惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Simon Willison's Weblog
Simon Willison's Weblog
Google DeepMind News
Google DeepMind News
CTFtime.org: upcoming CTF events
CTFtime.org: upcoming CTF events
P
Proofpoint News Feed
Recent Announcements
Recent Announcements
MongoDB | Blog
MongoDB | Blog
U
Unit 42
云风的 BLOG
云风的 BLOG
Recorded Future
Recorded Future
G
Google Developers Blog
I
InfoQ
Blog — PlanetScale
Blog — PlanetScale
A
About on SuperTechFans
Jina AI
Jina AI
量子位
宝玉的分享
宝玉的分享
The Cloudflare Blog
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
博客园 - 聂微东
Last Week in AI
Last Week in AI
WordPress大学
WordPress大学
美团技术团队
The Hacker News
The Hacker News
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
D
Darknet – Hacking Tools, Hacker News & Cyber Security
T
Tailwind CSS Blog
博客园 - 司徒正美
博客园 - 叶小钗
Hugging Face - Blog
Hugging Face - Blog
P
Palo Alto Networks Blog
博客园_首页
阮一峰的网络日志
阮一峰的网络日志
博客园 - 【当耐特】
Spread Privacy
Spread Privacy
The GitHub Blog
The GitHub Blog
Y
Y Combinator Blog
Vercel News
Vercel News
Martin Fowler
Martin Fowler
cs.CL updates on arXiv.org
cs.CL updates on arXiv.org
Forbes - Security
Forbes - Security
Attack and Defense Labs
Attack and Defense Labs
Google DeepMind News
Google DeepMind News
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
Microsoft Azure Blog
Microsoft Azure Blog
P
Privacy International News Feed
G
GRAHAM CLULEY
The Last Watchdog
The Last Watchdog
C
Cyber Attacks, Cyber Crime and Cyber Security
AI
AI
V2EX - 技术
V2EX - 技术

Hacker News - Newest: "AI"

