iCreatAI News
Back to main site
Skip to content
#On device
Nov 4Wed
  1. Hacker News AI31

    EdgeDAE: Acceleration of Diffusion Action Experts for Real-Time Physical AI

    EdgeDAE is a heterogeneous FPGA-GPU system that offloads the perception-heavy vision transformer to the GPU while accelerating Diffusion Action Expert (DAE) inference on an FPGA by storing all parameters on-chip in BRAM/URAM, eliminating the repeated DRAM parameter loading of iterative denoising. It co-designs quantization, fixed-point arithmetic, and hardware-efficient random number generation for the FPGA fabric. Compared with an edge GPU baseline, it reduces end-to-end inference latency by 52.5% for Octo-Small and 38.7% for Octo-Base with up to 2.10x higher throughput, and achieves about 17x higher energy efficiency than an RTX 4090.

  2. Hacker News AI75

    Show HN: Umbra – An AI agent that uses Android apps while you use your phone

    Umbra is an Android agent that reads the screen via the accessibility tree and screenshots, then taps, types, scrolls and opens apps in an invisible background display so you can keep using the phone. It also runs a Debian Linux environment on-device for building and testing code, supports Anthropic, OpenAI, Google, OpenRouter or any OpenAI-compatible model, and requires your approval for risky actions like payments, messages or deleting data. It requires Android 16 (API 36), root is optional, and distribution is by sideload from umbra-agent.pages.dev.

  4. Hacker News AI22

    Would Local AI Become the Dr.Pepper?

    Using Dr Pepper's rise to America's No. 2 soda as an analogy, the post argues that local AI can earn a lasting place by fitting specific needs rather than outperforming frontier models. It notes limits of scaling—finite public training data and diminishing returns from added compute—and suggests tasks like routing, personal knowledge management, and dictation could move to edge devices where privacy, speed, and local control matter. It cites Google AI Edge Foresight, Fluid voice, LiteLLM, Ollama, and DSH's creator mode as examples for experimenting with locally hosted models.

  6. The Verge · AI58

    Apple’s latest watchOS beta includes its new always-on Audio Intelligence features

    The new watchOS 27.3 developer beta adds Audio Intelligence features that let Siri use the Apple Watch microphone: Siri Recap, which uses ambient listening to generate conversation summaries, and Live Rewind, which produces text transcripts of the last 15 seconds of audio the mic detected. Apple also released a redesigned Health app for testing in the iOS 27.3 developer beta and announced expanded language support in global English, Chinese (Simplified and Traditional), French, Japanese, Korean, Norwegian, Portuguese, Spanish, Turkish, and Vietnamese. Apple published an Audio Intelligence Privacy Overview stating that raw audio is handled within dedicated hardware, is not saved as a file, and is not accessible to the operating system, apps, or Apple.

  7. Apple Newsroom58

    Redesigned Health app expands language support

    Apple's redesigned Health app expands language support to global English, Chinese (Simplified and Traditional), French, Japanese, Korean, Norwegian, Portuguese, Spanish, Turkish, and Vietnamese. Built with Apple Intelligence, it adds a Longevity tab with a Health Age feature derived from Apple Watch metrics and at-home movement evaluations that run entirely on device. The app will be available on Apple Intelligence-enabled iPhone and iPad models with the latest software, with some features limited by region, language, or device.

  8. Simon Willison50

    Qwen 3.8 27B on my laptop can interpret model weights, too

    Simon Willison tested the "Sparks of Weight-Space Programming" paper's game-identification prompts, which ask a model to name which of four Atari-style games a policy network was trained on using only its raw weights and no tools. Running Qwen 3.8 27B locally via LM Studio, he got correct answers for Pac-Man (46m 12s, 43,506 reasoning tokens), Pong (1h 4m 54s) and Breakout (52m 25s), but it failed on Freeway. He notes critics like Matthew Honnibal and Sean Goedecke dispute the paper's framing, and he focuses on the no-tools question rather than claims that source code is obsolete.

