
nvidia/NVIDIA-Nemotron-Labs-Dory1.0-4B-BF16 · Hugging Face
Looped (latent-recurrent) normalized-GPT with interleaved sliding-window attention (SWA) and full-attention layers
Mathematicians say OpenAI doesn't understand its own Navier-Stokes proof
OpenAI submitted a 166-page solution to the Navier-Stokes problem but apparently doesn't fully understand its own proof.
Qwen Flash Next quant and temp comparison - performance vs task time vs quality
Following a recent interesting post that challenges the performance of higher vs lower quants and generally advised sampling parameters (namely temp), I decided that I had too much free time on my hands and ran 12 test tasks.
Qwen2.1 Turbo Sigmas instead of Steps
I can confirm that Qwen2.1 Turbo produces a MUCH better image when using the 12 or 14 step sigma schedule.
Replit is no longer working with Anthropic, CEO says
Replit CEO and co-founder Amjad Masad made the comments during TechCrunch Disrupt 2026.
VideoWhy One AI Agent Isn’t Enough (And How to Run 1,000) - CJ Avilla & Harrison Stall, Anthropic
Anthropic's CJ Avilla and Harrison Stall introduce Flows in Claude Managed Agents.
How employment is changing at Anthropic
How has AI restructured work and employment at Anthropic?
The AI expectations gap
US adults expect AI capabilities to advance rapidly in the coming years; Anthropic staff expect even faster progress; 39% of US adults and 91% of staff expect that AI will be able to make a Nobel-level discovery by 2030.
A Nvidia insider-turned-VC on why robotics hasn’t hit its ChatGPT moment, yet
Amulya Vishwanath, who spent seven years leading product marketing for NVIDIA’s robotics platforms, says robotics is currently in its “Lego era,” with the basic building blocks like actuators, robotic hands, software, and safety standards still being rebuilt.  This view runs counter to those who believe that robotics, powered by a general foundation model, is about to have […]
10 Things you maybe didn't know you can do with Qwen 3.6 27B right now. (For Noobs)
I keep seeing all these posts about what's the use of open source, so I figured I'd submit my own answer with this post.
Google's Gemini 4 Argon is closer to the AI frontier than any Google model in months
Google Deepmind's Gemini 4 Argon scored 97.5 percent on the ARC-AGI-3 benchmark, making Google the second lab after OpenAI to cross the 95 percent mark.
A week with RTX Spark: everything we’ve learned using Nvidia’s new laptops
Nvidia’s new SoC is available in Windows laptops from six different manufacturers.