AI can't read an investor deck AI as an attorney? Student uses ChatGPT, Gemini to sue UW over alleged racial discrimination Hacking MCP Servers in AI Systems – The Rug Pull: Tool Changes After Approval GitHub - MeepCastana/KubeezCut: Free Web based video editor GitHub - GenAI-Gurus/awesome-eu-ai-act: Curated tools, official sources, OSS, templates, and guides for EU AI Act compliance. Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers Coming soon: 10 Things That Matter in AI Right Now DARPA built an AI to fact-check enemy weapons claims What explains heterogeneity in AI adoption? When AI Meets Muscle: Context-Aware Electrical Stimulation Promises a New Way to Guide Human Movements - Department of Computer Science AI Changed How We Build. It Did Not Change What Matters. Linux rules on using AI-generated code - Copilot is OK, but humans must take 'full responsibility for the… Meta spins up AI version of Mark Zuckerberg to engage with employees Code Mode: Let Your AI Write Programs, Not Just Call Tools | TanStack Blog GitHub - Delavalom/graft: Go framework for building AI agents. Type-safe tools, multi-provider (OpenAI, Anthropic, Gemini, Bedrock), zero vendor SDKs. India's TCS tops estimates, says new AI models did not dent services demand Gen Z's fading AI hype Strong feeling: we are in a folded AI reality GitHub - machinarii/total-recall-catalog: A reference catalog of latest knowledge retrieval, memory & RAG systems GitHub - mensfeld/code-on-incus: Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically.. Quantization, LoRA, and the 8% Problem: Benchmarking Local LLMs for Production AI Iran war: We spoke to the man making Lego-style AI videos that experts say are powerful propaganda Powell, Bessent discussed Anthropic's Mythos AI cyber threat with major U.S. banks GitHub - immartian/bellamem: Persistent belief-graph memory for AI agents. Retrieves decisive context by importance — not recency, not RAG, not /compact. recursive-mode: The Repo-Native Operating System for AI Engineering After the attack on Sam Altman's home, will AI CEO's go on the offensive? The biggest advance in AI since the LLM Opus 4.6 vs GPT 5.4 One Prompt Unity World Generation Test “AI polls” are fake polls Client Challenge Can AI be a 'child of God'? Inside Anthropic's meeting with Christian leaders How to Switch AI Chatbots and Why You Might Want To GitHub - MattMessinger1/agentic_refund_guardrail: Safe refund policy layer for AI agents — Python + TypeScript. Same behavior, shared tests. Adam/papers/emergent_values_whitepaper.md at master · strangeadvancedmarketing/Adam Ask HN: How do you stop playing 20 questions with your AI coding tools How far can automation and AI support psychotherapy? - @theU GitHub - stagas/rtdiff: realtime git diff gui and AI-assisted commits A Mac Studio for Local AI — 6 Months Later A History of the Early Years of AI at the University of Edinburgh Why AI Coding Tools Still Feel Stuck on Localhost MSN AI Datacenters Are Becoming Strategic Targets twitter.com Penn Researchers Use AI to Surface Unreported GLP-1 Side Effects in Reddit Posts Show HN: MoodSense AI (ML and FastAPI and Gradio, Deployed on Hugging Face) Moodsense Ai - a Hugging Face Space by aman179102 AI models are terrible at betting on soccer—especially xAI Grok GitHub - xialeistudio/echoic GitHub - HimashaHerath/github-dev-wrapped: AI-powered weekly GitHub activity reports deployed to GitHub Pages GitHub - alejandrobalderas/claude-code-from-source: Architecture, patterns & internals of Anthropic's AI coding agent — reverse-engineered from source maps AI and Tech brief: Ireland ascendant GitHub - Titovilal/context0: Context0 - Never Surrender Training for a Marathon with an AI Coach: What Worked and What Didn't Cyber Pulse: Agentic Intel - Apps on Google Play I Built an AI PR Reviewer That Catches Bugs by Not Looking for Bugs Gen Z workers are so fearful AI will take their job they’re intentionally sabotaging their company’s AI rollout | Fortune How AI Is Reimagining the Game of Golf–For Both Players and Courses GitHub - nattergabriel/reseed: A CLI tool for managing and distributing agent skills across projects Is SVG the final frontier? My AI workflow evolved from prompts to a near-autonomous workflow MLSharp Help - 3DGS Viewer & Generator I put my cognitive field based AI's runtime on GitHub Is Numble the first AI-proof game? A3: Kubernetes for autonomous AI agent fleets | Emergent Principles Deepali Vyas ("The Elite Recruiter") GitHub - msmarkgu/RelayFreeLLM: A restful API designed to route user prompts to various AI model providers. Unionized ProPublica staff are on strike over AI, layoffs, and wages Unleashing the Advantage of Quantum AI We're heading for an AI-fueled 'dementia crisis,' brain scientist warns The AI-Assisted Breach of Mexico's Government Infrastructure [pdf] GitHub - stef41/lmscan: 🔍 Detect AI-generated text and fingerprint which LLM wrote it. Open-source GPTZero alternative. Zero dependencies, works offline. MSN GitHub - visionscaper/collabmem: Enabling long-term collaboration with Agentic AI - building up episodic and world model memory over time with in-context awareness We gave an AI a 3 year retail lease in SF and asked it to make a profit | Andon Labs AI Code is Hollowing Out Open Source, and Maintainers are Looking the Other Way What leaked "SteamGPT" files could mean for the PC gaming platform's use of AI AI is the boss at this retail store. What could go wrong? GitHub - Wuzu11517/agentic-proxy: Local proxy meant to help reduce With Drones, Geophysics and ArtificiaI Intelligence, Researchers Prepare to Do Battle Against Land Mines A Single Operator, Two AI Platforms, Nine Government Agencies: The Full Technical Report 在 Steam 上购买 FriedrichAI: Offline AI 立省 10% GitHub - inevolin/resume-cli: Hit Claude usage limits? Resume any AI coding session elsewhere. Switch tools at zero friction. GitHub - atripati/ark: AI Runtime Kernel — a context operating system for AI agents. Eliminates tool bloat, loads only what’s needed, and gives LLMs their reasoning space back. How to Build a Secure AI PR Reviewer with Claude, GitHub Actions, and JavaScript This Startup Wants You to Pay Up to Talk With AI Versions of Human Experts Intel Arc Pro B70 Brings 32GB VRAM to Local AI for $949 WordPress 7.0: The Good, the AI, and the Still Missing AI on the couch: Anthropic gives Claude 20 hours of psychiatry IatroBench: Pre-Registered Evidence of Iatrogenic Harm from AI Safety Measures AI Agents Know About Supabase. They Don't Always Use It Right. The history and future of AI at Google, with Sundar Pichai Inside an AI‑enabled device code phishing campaign How Meta Used AI to Map Tribal Knowledge in Large-Scale Data Pipelines AI for Systems: Using LLMs to Optimize Database Query Execution Forecasting the Economic Effects of AI Introducing Tinker: Play with AI, bring your ideas to life AI sheds light on an ancient gaming mystery People really hate AI but not as much as Iran—or Democrats | Fortune What is an AI Product Engineer? Phoebe Gates wants her $185 million AI startup to succeed with 'no ties to my privilege or my last name': 'I have a chip on my shoulder' | Fortune
Project Huginn — Distributed GPU Sharing for AI Training
SMware ApS · 2026-06-17 · via Hacker News - Newest: "AI"

Unify heterogeneous GPUs (2GB–32GB+) into a single compute pool. Train AI models at lower cost with measurable sustainability impact.