  9. Hacker News AI38

    HPE Spaceborne Computer-4 Will Bring AI and HPC to the Moon on Astrolab Flip

    HPE announced that its Spaceborne Computer-4, a compact fanless compute module with a low-power CPU and GPU and a Kioxia BG-series M.2 SSD, will run AI workloads including LLM retrieval-augmented generation, vision inference and LiDAR processing aboard Astrolab's FLIP rover, landing near the Moon's South Pole under NASA's CLPS initiative. The mission is intended to launch in late 2026 on Voyager Technologies' Griffin-1 lander, with NASA running its FLUID federated learning experiment and a tuned LLM test, and Deloitte providing its Silent Shield predictive cyber-analytics system. HPE will support Astrolab's mission control in Hawthorne, California, using HPE Private Cloud and HPE CloudOps Software.

  10. Hacker News AI45

    Show HN: I built Knemora – An Open-source AI Second brain

    Knemora is an open-source, local-first knowledge and agent platform with a native Second Brain graph, a portable Markdown Wiki, and local AI inference that runs without a mandatory cloud service, external LLM API, or vector database. The repository is prepared for its first public release, v0.1.0-alpha, and supports managed setup on macOS Apple Silicon with Node.js 24, local Qwen3-8B inference via llama.cpp, document ingestion, BM25 search, and backup/restore workflows.

Nov 3Tue
  1. Hacker News AI20

    Consumer Appetite for Personal AI Goes Mainstream

    A Qualcomm Technologies-commissioned Personal AI report, based on a survey of more than 12,000 people across the U.S., U.K., Germany, India, Brazil, and China, finds consumers shifting from single devices toward connected ecosystems coordinated by a personal AI agent, with 70 percent expressing strong interest in an AI assistant that interacts naturally through voice and manages devices, apps, and everyday tasks. The report says 60 percent of wearable owners use AI at least weekly, while privacy, security, and user control are the top adoption concerns, with almost 60 percent wanting wearable data kept private or shared only with approval. Purchase drivers vary by category, from audio quality for earbuds and headphones to display quality for smartwatches, materials and finish for rings and pendants, and control type for smart glasses.

  3. TechCrunch · AI58

    Apple Watch’s AI features for rewinding and recapping conversations begin beta testing

    Apple has begun beta testing its AI-powered "Audio Intelligence" features for Apple Watch Series 12 and Apple Watch Ultra 4, initially for developers in the iOS 27.3 update, after announcing them at its September hardware event. The four opt-in features include sound recognition, Shazam-based song recognition, Live Rewind (which shows the last 15 seconds of speech as a text snippet when you double-press the digital crown), and Siri Recaps, which summarize conversations into short notes without saving raw audio or identifying speakers. Apple says processing relies on the S11 chip, the iPhone, and Private Cloud Compute, with snippets deleted after 30 seconds if unsaved and recaps deleted within seven days if not reviewed.

  5. Allen Institute for AI Blog34

    How Olmo 3 earned the highest ethical rating from Nextcloud

    Nextcloud awarded Olmo 3 its highest Ethical AI rating, green, which requires open-source training and inference software, a self-hostable trained model, and publicly available training data. Nextcloud bundles Olmo 3 with its llm2 local LLM backend, citing the 7B model's ability to run on consumer hardware without high-end GPUs and its fit for customers seeking independence from big tech. Nextcloud engineer Marcel Klehr said fully open models make AI transparent, explainable, auditable and controllable, and that the company is working on expanding its AI features.

Nov 2Mon
  1. Hacker News AI40

    Sam Altman on Local AI

    In an interview on the Cheeky Pint podcast, OpenAI CEO Sam Altman said running AI models locally pulls limited DRAM away from data centers, arguing for maximum cloud utilization in a supply-constrained environment. He said the memory shortage is worsening, calling the situation "bad" and saying the "invisible hand is failing right now," while noting memory makers remember past bankruptcies from overbuilding fabs. The article notes local AI offers better privacy and could let users avoid paying for cloud AI, and that local-AI PCs like Nvidia's RTX Spark laptops remain expensive.

