NVIDIA's RTX Spark N1X puts up to 128 GB of unified memory and a Blackwell GPU in a Windows laptop. In Artificial Analysis's first AA-AgentPerf-Local results, an RTX Spark laptop completes its local agent workload on Qwen3.8 27B in 18.7 minutes. Artificial Analysis describes RTX Spark's N1X chip as essentially the GB10, linking to https://t.co/520A4lOOdB
NVIDIA AI introduces Nemotron 3.5 Bolt 24GB, saying it fits into 24GB of memory. The quoted @NVIDIARTXSpark post describes 74B-A9B parameters packed into 24GB of memory, optimized for RTX Spark and designed to run alongside your apps. https://t.co/SOiBgeMMjD https://t.co/yPsBt9N58A
great doing business with you john apple 🍎🤝 https://twitter.com/johnternus/status/2110789142326165518
Muse Spark Open — weights out today! Best US open-source model, and runs on the new Mac Studio 🍎 https://t.co/C8PU0RvlUA
Huawei says its mass-produced Kirin 9050 Pro uses logic folding to shorten signal paths through vertically connected circuit layers, supporting performance and efficiency gains in the Mate 90 Pro Max. The article explains how dense hybrid bonding enables this layout and how memory coordination allows Kirin 9050 chips to run a 30B MoE model with less than 2.5GB of additional memory usage. Huawei also says Kirin 2027 chips are in testing, with results meeting expectations so far.
Anthropic warns that small AI models may soon enable self-replicating malware to spread across IoT devices, potentially forming computer worms that infect unfamiliar devices without human interaction. The company said such models may be able to replicate autonomously across IoT devices by next year. A recent experiment found significant challenges to building an AI malware replicator, tempering the concern; Anthropic also said one could be difficult to shut down, though its doubling time may be days.
Over half of reported crimes in Singapore use sophisticated social engineering. With @GovTechSG, we’re building on-device fraud detection using our open model Gemma. By fine-tuning it on local scam tactics, it identifies threats in real time without compromising personal privacy → https://goo.gle/4ygVkJa
Google DeepMind and Singapore’s Government Technology Agency (GovTech) have delivered a Gemma model fine-tuned on local scam conversations and optimized for typical smartphones in Singapore. A lightweight classifier screens incoming messages before the model evaluates broader conversation context and triggers warnings about suspected manipulation, with conversation processing kept on the device. Training excluded private or personally identifiable information, using anonymized datasets derived from scam reports and intelligence feeds.
Why it matters: The two-stage detection pipeline offers a transferable approach to balancing smartphone power consumption with contextual scam analysis while keeping private conversations on the device.
Anthropic researchers William Zhou and Daniel Freeman report that small AI models could become capable of autonomous replication across poorly secured consumer devices by the end of 2026. Gemma4-26B-A4B ran across 15 Android TV boxes costing about $55 each, while GLM-5.2 replicated across all eight test nodes in four of five runs; the experiments omitted actual hacking. No current model combines the small size and replication capabilities needed for a fully autonomous AI worm.
GPT-4 was the undisputed best model in the world for a full year and we all debated whether anyone could catch up or if OpenAI had some secret formula that nobody else knew about. Now there are are 19 AI models that can run on a modern smartphone that are better than GPT-4!
A Verge writer began running local LLMs on an M5 Ultra Mac Studio with 256GB of unified memory, installing the open-source Hermes Agent desktop app and the 125-billion-parameter Qwen 3.8 Flash Next model, about 105GB in size. Early tasks included a daily briefing via Telegram, reorganizing a 400-plus-game Steam library, and local data analysis; the writer is now working on automating laptop benchmark tests.
Honor launched the Magic9 Pro Max, Magic9, and Magic9 Ultra at a Beijing event on September 28, 2026, priced from 4,499 yuan and on sale the same day. The phones pair ARRI co-tuned dual 200MP camera systems with MagicOS 11, which commercially debuts an Agent Harness framework built on a 397B agent model co-developed with Alibaba.
AI hardware startup Natura launched Interface, a $99 smart ring. A press on the finger can ask an AI agent to complete tasks or record thoughts, without taking out a phone.
Google released an experimental note-taking app, Google AI Edge Foresight, that transcribes meetings and audio files completely offline. It is free, powered by its on-device EmbeddingGemma 2 model, and currently runs on macOS.
Cal AI co-founder Zach Yadegari raised $10 million for his new AI agent company, Persona, in a round led by Vine Ventures. Persona is positioned as a personal AI assistant and plans to launch a $179 wearable wristband in December. It has already attracted thousands of users through a free iMessage beta.
Google released the meeting notes app Google AI Edge Foresight, positioned as a competitor to Granola. It runs entirely offline on Mac and uses the on-device EmbeddingGemma 2 model with 740 million parameters to capture meeting content across apps, including in-person meetings.
Blind preorders for the Lenovo YOGA Pro 15 RTX Spark opened nationwide at 9:00 on October 8. It is one of the first local-agent AI PCs to feature the NVIDIA RTX Spark N1X superchip.
StepFun Terminal will hold a 'Ready Builder One' launch event in Shanghai on October 13 to introduce its first foundation-model-native agent phone, STEPX Neo. The model, operating system, and hardware were all designed natively for agents. The event will also announce the company's latest ecosystem partnerships.
At its first in-person event in two years, Microsoft launched the new Surface Laptop Ultra and announced a series of Windows 11 changes, while outlining its vision for local AI and agent workflows in personal computing.
At a Tech Week event in San Francisco, Microsoft announced the specifications and prices for the new Surface Laptop Ultra. The two base models start at $2,600 and $3,700; optional memory and storage upgrades raise the price to as much as $5,900. Microsoft says the top configuration has sold out.