Why Choose Hugin?

One network to train your AI — cheaper, private, and built for the physical world.

🔒

Private by design

You choose the protection level on every job: Shield+ keeps your data encrypted and split so no machine ever sees the whole — or Vault runs on sealed hardware no one else can read.

Built for Physical AI

More than fine-tuning. Hugin Learning trains robot, drone and automation control from scratch — our own breakthrough, for the problems others can’t train.

Verified & reliable

Every result is independently re-checked, and a slow or dropped machine never stalls your job. You get a model you can trust — completely hands-off.

What you can build

Built for real AI projects

Bring your data, train production-ready models across the network, then download and use them anywhere.

Custom language models

Fine-tune open models (LLaMA, Mistral, Phi, Gemma, Qwen) on your own data with LoRA & QLoRA.

Chatbots · Assistants · Code · Multilingual

Computer vision models

Train image classification & object-detection models on your own images, then download the trained model.

Quality control · Drones · Cameras · Inspection

Robotics & automation

Build perception models and prepare data — augmentation, validation and benchmarking for robotics teams.

Perception · Dataset prep · Model evaluation

Hugin Learning

Train control policies for robots and agents — the system discovers behaviour from scratch by trial-and-error, no labelled data needed.

Robots · Control · Agents · Automation

How Hugin Works

Six-step pipeline from job definition to verified delivery with transparent billing.

Job Definition

Data classification, target metric, and pool selection.

Cost Estimate

HU & GPU-seconds estimate with upper bound and SLA.

Sharding & Scheduling

Capability-aware distribution, node selection, redundant execution.

Execution

Sandboxed micro-shard execution with telemetry, health checks, auto-retry.

Aggregation & Verification

Quality threshold, spot-check, and anomaly detection validation.

Billing & Payment

Verified job billing, automatic GPU owner payout.

The journey

From your data to your trained model

You upload, we take care of everything, and your finished model is ready to download.

Upload your data

Secure & private

We prepare it

Fully automatic

Trained across our network

Powered at scale

Turned into your model

Brought together

Download the result

Ready to use

Our breakthrough

Hugin Learning

Our own innovation: control intelligence that learns from scratch — by trial-and-error, with no labelled data — trained and verified across the network. Built for robots, drones and automation.

Learns from zero

Discovers the right behaviour by trial-and-error — no demonstrations or labelled data required.

Built for Physical AI

Balancing, positioning, navigation and your own custom tasks — the brain that runs robots, drones and automated lines.

Distributed & verified

Trained across the network and independently re-checked, so the result can be trusted.

⬡ Engineered by Hugin

Your data, protected by design

Most GPU networks ship your raw data to strangers' machines. Hugin doesn't. Every job runs under Hugin Shield+ — a multi-layer protection stack we built — or, for the most sensitive work, inside a hardware-isolated Vault.

🛡️ Hugin Shield+

Included on every job

Protection is built in on every job — at no extra cost. Your data stays private, safe, and completely yours.

  • 🔒Private by default — protected automatically the moment you upload
  • 🙈Yours alone — no one ever sees your full data
  • 🤝Trusted partners only — your work runs on verified, reliable machines
  • 😌Total peace of mind — no setup, no risk, no surprises

🔒 Hugin Vault

Maximum confidentiality

Maximum confidentiality. Your data stays invisible — even to the computer that trains it.

  • 🧠Sealed private space — your data is only opened inside a protected area
  • 🙈Invisible to the owner — the machine’s owner can never see it
  • 📜Proven genuine — every machine is verified before it touches your data
  • 🏥For your most sensitive data — health, finance & personal records

You choose the protection level per job — Shield+ is included at the standard rate; Vault adds hardware isolation for the most sensitive data.

Transparent HU-Based Pricing

HU (Hugin Unit) is your simple billing unit. 1 HU ≈ 3,600 normalized GPU-seconds. Pre-estimate and upper bound before every job.

0.21EUR / HU

Transparent Pricing for Training

  • More iterations with same budget
  • Transparent HU + GPU-seconds billing
  • Pre-estimate & upper bound guaranteed
  • Community, Verified, or Dedicated pools
  • ESG reporting: kWh/HU & CO₂e/HU

Start Training Now

Share Your Compute

Turn idle devices into passive value. Hugin network harnesses compute power from any connected device globally.

