AI Digest

Digest curado

jueves, 23 de abril de 2026·light·short·6,590 tokens

🔥 TOP — lo que SÍ o SÍ tenés que ver

  • [Preflight] — Herramienta clave para testear tu MCP server antes de mandarlo a Claude/OpenAI y evitar rechazos por bugs boludos. link
  • [Claude Code for the Outer Loop] — Un playbook concreto de cómo usar Claude Code para workflows de SRE y on-call, aplicado a tu interés en agentes. link
  • [Skills CLI] — Herramienta de Vercel para instalar skills en agentes como Claude Code, directamente relevante para tu tooling. link
  • [What 81k Claude users said] — Investigación oficial de Anthropic sobre la economía de la IA, data directa de la fuente. link
  • [Pgrust] — Un proyecto para reescribir Postgres en Rust con IA, toca justo tu interés en bases de datos e ingeniería de software. link

📦 Claude / Anthropic ecosystem

  • [SportScore MCP] — Un server MCP de código abierto que da datos de deportes en vivo gratis para usar con Claude. link
  • [Discord group says it accessed Claude Mythos] — Noticia sobre un supuesto acceso a una versión no lanzada de Claude, pura curiosidad. link

🛠️ Dev tools & coding

  • [Qwen3.6-27B] — Un modelo open-weight de 27B parámetros que promete performance de "flagship" en coding, competencia directa. link
  • [RedAI] — Una herramienta de seguridad que usa agentes de validación para probar vulnerabilidades en entornos reales, no solo detectarlas. link
  • [Langfuse] — La plataforma open-source de observabilidad para LLMs sigue siendo trending, clave para monitorear agentes. link

🏗️ Software engineering

  • [SAW-INT4] — Paper de arXiv sobre cuantización 4-bit del KV-Cache para servir LLMs, toca optimización de performance a bajo nivel. link
  • [OpenMetadata] — Plataforma unificada open-source para metadata, descubrimiento y gobernanza de datos, arquitectura de data a gran escala. link

💤 Skippeable pero conviene saber

  • [Hermes Agent] — Un framework de agente "self-improving" de NousResearch, para tener en el radar. link
  • [CubeSandbox] — Sandbox de Tencent para probar agentes de IA, un proyecto grande de un gigante tech. link
  • [Shopify’s AI Phase Transition] — Entrevista larga con el CTO de Shopify sobre su adopción masiva de IA, data interna interesante pero densa. link

Artículos fetched (39)

  • open-metadata/OpenMetadata
    github-trending

    OpenMetadata is a unified metadata platform for data discovery, data observability, and data governance powered by a central metadata repository, in-depth column level lineage, and seamless team collaboration. Empower your Data Journey with OpenMetadata What is OpenMetadata? OpenMetadata is a unified metadata platform for data discovery, data observability, and data governance powered by a central metadata repository, in-depth column-level lineage, and seamless team collaboration. It is one of the fastest-growing open-source projects with a vibrant community and adoption by a diverse set of companies in a variety of industry verticals. Based on Open Metadata Standards and APIs, supporting connectors to a wide range of data services, OpenMetadata enables end-to-end metadata management, giv…

  • KeygraphHQ/shannon
    github-trending

    Shannon Lite is an autonomous, white-box AI pentester for web applications and APIs. It analyzes your source code, identifies attack vectors, and executes real exploits to prove vulnerabilities before they reach production. Note 📢 Sunsetting Router Mode (claude-code-router)`. → Shannon — AI Pentester by Keygraph Shannon is an autonomous, white-box AI pentester for web applications and APIs. It analyzes your source code, identifies attack vectors, and executes real exploits to prove vulnerabilities before they reach production. What is Shannon? Shannon is an AI pentester developed by Keygraph. It performs white-box security testing of web applications and their underlying APIs by combining source code analysis with live exploitation. Shannon analyzes your web application's source code to …

