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

H
Help Net Security
月光博客
月光博客
IT之家
IT之家
B
Blog RSS Feed
T
Tailwind CSS Blog
The GitHub Blog
The GitHub Blog
博客园 - 三生石上(FineUI控件)
MyScale Blog
MyScale Blog
J
Java Code Geeks
Stack Overflow Blog
Stack Overflow Blog
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
博客园 - Franky
博客园 - 叶小钗
阮一峰的网络日志
阮一峰的网络日志
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
U
Unit 42
博客园_首页
B
Blog
V
V2EX
腾讯CDC
Vercel News
Vercel News
量子位
Microsoft Security Blog
Microsoft Security Blog

Hacker News: Show HN

PurrrrrFocus: Pomodoro Timer App - App Store Workflow Engine — Multi-Step Orchestration for Bun RapidPhoto: Pro Photo Editor App - App Store GitHub - DheerG/swarms: Achieve extraordinary results with claude code across a variety of tasks SPICE simulation → oscilloscope → verification with Claude Code — Lucas Gerads Show HN: VCoding – A 5 MB native Windows IDE with no dynamic dependencies Show HN: LLMs don't hallucinate because they're bad at math, it's the format GitHub - Agent-FM/agentfm-core: AgentFM is a peer-to-peer network that turns everyday computers into a decentralized AI supercomputer. AgentFM lets you run massive AI workloads directly across a global mesh of idle CPUs and GPUs. Show HN: Tracking Top US Science Olympiad Alumni over Last 25 Years GitHub - Potarix/agent-hub: One place to talk to all your agents Show HN: Runtime security for AI agents(injection,tool abuse, data exfiltration) GitHub - dubeyKartikay/lazyspotify: Terminal Spotify client for macOS and Linux GitHub - the-banana-tool/king-louie: Easy to use GUI Personal AI Assistant. Win/Linux/Mac. Show HN I made my vacation rental bookable by AI agents–no Airbnb, 0% commission GitHub - basteez/jsf-autoreload: maven plugin to enable hot reload on jsf projects uvm32/hosts/host-gdbstub at main · ringtailsoftware/uvm32 GitHub - labsai/EDDI: Config-driven engine that turns JSON into production-grade AI agents. Multi-agent orchestration, 12+ LLM providers, MCP/A2A protocols, RAG, persistent memory, and enterprise compliance (EU AI Act, GDPR, HIPAA). Built on Quarkus. GitHub - glitchnsec/fortyone-oss: AI Executive Assistant Platform Quickstart | Alien GitHub - muxshed/shed: One stream in, or many. Every destination, simultaneously. No cloud middleman, no per-channel fees, no limits. GitHub - ocrbase-hq/ocrbase: 📄 PDF/IMG ->.MD/JSON Document OCR API for PaddleOCR and GLMOCR. Self-hostable. GitHub - impactjo/home-memory: MCP server that lets your AI assistant remember everything about your home. GitHub - Sets88/dbcls: DbCls is a powerful terminal database client that supports various databases GitHub - neptun2000/heor-agent-mcp GitHub - SeanFDZ/macmind: Single-layer transformer in HyperTalk for the classic Macintosh RollQuation: Math Puzzles - Apps on Google Play GitHub - dropbox/witchcraft Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis GitHub - opentalon/opentalon: OpenTalon is an open-source platform built from the ground up in Go as a robust alternative to OpenClaw LinkedIn™ 职位抓取工具 - Chrome 应用商店
GitHub - milind-soni/Decaf
milindsoni20 · 2026-04-29 · via Hacker News: Show HN

Take the boil out of comment threads.

A Chrome extension that uses Gemini Nano via Chrome's Prompt API to rewrite any page live to whatever tone you ask for. No API keys. No backend. Nothing leaves your browser.

Works on Hacker News, X, Reddit, and most article/comment pages.

Use it

  1. Open HN or X.
  2. Click the floating coffee cup (bottom-right).
  3. Type a rule — "make every comment polite", "3-word summary", "pirate speak".
  4. Watch each comment / post stream a rewrite in real time.
  5. Click the next to any block to see the original.

What it preserves

Facts, stance, and voice. Only the tone changes. A pro-X take stays pro-X — Decaf defangs the delivery, not the position. Click ↺ to see the unfiltered original anytime.

Setup

Needs Chrome 138+ + ~22 GB disk / 4 GB VRAM / 16 GB RAM (Gemini Nano's hardware floor).

  1. chrome://flags/#optimization-guide-on-device-modelEnabled BypassPerfRequirement
  2. chrome://flags/#prompt-api-for-gemini-nanoEnabled, then relaunch
  3. git clone https://github.com/milind-soni/Decaf.git
  4. chrome://extensionsDeveloper modeLoad unpacked → pick the folder
  5. Click the toolbar icon → Download Gemini Nano (~4 GB, one time)

If anything's stuck: chrome://on-device-internals is the diagnostic page.

How it works

Vanilla JS, no build step. The service worker keeps one warm LanguageModel session and clones it per request. Each rewrite streams via promptStreaming with a JSON-schema-constrained response; a hand-rolled parser extracts the rewritten field character-by-character and pipes it over a chrome.runtime Port to the content script, which swaps element.textContent live. Original HTML is stashed on each element's dataset so the ↺ can put it back exactly.

Limits

  • English-strong only (en / es / ja).
  • First rewrite is slow (model warmup); subsequent ones are sub-second.
  • Nano-quality rewrites — fast, free, fully local. Not Claude-quality.

License

MIT.