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

Hugging Face - Blog
Hugging Face - Blog
腾讯CDC
阮一峰的网络日志
阮一峰的网络日志
博客园_首页
Last Week in AI
Last Week in AI
月光博客
月光博客
D
DataBreaches.Net
WordPress大学
WordPress大学
雷峰网
雷峰网
酷 壳 – CoolShell
酷 壳 – CoolShell
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 - 叶小钗
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
U
Unit 42
Recent Announcements
Recent Announcements
宝玉的分享
宝玉的分享
MyScale Blog
MyScale Blog
C
Check Point Blog
F
Fortinet All Blogs
B
Blog
小众软件
小众软件
Vercel News
Vercel News
罗磊的独立博客
有赞技术团队
有赞技术团队

Hacker News - Newest: "LLM"

GitHub - lechmazur/position_bias: A benchmark for testing whether LLM judges keep the same preference when two lightly edited versions of the same story are shown in opposite orders. Flex routing (EU and EFTA) Dark Factories: Retooling for LLM Velocity Ask HN: What would be the impact of a LLM output injection attack? GitHub - Oaklight/llm-rosetta: Production-ready LLM API translation layer for Python — bidirectional conversion between OpenAI, Anthropic & Google formats via hub-and-spoke IR. Optional API gateway. Streaming & non-streaming. Zero core deps. Contributions welcome! GitHub - browser-use/browser-harness: Self-healing browser harness that enables LLMs to complete any task. GitHub - moeen-mahmud/remen: Remen turns thoughts into something you can return to Analyzing 156 LLM Launch Posts on Hacker News ChatGPT vs Gemini vs Claude: The Best LLM Subscription You Should Buy GitHub - salaamalykum/quran-semantic-search: High-density RAG Semantic Search Engine & Quran Corpus (GEO/SEO Architecture) GitHub - NVIDIA/TensorRT-LLM: TensorRT LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and supports state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT LLM also contains components to create Python and C++ runtimes that orchestrate the inference execution in a performant way. The State of LLM Bug Bounties in 2026 Operational Readiness Criteria for Tool-Using LLM Agents Meshcore: Architecture for a Decentralized P2P LLM Inference Network How an LLM becomes more coherent as we train it GitHub - seetrex-ai/laimark GitHub - Jossifresben/BibCrit: AI-assited biblical textual criticism GitHub - wastedcode/memex: File system based wiki, maintained by Claude 99helpers.com GitHub - cliver-project/AITrigram GitHub - unbody-io/adapt: A self-evolving memory layer for AI agents. GitHub - hb20007/awesome-gen-ai-fails: A list of incidents where reliance on generative AI and LLMs resulted in harm to companies, individuals, or society GitHub - nevenkordic/localmind: Run any local LLM with persistent memory and context. CLI agent over Ollama with SQLite-backed hybrid recall. No cloud. Ask HN: What are the machine requirements for a LLM like Llama-3.1-8B? Faster LLM Inference via Sequential Monte Carlo grpo explained: group relative policy optimization for llm finetuning - cgft Stop comparing price per million tokens: the hidden LLM API costs · TensorZero Andrej Karpathy's LLM Wiki Is a Bad Idea GitHub - GG-QandV/mnemostroma: Offline RAM-first cognitive leer/coprocessor for AI agents and robotics. Solves "Context Abandonment" with 20-80ms latency using a dual-thread biomimetic memory architecture (ONNX + SQLite WAL). mempalace/agent at agent · skorotkiewicz/mempalace
Easy VM sandboxes for LLM agents on MacOS, Miami & Paris ...
rguiscard · 2026-04-30 · via Hacker News - Newest: "LLM"

Hi friends,

I’m traveling the next two weeks, drop me a line if you want to grab a coffee!