  • 📱 Smartphones (Browser or App)
  • 🚗 Electric Vehicles (EV Compute)
  • 🌐 Any Web Browser Tab
  • 💻 Desktop & Server GPUs

Join the Network

HU Rate by Device Class

Device Class~HU/HourExamples
Smartphones (Idle)0.05 - 0.15iPhone 15 Pro, Pixel 8
Electric Vehicles (EV)0.2 - 0.4Tesla MCU, Polestar 2
4-6GB Consumer0.4 - 0.6GTX 1650, RTX 3050
8-12GB Consumer0.6 - 0.9RTX 3060, RTX 4070
16-24GB Workstation1.6 - 2.8RTX 3090, RTX 4090
24-48GB+ Datacenter2.8 - 4.5A100, H100

Frequently Asked Questions

Everything you need to know about training AI models on Huginn.

What is Project Huginn?+

Project Huginn is a distributed GPU sharing platform that lets you train and fine-tune AI models at up to 50% lower cost than traditional cloud providers. It pools heterogeneous GPUs (from RTX 3080 to H100) into a unified compute grid. Built by SMware ApS, a Startup Denmark approved company.

How much does it cost to train an AI model on Huginn?+

Huginn uses Hugin Units (HU) for billing. 1 HU = €0.21. A typical LoRA fine-tuning of a 7B model with 1,000 training examples costs approximately 5-8 HU (€1-2). Costs are estimated before you start, with an upper bound guarantee — no surprise charges.

What AI models can I fine-tune?+

Huginn provides 7 base models for free: LLaMA 3.1 (8B & 70B), Mistral 7B, Phi-3 Mini 3.8B, CodeLlama 7B, Gemma 2 9B, and Qwen 2.5 7B. You select a base model and upload your own training data — Huginn handles the rest.

What training methods does Huginn support?+

Three methods: LoRA (Low-Rank Adaptation) for fast, efficient training with ~12GB VRAM; QLoRA (4-bit quantized) for even lower memory usage with ~8GB VRAM; and Full Fine-Tune for highest quality with ~40GB+ VRAM. LoRA with rank 16 is recommended for most use cases.

What data formats are supported for training?+

JSONL (Alpaca format with instruction/input/output fields), CSV (with instruction, input, output columns), and Chat Format JSONL (with messages array). Maximum upload size is 500MB. We recommend at least 1,000 examples for good results.

How is Huginn different from other GPU cloud platforms?+

Unlike traditional GPU rental platforms that only offer raw compute, Huginn provides an all-in-one AI training experience. You get a built-in Model Studio with 7 free base models, an integrated fine-tuning pipeline, automatic cost estimation with upper bounds, and a ChatGPT-style playground to test your model instantly — no setup, no configuration, just upload your data and start training.

Can I earn money by sharing my device?+

Yes! Huginn lets you earn from virtually any device. Share your desktop GPU, laptop, tablet, smartphone, or even your electric vehicle's compute power while you drive. Connect via the Huginn Agent app or simply open a browser tab. Every connected device contributes to the network and earns you HU. Desktop GPUs earn the most (up to €0.60/hour for high-end cards), but even a smartphone or browser tab generates passive income 24/7.

How do I share my GPU and start earning?+

It takes under 5 minutes: 1) Sign in with Google and choose GPU Owner role, 2) Go to GPU Agent page in your dashboard, 3) Download the agent file and generate a token, 4) Run the command in your terminal. The agent auto-detects your GPU and starts processing jobs. You can also use the browser-based Web Engine with zero downloads. Read our full guide at projecthuginn.com/how-to-share for step-by-step instructions.

Is my training data secure?+

Yes. All training jobs run in sandboxed environments with complete data isolation. Each user's models and datasets are private. We use session-based authentication and encrypted connections for all data transfers.

Can I download my trained model?+

Yes. After training completes, you can download your fine-tuned model in .safetensors format. You can run it locally with Ollama, llama.cpp, vLLM, or HuggingFace Transformers. You can also keep it on Huginn and test it via the built-in playground.

What GPU tiers are available?+

Five tiers: T1 (H100/A100 80GB+, 4.50 HU/hr), T2 (A100 40GB/RTX 4090, 2.80 HU/hr), T3 (RTX 4080/3090, 1.60 HU/hr), T4 (RTX 3080/2080 Ti, 0.90 HU/hr), and T5 (entry-level, 0.40 HU/hr). Higher tiers are faster but cost more per hour.