If you need anymore evidence about how exciting local AI (and llamacpp) is! From the @Microsoft windows launch today with @satyanadella @sriramk @JensenHuang
We’re supercharging Copilot on Windows with Hybrid Intelligence. With your permission, Copilot can tap into the context on your PC, take action for you, and use local models when it makes sense, giving you more capability while helping your tokens go further.
Microsoft opened preorders for the high-end Surface RTX Spark Dev Box AI mini-PC for developers at about $6,000, with shipping starting in November. It uses Nvidia's Arm-based RTX Spark platform and 128 GB of unified memory, is optimized for local AI, and can run models with more than 120B parameters. Use cases include testing AI models, prototyping agents, and running AI coding tools locally.
At its Windows and Surface launch event in San Francisco, Microsoft announced the Surface Laptop Ultra. The base model, with an Nvidia RTX Spark Arm-based chip, an 8-core CPU, 24 GB of memory, and 512 GB of storage, starts at $2,599. It goes on sale October 16 and replaces the Surface Connect port with built-in magnetic USB-C charging.
At a Windows AI and Surface event in San Francisco, NVIDIA and Microsoft said they are designing software and hardware together for the agent era on Windows PCs. Microsoft announced that its OS-level Microsoft Execution Containers (MXC) infrastructure is generally available, allowing agents to run persistently in the background under system control.
Why it matters: NVIDIA and Microsoft are delivering both the system-level containers and local compute needed to run agents, helping readers assess the conditions for local agent workflows.
Liquid AI released two open-weight models in its d1 decision-model family: d1-3B and the experimental d1-omni-600M.
Why it matters: Liquid AI open-sourced two decision models and reported on-device latency and results across seven benchmarks, providing evidence on the use of small models for structured decisions.
At MIT Future Fest, Tony Fadell said first-generation AI hardware such as Rabbit R1, Humane AI Pin, and Limitless Pendant failed because it did not solve real needs and was merely 'technology that geeks find interesting'.
.@GoogleDeepMind EmbeddingGemma 2 is now available on Ollama. Made for consumer devices and is multimodal. ollama pull embeddinggemma-2 Quoted post by @GoogleDeepMind (context, not the main author's work): Meet EmbeddingGemma 2, our first natively multimodal open model for on-device embeddings. It expands beyond text to unify code, images, audio, and video in a shared space. 🧵
Google DeepMind released EmbeddingGemma 2, a 740M-parameter open model based on the Gemma 4 architecture and licensed under Apache 2.0. It maps text, images, audio, and video into a shared embedding space.
Why it matters: EmbeddingGemma 2 maps text, images, audio, and video into one embedding space and reports concrete on-device memory and vector truncation figures, helping assess local multimodal retrieval.
Google released EmbeddingGemma 2, an open multimodal embedding model that converts text, images, video, audio, and code into numerical vectors. It has 740 million parameters. Google says it outperforms competitors with more than twice as many parameters on multimodal embedding benchmarks and scores 78.68 on the Massive Text Embedding Benchmark (Code), nearly 10 points above the previous model's 68.76.
Startup Hark officially launched Hark Pro, an AI personal assistant that is free to use, with a paid subscription for heavy users. Built on a model trained specifically to operate computers, it can connect to email, calendars, hard drives, and credit cards to handle digital tasks for users. It shows the agent browsing the web in a small window to build trust.
Apple is reportedly developing a smart-home camera that does not save video, instead using AI to convert visual data into text snippets describing what is happening at home, with facial recognition support. Sandeep Waraich, who leads Google's wearable-device work, has also suggested that a smart-glasses camera could act only as an image sensor without saving footage.
Google DeepMind released EmbeddingGemma 2, which it calls the first natively multimodal open embedding model for on-device use. It combines code, images, audio, and video with text in a shared space.
Sundar Pichai announced EmbeddingGemma 2, described as Google's first open, natively multimodal embedding model. It handles text, code, image, video, and audio tasks in a modular 740M parameter form factor, is positioned for offline, privacy-first RAG when paired with Gemma 4, and is said to outperform some specialist models more than twice its size. Weights are available now on Hugging Face.
The open-source RemoveMacAI command-line tool can quickly remove Apple Intelligence from macOS 27. Its developer says it uses a configuration profile approved through System Settings together with Apple's own resource service, keeps System Integrity Protection enabled, and does not directly modify /System.
Australia is considering banning smart glasses in some public places, and courts in several countries have issued bans. Norway's government says it does not plan a blanket ban but considers a proposed restriction a 'highly prioritized matter'. A draft law will soon go to parliament; the government says it does not want a society where people fear being secretly recorded, photographed, or filmed.
The open-source RemoveMacAI command-line tool can turn off Apple Intelligence features, delete its local models, and prevent macOS from downloading them again. Its developer says it can free up about 12 GB of space.
At its October 1 Mate 90 series and all-scenario product launch, Huawei introduced the Mate 90 lineup, all powered by its flagship τ chip. Its stacked logic-chiplet architecture delivers 125 TB/s of cross-die bandwidth and cuts critical-path latency by 30%. Overall performance on the Mate 90 Pro Max is 31% higher than the previous generation.
Meta announced the open-source Muse Gadgets project, providing ESP32 firmware and a Linux SDK for hobbyists to build their own devices connected to its AI agent Muse. The code is licensed under Apache 2.0.
Meta launched the open-source Muse Gadgets project, providing open firmware and a Linux SDK for developers to build devices connected to its personal AI agent Muse using inexpensive hardware such as Raspberry Pi or ESP32. It offers examples including a color e-ink display, an HDMI TV stick, and other projects, and has opened a Discord channel to support users.
