Search: “hardware”

We started running monthly research sessions at the @agi_inc office. First one was last Friday on PNM compilation. If you do on-device research and want to come present, apply here: https://form.typef

Hello again, everyone! Our latest Qwopus3.6-35B-A3B-v1 is now live, and it is once again breathtaking! Full HF space benchmark showcase and write-up is in the comments, so you can make conclusions for

SpaceX's Super Heavy Booster 19 was transported to the launch site for testing purposes, as documented in a post by Elon Musk on X. This represents progress in SpaceX's development of the Starship lau

Reachy-Mini is a compact robotic arm under development or assembly, likely a smaller version of the Reachy robot platform. The post from Hugging Face's leadership suggests this is part of their roboti

A user reported experiencing a false positive from their vehicle's driver monitoring system, which incorrectly detected their eyes as closed when they were actually open, prompting a humorous reaction

These folks are trying to ban open source. They're looking to take away your freedom to choose. They're also looking to take away the rights of businesses like Cursor to fine tune and make their produ

🚀 Introducing FlashQLA: high-performance linear attention kernels built on TileLang. ⚡ 2–3× forward speedup. 2× backward speedup. 💻 Purpose-built for agentic AI on your personal devices. 💡Key insights

Our cofounder @the_bunny_chen joined @GregorVand to talk about open models, production inference, and how RFT is unlocking model customization for teams without a dedicated ML org. They covered custom

This resource provides instructions for setting up and running a local coding agent using LM Studio's serving engine, featuring the Pi agent framework and Google's Gemma 4 26B language model. It demon

Nemotron-3 is an open-source language model available through Ollama's model library, likely representing NVIDIA's Nemotron-3 family of instruction-tuned models designed for various NLP tasks. The mod

The $15B Physical AI Company: 4 stack rewrites, end-to-end RL, neural sim, android for vehicles, trucks in Japan https://latent.space/p/appliedintuition @AppliedInt co-founders @qasar (CEO) and Peter

China is blocking the Manus acquisition by Meta. A few things stand out: - China has historically been more opaque and less public in restricting foreign acquisitions of Chinese tech assets, while the

Congrats to @deepseek_ai team! Doing the numbers I would estimate: Pro < 14m for the final training run Flash < 4m Ratio of active params x total training tokens vs v3 Total compute costs (data prep,

Elon's team just did something nobody's talking about. They replaced Starlink's call center with an AI. 🤯 And the results are insane. 1 in 5 people who called Starlink bought Starlink on the call. 70%

Qwen3.6-27B-TQ3_4S is a quantized 27 billion parameter language model that fits on 16GB of VRAM while supporting a 32k token context window, demonstrating strong performance across tested prompts. The

> be Yann LeCun > spend years building JEPA at Meta > company focuses on LLaMA instead > his idea stays complicated and unused > robotics plans get dropped > decides to leave and start AMI Labs > buil

We're open-sourcing FlashKDA — our high-performance CUTLASS-based implementation of Kimi Delta Attention kernels. Achieves 1.72×–2.22× prefill speedup over the flash-linear-attention baseline on H20,

A sophomore engineering student built a platform, landed corporate sponsors, and started a company. In three weeks. 🚀🤯 Engineering students have no way to build in public and get noticed by companies.

Okay this one is insane. A new 18B frankenstein model was just released on @huggingface — Beats the new Qwen3.6-35B-A3B on a 44-test suite despite requiring 12GB VRAM instead of 24GB 🤯 Runs on a SINGL

Nous Research demonstrated strong performance results using their Hermes agent with Qwen 3.5 35B model on an NVIDIA RTX 4090 GPU, highlighting competitive capabilities of locally-run models compared t

Hi! I'm here with *another launch*, it just happens to be extremely niche, nerdy, and probably only for a handful of people. In the desktop app, Claude Cowork and Code now have a little Bluetooth API

Qwen3.6 is a model available through Ollama's model library, representing Alibaba's Qwen series of large language models optimized for local deployment. The model can be accessed and run through the O