  • vercel-labs/skills
    github-trending

    The open agent skills tool - npx skills skills The CLI for the open agent skills ecosystem. Supports OpenCode, Claude Code, Codex, Cursor, and 41 more. Install a Skill npx skills add vercel-labs/agent-skills Source Formats # GitHub shorthand (owner/repo) npx skills add vercel-labs/agent-skills # Full GitHub URL npx skills add https://github.com/vercel-labs/agent-skills # Direct path to a skill in a repo npx skills add https://github.com/vercel-labs/agent-skills/tree/main/skills/web-design-guidelines # GitLab URL npx skills add https://gitlab.com/org/repo # Any git URL npx skills add git@github.com:vercel-labs/agent-skills.git # Local path npx skills add ./my-local-skills Options Option Description -g, --global Install to user directory instead of project -a, --agent <agents...> Target spe…

  • Z4nzu/hackingtool
    github-trending

    ALL IN ONE Hacking Tool For Hackers All-in-One Hacking Tool for Security Researchers & Pentesters What's New in v2.0.0 Feature Description 🐍 Python 3.10+ All Python 2 code removed, modern syntax throughout 🖥 OS-aware menus Linux-only tools hidden automatically on macOS 📦 185+ tools 35 new modern tools added across 6 categories 🔍 Search Type / to search all tools by name, description, or keyword 🏷 Tag filter Type t to filter by 19 tags — osint, web, c2, cloud, mobile... 💡 Recommend Type r — "I want to scan a network" → shows relevant tools ✅ Install status ✔/✘ shown next to every tool — know what's ready ⚡ Install all Option 97 in any category — batch install at once 🔄 Smart update Each tool has Update — auto-detects git pull / pip upgrade / go install 📂 Open folder Jump into any t…

  • koala73/worldmonitor
    github-trending

    Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface World Monitor Real-time global intelligence dashboard — AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface. Documentation · Releases · Contributing What It Does 500+ curated news feeds across 15 categories, AI-synthesized into briefs Dual map engine — 3D globe (globe.gl) and WebGL flat map (deck.gl) with 45 data layers Cross-stream correlation — military, economic, disaster, and escalation signal convergence Country Intelligence Index — composite risk scoring across 12 signal categories Finance radar — 92 stock exchanges, commodities, crypto, and 7-…

  • AIDC-AI/Pixelle-Video
    github-trending

    🚀 AI 全自动短视频引擎 | AI Fully Automated Short Video Engine 🎬 Pixelle-Video —— AI 全自动短视频引擎 English | 中文 https://github.com/user-attachments/assets/a42e7457-fcc8-40da-83fc-784c45a8b95d 只需输入一个 主题,Pixelle-Video 就能自动完成: ✍️ 撰写视频文案 🎨 生成 AI 配图/视频 🗣️ 合成语音解说 🎵 添加背景音乐 🎬 一键合成视频 零门槛,零剪辑经验,让视频创作成为一句话的事! 🖥️ Web 界面预览 📋 最近更新 ✅ 2026-01-26: 新增「动作迁移」模块,上传参考视频和图片进行动作迁移 ✅ 2026-01-14: 新增「数字人口播」和「图生视频」流水线,新增多语言 TTS 音色支持 ✅ 2026-01-06: 新增 RunningHub 48G 显存机器调用支持 ✅ 2025-12-28: 支持 RunningHub 并发限制可配置,优化 LLM 返回结构化数据的逻辑 ✅ 2025-12-17: 支持 ComfyUI API Key 配置,支持 Nano Banana 模型调用,API 接口支持模板自定义参数 ✅ 2025-12-10: 侧边栏内置 FAQ,锁定 edge-tts 版本修复 TTS 服务不稳定问题 ✅ 2025-12-08: 支持固定脚本多种分割方式(段落/行/句子),优化模板选择交互逻辑支持直接预览选择 ✅ 2025-12-06: 修复视频生成 API 返回 URL 路径处理,支持跨平台兼容 ✅ 2025-12-05: 新增 Windows 整合包下载,优化图片与视频反推工作流 ✅ 2025-12-04: 新增「自定义素材」功能,支持用户上传…