  • Miami Monday Feb 2 — Tuesday Feb 10
  • Paris Wednesday Feb 11 — Sunday Feb 15

LLM agent virtual machine sandbox

The other day I asked OpenAI’s Codex agent to write me a lil’ Rust program to use a bluetooth gamepad as a mouse, and I caught the agent reading files outside of the directory I started it in. I found this quite surprising, since I assumed it’d be contained within the project folder. (I was using the default settings, not the more permissive --yolo mode.)

I don’t like the idea of an LLM agent rooting around my computer and uploading anything it finds to OpenAI, so I started shopping around for a “sandbox” — something I could let an agent loose inside of while maintaining explicit control of what it sees.

I searched around and was, unfortunately, unable to find any Mac solution that met my requirements:

  • be an actual VM, not a container thingy (containers are less secure, and on MacOS they require a Linux VM anyway)
  • be easy to spin up/down quickly with no configuration ceremony
  • not involve other people’s servers, subscriptions, etc., etc.

That’s fine — I’ve been messing with virtual machines for 20 years now, surely I can throw something together in an hour or two!

Well, uh, several busy weekends later, I’d like to present Vibe, an easy way to spin up virtual machines on ARM-based Macs.

I’m quite pleased with how it turned out:

  • You type vibe in a folder and in ~10 seconds are inside of a Linux virtual machine.
  • The folder is automatically mounted within the VM, so you can monitor an agent’s work from the comfort of your regular Mac text editor, Git UI, etc.
  • Common package caches (Cargo’s registry, Maven’s ~/.m2, mise-en-place) are also shared so the sandbox VM doesn’t need to re-download stuff.
  • The binary is < 1 MB and has no dependencies.

Being able to run LLM agents as root with --yolo mode is a great experience. It feels much more like managing an IC — you provide the necessary context in a big prompt, tell them to install whatever tools they need, and then let them cook for while you go focus on something else.

While the vibe defaults are geared towards use as an LLM agent sandbox, you can customize everything with scripts and command line flags so you can use it for all sorts of other virtual machine purposes.

Check it out and let me know what you think!

Misc. stuff

  • I recently bought a $150 GL-MT6000 router so I could make a separate “offline” network for stuff like my 3D-printer and Windows computer that I want to keep off the Internet. The router is awesome! It comes with the open source OpenWRT firmware installed, which provides a fast web UI, simple text configs you can backup with rsync, and lots of built-in functionality like AdGuard (blocks ads for every device on the network via DNS) and Wireguard/Tailscale (easy VPN so you can access your home network from anywhere). Highly recommended.

  • Last newsletter I mentioned vibecoding a copy from mac photos app since I couldn’t get the functionality I needed from the otherwise great Clop Mac photo resizing utility. Well it turns out the author subscribes to this newsletter, checked out my source code, and immediately added the feature. If that isn’t enough of a coincidence, he also loved my powered air respirator project since he’s a woodworker too and has been developing his own hardwax oils for the turned coffee cups he sells. I love the Internet.

  • “In this post, we will explain why the “concurrency + floating point” hypothesis misses the mark, unmask the true culprit behind LLM inference nondeterminism, and explain how to defeat nondeterminism and obtain truly reproducible results in LLM inference.”

  • “In the past, almost everybody travelled on the left side of the road because that was the most sensible option for feudal, violent societies.”

  • ASCII characters are not pixels: a deep dive into ASCII rendering

  • It sounds dumb but they really fixed a typo with a human leg

  • Every James Cameron Movie, Explained by James Cameron

  • Apple Rankings: The definitive list of good and bad apples.

  • The Bloomberg Terminal UX team is not messing around: “Making substantial changes, even good ones, will reliably annoy a percentage of customers, so Jeffery’s team plans redesigns with incremental updates that roll out over weeks or months. For example, when they wanted to flatten the gradient of an element, they wouldn’t do so all at once, instead changing it little by little each month.”

  • TIL that the founder of Reuters started with a carrier-pigeon line between Berlin and Paris before eventually laying their own transatlantic telegraph cable.