  2. Hacker News AI36

    Blaize GSP [Graph Streaming Processor] Streams AI Graphs at the Edge

    Blaize's Graph Streaming Processor runs neural networks as directed graphs, with an on-chip hardware scheduler streaming the computational graph depth-first through 16 cores and keeping intermediate data in cache rather than external DRAM, delivering 16 TOPS at 7 W on Samsung 14 nm in the El Cano product. The next-generation GSP2 targets 1,600 TOPS on Samsung 5 nm for automotive ADAS, with production availability no earlier than 2028 per Blaize's 2025 10-K, which also flagged going-concern doubt amid a $103.8 million operating loss. DENSO and Mercedes-Benz have collectively paid $32 million in engineering fees against anticipated GSP2 delivery.

Nov 1Sun
  1. The Verge · AI38

    I built an AI Tamagotchi before Meta could

    The Verge's Sean Hollister bought a $50 Waveshare ESP32-S3 1.75C, a Meta-sanctioned devkit for building "Muse Gadgets," and found it nearly useless out of the box until he used Claude Code to add conversation history, spoken replies via ElevenLabs, a camera preview, multiple avatars, and a Tamagotchi-style minigame. He notes he has no coding experience and can't vouch for the code's quality, and that he keeps the device on guest Wi-Fi over privacy concerns. He concludes the appeal of dedicated hardware is reprogrammability rather than Muse itself, and wonders whether the real Muse Charm will be as freely programmable.

Oct 31Sat
  1. The Verge · AI38

    Sam Altman says it’s better for RAM to be in a data center than on your desk

    In an interview on the Cheeky Pint podcast, OpenAI CEO Sam Altman said it is "quite terrible" from a global allocation perspective for people to have the world's limited DRAM supply sitting in their own computers and used only five percent of the time, arguing for maximum utilization in the cloud amid supply constraints. The article notes that Samsung and SK Hynix have long-term deals with cloud operators and projects like OpenAI's Stargate that have reserved much of the world's RAM output for the next few years, which is why laptops, phones, and other tech are getting more expensive.

  2. NVIDIA Technical Blog22

    Run AI Models for Edge Application Development with Just a Few Commands

    AI is moving beyond the data center as developers build agents that act at the edge, including monitoring driver attention and cabin conditions, helping robots understand surroundings and plan tasks, and coordinating fleets of machines on production lines. These physical AI applications rely on models that perceive, reason, and act within tight memory, power, and latency budgets.

  3. Hacker News AI20

    ASRock Industrial AI Box-A395 Review: Taking Strix Halo to Work – ServeTheHome

    ServeTheHome reviews ASRock Industrial's AI Box-A395, a barebones mini-tower built around the AMD Ryzen AI Max+ 395 (16C/32T, 5.1GHz) with Radeon 8060S graphics and 128GB of soldered LPDDR5X-8000 memory. The aluminum, vertically oriented mini-ITX chassis has a 400W internal PSU, two empty M.2 2280 PCIe Gen4 x4 slots, Wi-Fi 7, and rear I/O including two 40Gbps USB-C ports, two HDMI 2.1 outputs, 10GbE (AQC113) and 2.5GbE (RTL8125BG); the only front-accessible ports are two USB 2.0 Type-A ports on top. The review frames it less as the "AI workstation" ASRock markets and more as a portable, self-contained Ryzen AI Max system for industrial environments.

  4. Hacker News AI20

    Show HN: NKTg AI – An LLM that decodes text before generating

    NKTg AI presents itself as a language decoding system that uses physical-style algorithms to measure semantic energy and extract a text's "Core Content" — the original sentence with the highest density of executable actions — classifying text as AMP, DAMP or STABLE. It pairs this decoding layer with an LLM that generates language from the standardized structure, offered as a browser-based local model running via WebAssembly (no server, history in localStorage) and as a free cloud mode. The post also lists output modes, claimed benefits such as hallucination mitigation, system requirements (16 GB RAM, 6-core CPU, WebGPU GPU with 6–8 GB VRAM), and a stated limitation on handwritten documents due to OCR accuracy.