🎬 Ollama Gemma Day Recap: SGLang at the Ollama Gemma 4 Party in Palo Alto 🍾 Last night, @ollama hosted a packed Gemma Day at the Palo Alto office alongside the @GoogleDeepMind Gemma team. SGLang was i

📽️ IC LoRAs: Control Your Video 🎤 Hosts: Purz & Julien ⏲️ April 15th – 3pm PDT / 6pm EST 📍 Live on YouTube, X, and Twitch LTX 2.3 + IC LoRAs = video control unlocked. Outpainting, character stylizatio

IC LoRAs are a technique for controlling video generation, likely enabling fine-grained stylistic or motion control through specialized LoRA (Low-Rank Adaptation) models applied to video workflows. Th

Ollama announced they are expanding cloud infrastructure capacity by adding more GPUs to meet demand, asking users for patience during the scaling process. This indicates Ollama's cloud service experi

We believe the future of intelligence is decentralized. Tonight we're opening the new @agi_inc office in SF with a dinner, gathering chipmakers, OEM partners, and investors to discuss what happens whe

As of April 2026, Latent Space published a curated ranking of the top locally-runnable AI models, providing a research-based overview of the best open-weight or self-hostable models available at that

MiniMax M2.7 is a large language model now available for local deployment through LM Studio, notable for its strong performance in agentic tool calling tasks. The model requires a substantial minimum

Split-stack videos in one click. Upload an image and it generates multiple angles, stacked and ready to go. Want more control? Jump into graph mode and tweak every step. Link to workflow is below 👇 Me

After seeing that Claude Mythos marketing turned out to be, as expected, a scam, I wanted to make a master list of tricks being used to market LLMs. The master list includes statements directly from l

I see every week on X an announcement or demo which implies that robotic manipulation has been solved. The only reason I don't believe it is because manipulation had already been solved last week by s

👇 “I still maintain that LLMs are really dumb and of limited use. it’s my opinion as a practitioner and professional that the hype being peddled by the AI corporations and self promoters (from X grift

In a world where writing code to build websites and apps is trivial (thank you Lovable, Cursor, Claude,...), the real differentiation for you and your company (and what makes you successful) will be h

Tesla is down to the last few hundred Model S & X cars in inventory. Poignant end of an era. Traded in my 2020 Model S for a brand new plaid X before they discontinue it. Car is amazing, but the FSD h

Transformers.js, a library that enables running state-of-the-art Hugging Face machine learning models directly in the browser with no server required, has experienced explosive NPM download growth ...

🚀 GLM-5.1 is now available on SiliconFlow Start a task, go to sleep. GLM-5.1 plans, executes, and self-improves for 8 hours and delivers high-quality results by morning. Still open-source. Big kudos t

Google's Gemma 4 is pretty wild. You can now run it locally with OpenClaw in 3 steps. 1. Install Ollama 2. Pull Gemma 4 model 3. Launch OpenClaw with Gemma as the backend Private local AI agents in mi

Strong release! GLM-5.1 is a DeepSeek-V3.2-like architecture (including MLA and DeepSeek Sparse Attention) but with more layers. And the benchmarks look better throughout! Looks like THE flagship open

We partnered with @Zai_org to bring GLM-5.1 to Modal. Free to try as an endpoint for the next month. GLM-5.1 further improves upon GLM-5's coding abilities and long-horizon effectiveness. Introducing

GLM-5.1 is now available through OpenCode Go, a low-cost subscription service ($5 first month, then $10/month) that provides access to open-source coding models with a zero data retention policy, m...

GLM-5.1 is Z.ai's (zai-org) next-generation open-source flagship model for agentic engineering, achieving state-of-the-art performance on SWE-Bench Pro and significantly outperforming its predecess...

INCREDIBLE GLM-5.1 weights are now opensource > i’ve had early access to the weights for the past few days > and yeah… this one matters a lot benchmarks? > SWE-Bench Pro: 58.4 > beats Opus 4.6 (57.3)