Oct 29Thu
  1. Hacker News AI38

    Show HN: Howler, a macOS menu bar AI that burns the chat after reading

    Howler is a macOS menu bar AI app that opens a parchment-style letter with ⌥H, streams answers at a hand's pace, and burns the conversation when you press Esc or click outside, saving nothing. It supports chat-completions, Anthropic and Gemini formats with configurable base URL, model, body and headers, and stores only the model setup, prompt instructions, shortcut and API key in the Keychain. It is signed with a Developer ID and notarized by Apple, self-updates via Sparkle from v0.1.7, and is MIT licensed.

  2. Hugging Face Blog58

    Introducing AnySpeechRecognizer and AnyDecisionModel: Open Models for Speech and Decisions on Apple Platforms

    The author announces two Swift packages: AnySpeechRecognizer, which mirrors Apple's SpeechAnalyzer API while allowing alternative engines such as NVIDIA's Parakeet and Alibaba's Qwen3-ASR running locally on Core ML, plus speaker diarization via NVIDIA's Streaming Sortformer; and AnyDecisionModel, which answers yes-or-no, multiple-choice, and rating-scale questions with typed results and probabilities from backends including MLX, Core AI, llama.cpp, Jev, and Ollama. Both can download open models from the Hugging Face Hub and run them on device, require Swift 6.2 and macOS 26 or iOS 26 or later, and are published on GitHub.

  3. Hacker News AI60

    Opera adds local AI to its browsers

    Opera began shipping local AI models in its browsers as an early preview in Opera One, with Opera GX and Opera Neon to follow, using the WebAI standard to let sites and extensions run open models such as Google's Gemma and, on Macs, Apple's on-device model. WebAI is opt-in with site-specific permissions, and Opera also launched HTML-in-Canvas, which places interactive web pages inside 2D and 3D scenes.

  4. Google Developers Blog58

    Smaller apps and faster GPU inference with LiteRT in Google Play services

    LiteRT in Google Play services now offers GPU inference about 1.4x faster than the legacy GPU delegate via ML Drift and the new CompiledModel API, and apps that switch are about 8 MB smaller because the runtime ships with Play services instead of inside each app. The runtime supports phones back to Android 7, plus Wear OS 4+ and Android Automotive OS (Android 11+), and the legacy GPU delegate and custom delegate APIs are deprecated, with NNAPI to be removed starting in 2027.

  5. The Decoder72

    Google opens Android development to third-party AI agents and unveils new Gemini APIs

    At the Android Dev Summit, Google introduced Android Studio Rabbit 2 with a "Bring Your Own Agent" feature that lets developers run over 40 coding agents, including Anthropic's Claude and OpenAI's Codex, via the open Agent Client Protocol. Google also announced an alpha Agentic API powered by Gemini Nano 4 in ML Kit, the beta AppFunctions API built on the Model Context Protocol, and the XREAL AURA wired smart glasses running Android XR, priced at $1,099 and shipping mid-November.

Oct 28Wed
  1. Hacker News AI18

    Show HN: Curate. free, no ADS, with AI, private photo cleaner for iPhone

    Curate is a free iPhone and iPad app that lets you swipe right to keep or left to delete photos, videos and Live Photos, with nothing deleted until you review and confirm the delete pile. It offers filters for screenshots, screen recordings, selfies and other types, size and date sorting, duplicate grouping with a suggested keeper, and on-device Apple Intelligence suggestions for clutter like blurry shots or receipts. The app requires no account, shows no ads, does no tracking, uses no servers, and states that no data is collected.

  2. MarkTechPost69

    Liquid AI Releases LFM3: 6 Open-Weight 2B-Class Multimodal Models

    Liquid AI released LFM3, a family of six open-weight 2B-class models built on one shared tri-hybrid backbone covering text, vision, audio, omni and decision tasks, each fine-tunable on a single GPU. LFM3-1.7B has 1.73B parameters and a 256K context window, runs at about 1 GB of memory at 4-bit, and averages 49.37 across 29 benchmarks (Liquid's own eval), above Qwen3.5-9B at 47.24, while trailing it on GPQA Diamond (63.37 vs 80.40) and SWE-bench Verified (43.40 vs 51.20). Weights are on Hugging Face, the Liquid AI Playground and the Liquid API, with GGUF, MLX, ONNX, FP8 and NVFP4 checkpoints, and free commercial use stops at $10M annual revenue under the LFM Open License v1.0.

  3. Hacker News AI58

    Running Local AI on Nvidia DGX Spark Without CUDA

    CUDA++ is a research-stage toolchain and runtime that encodes GB10 (sm_121) machine code directly, replacing nvcc, PTX and libcuda with its own encoder and a client for NVIDIA's open kernel module ioctls, while NVIDIA's kernel module and signed GSP firmware remain. It currently supports only the DGX Spark's GB10, does not compile CUDA C++ or PTX, and reports partial coverage of CUDA's libraries; a head-to-head speed comparison against llama.cpp and vLLM has not been published.

  4. Hacker News AI58

    Survival LLM: AI assistant and emergency library that runs on your own computer

    Survival LLM is a free, open-source offline AI assistant and emergency reference library that runs entirely on the user's computer, answering questions from sources like FEMA, CDC, MedlinePlus, WikEM, military survival manuals, wikiHow, iFixit and about 930,000 Wikipedia articles while showing the exact passages used. It selects an AI model based on available memory (from IBM Granite 4.2 8B at 16–31 GB up to NVIDIA Nemotron 3.5 Super at 128 GB or more), requires Apple Silicon macOS 14+ or 64-bit Windows 10/11, and needs internet only during setup. The library is a 9 GB download (19 GB installed), or 1.3 GB (3 GB installed) without Wikipedia and iFixit, and users can also add their own PDFs, Word documents, ebooks and notes for the AI to search locally.

  5. Hugging Face Blog78

    Build specialized multimodal models with LFM3

    Liquid AI introduces LFM3, a generation of Liquid Foundation Models in the 2B parameter range that places text, vision, audio, omni, and decision models on one shared backbone built from LFM3-1.7B-Base, which was pre-trained on 22T tokens and uses a tri-hybrid layer mix with a 256K context window. The release includes six 2B-class models, new in-house vision and audio encoders, and LiquidCodec, with every model fine-tunable on a single GPU and the text model fitting in about 1 GB at 4-bit precision. All LFM3 models are open-weight and available on Hugging Face, with fine-tuning notebooks and agent harness guides provided.

    Why it matters: The post is a detailed technical release covering architecture changes, benchmark comparisons, and runnable code for six models, useful for developers evaluating small multimodal models for on-device deployment.

Oct 27Tue
  1. Apple Newsroom40

    Apple unveils new iMac with M6

    Apple announced a new iMac featuring the M6 chip, with an up-to-12-core CPU and GPU with Neural Accelerators in each core, a Dual 16-core Neural Engine, Wi-Fi 7, Bluetooth 6, and up to 2x faster storage, in a 24-inch 4.5K Retina all-in-one design running macOS 27. Apple claims up to 5x faster on-device AI performance than M4 and up to 18x faster performance than the most popular Intel-based iMac, with prices starting at $1,499 and pre-orders from October 28 ahead of availability on November 6.

  2. TechCrunch · AI58

    Opera introduces a new web standard to let users deploy AI locally in their browser

    Opera has introduced WebAI, a new web standard designed to let users run AI models directly in its browser without relying on the cloud, supporting models such as Google's Gemma and, on Macs, Apple's on-device foundation model. The feature is available as an early preview in Opera One, with Opera GX and Opera Neon to follow, and users must opt in. Opera cites privacy and no subscription or infrastructure costs as benefits, while noting that local model performance depends on the device's hardware.

  3. Ars Technica · AI58

    Apple releases iOS and macOS 27.2 with new Siri AI features, new Music logo

    Apple released the 27.2 updates across iOS, macOS, watchOS, tvOS, and visionOS, with iOS 27.2 adding a Siri memory feature that recalls past conversations, expanded data sources including voicemails and Messages attachments, an "Excluded Apps" option, and Siri AI support in French, Japanese, Korean, Portuguese, and Spanish. Other additions include context-aware suggestions in Messages, SynthID watermarking in Photos, Apple TV app user profiles, new Apple Music branding, and Apple Pencil support on the iPhone Duo foldable. macOS Golden Gate 27.2, watchOS, and visionOS 27.2 share the Siri updates, and macOS 27.2 beta references to gesture recognition in Photos, Notes, and Safari may relate to unconfirmed reports of touch support on the next MacBook Pro.

  4. Apple Newsroom48

    Siri AI is available in more languages starting today

    Apple Intelligence with Siri is available when the device and Siri language are set to Chinese (simplified or traditional), Danish, Dutch, English, French, German, Italian, Japanese, Korean, Norwegian, Portuguese, Spanish, Swedish, Turkish, or Vietnamese. Siri AI is in beta in iOS 27.2, iPadOS 27.2, macOS 27.2, watchOS 27.2, and visionOS 27.2, requires an Apple Intelligence-enabled device set to a supported language, and is currently unavailable in the EU on iOS, iPadOS, and watchOS. Some server-based Apple Intelligence features are subject to daily usage limits, with expanded access to be offered for a fee in the future.

  5. PyTorch Blog34

    Optimizing Streaming Speech Recognition on Arm® Edge Devices with ExecuTorch

    This PyTorch blog case study takes the multilingual Nemotron 3.5 streaming ASR model from PyTorch to ExecuTorch and profiles it on a Google Pixel 11, measuring how chunk size and precision affect streaming WER and processing time. Enabling Arm SME2 cut whole-recording processing time from 9.79 to 4.61 seconds for FP16 (2.12x) and from 5.20 to 3.24 seconds for QC8W (1.61x), with GEMM driving most of the gain, while longer chunks improved accuracy at the cost of latency. ETDump and Arm Streamline were combined to attribute speedups to operators and confirm which KleidiAI SME2 microkernels actually ran.

Oct 25Sun
  1. TechCrunch · AI65

    Block’s new voice model sounds eerily human

    Block released Dessa, a full-duplex speech-to-speech model trained entirely on human speech that runs locally on a single NVIDIA GPU or consumer hardware such as a MacBook M4 Max. The company says emotion is built into the model rather than driven by text or voice tags, and it has called for safety researchers to test Dessa, which is not publicly available.

Oct 24Sat
  1. The Decoder58

    Block's new AI model Dessa was trained entirely on human speech to catch emotions text-based models miss

    Block, the company behind Square and Cash App, has unveiled Dessa, an AI speech model it says is the first trained entirely on human speech with no text data, built from scratch. Dessa processes speech directly as audio rather than converting it to text first, letting it pick up on emotions, sarcasm, laughter and pauses in real time, and it supports full-duplex communication so it can listen and talk at the same time. It runs locally on a MacBook Pro M4 Max, was trained on a single NVIDIA GB200 NVL72, and is not yet publicly available, with the team first seeking AI security researchers to test it for unexpected risks.

  2. The Verge · AI58

    Jack Dorsey’s Block made an AI model that talks to you like an awkward new friend

    Block announced Dessa, an AI model trained "from scratch on nothing but human conversation," and shared sample chats that can be awkward and uncanny, including one where it mistakes the speaker for someone named "Ronaldo." The team describes Dessa as a "true speech-to-speech model" that can run locally on a MacBook Pro M4 Max "or better," citing curiosity, better understanding between humans and AI, and safety as reasons for building it. Block says it will give "a small group of AI safety researchers" early access to the model.