  • langfuse/langfuse
    github-trending

    🪢 Open source LLM engineering platform: LLM Observability, metrics, evals, prompt management, playground, datasets. Integrates with OpenTelemetry, Langchain, OpenAI SDK, LiteLLM, and more. 🍊YC W23 Langfuse Is Doubling Down On Open Source Langfuse Cloud · Self Host · Demo Docs · Report Bug · Feature Request · Changelog · Roadmap · Langfuse uses GitHub Discussions for Support and Feature Requests. We're hiring. Join us in product engineering and technical go-to-market roles. Proudly made with ClickHouse open source database Langfuse is an open source LLM engineering platform. It helps teams collaboratively develop, monitor, evaluate, and debug AI applications. Langfuse can be self-hosted in minutes and is battle-tested. ✨ Core Features LLM Application Observability: Instrument your app an…

  • Clippy Is Back, Your Desktop AI Paperclip Agent over GUI
    hn-ai· 23-abr

    Article URL: https://github.com/AmrDab/clippyai-desktop Comments URL: https://news.ycombinator.com/item?id=47871012 Points: 1 # Comments: 0

  • SportScore MCP – free live sports data for Claude
    hn-ai· 23-abr

    Article URL: https://github.com/Backspace-me/sportscore-mcp Comments URL: https://news.ycombinator.com/item?id=47871351 Points: 1 # Comments: 0

  • Show HN: RedAI – AI-driven vulnerability discovery and live validation
    hn-ai· 22-abr

    Most AI security tools stop at flagging code that looks vulnerable. You end up with a wall of "potential" findings and no way to know which ones are real without manually reproducing each one. Finding and fixing are both cheap now; actually proving which findings are real and impactful is the last remaining hurdle. RedAI is built to close that gap. After scanner agents surface candidates, validator agents take each one into a live environment — a running instance of the target — and try to prove or disprove the finding. They navigate UIs, hit endpoints, write PoC scripts, spin up helper servers, capture logs and screenshots. The end result is a report of real, reproducible vulnerabilities with PoC steps and screenshots to prove it. Comments URL: https://news.ycombinator.com/item?id=478697…

  • CubeSandbox: Tencent's sandbox for ai agents
    hn-ai· 23-abr

    Article URL: https://github.com/tencentcloud/CubeSandbox Comments URL: https://news.ycombinator.com/item?id=47871063 Points: 1 # Comments: 0

  • Failed Companies Are Selling Old Slack Chats and Email Archives to Train AI
    hn-ai· 22-abr

    Article URL: https://gizmodo.com/failed-companies-are-selling-old-slack-chats-and-email-archives-to-train-ai-2000747916 Comments URL: https://news.ycombinator.com/item?id=47869253 Points: 21 # Comments: 7

  • Hermes Agent – A self-improving AI agent
    hn-ai· 23-abr

    Article URL: https://hermes-agent.nousresearch.com Comments URL: https://news.ycombinator.com/item?id=47871176 Points: 1 # Comments: 0

  • HuntifyAI – I built an AI tools while deep in AI research and vibe coding
    hn-ai· 23-abr

    Article URL: https://huntifyai.com/ Comments URL: https://news.ycombinator.com/item?id=47870959 Points: 1 # Comments: 0

  • Design.md: The New Open Contract Between Designers and AI
    hn-ai· 22-abr

    Article URL: https://kyanfeat.substack.com/p/designmd-the-new-open-contract-between Comments URL: https://news.ycombinator.com/item?id=47870227 Points: 1 # Comments: 0

  • Show HN: Preflight – Test your MCP server before submitting to Claude/OpenAI
    hn-ai· 23-abr

    I submitted an MCP server to Claude and OpenAI, got rejected from OpenAI after 4 weeks for a stupid OAuth redirect bug. Fix took an hour. Resubmit wait was another 3 weeks. Every check they run is something you can test locally in 15 seconds, so I built preflight for everyone to use. FREE. So: https://m8ven.ai/preflight — paste your URL, get the report. Free, no signup, probes your live endpoint only (doesn't touch your code). Built this out of M8ven, which I'm working on as an trust layer for the agent economy. Keeping Preflight free because the rejection loop is a stupid tax on people shipping. Would love to hear failure modes I'm not catching. Comments URL: https://news.ycombinator.com/item?id=47871631 Points: 2 # Comments: 0

  • Pgrust: Rebuilding Postgres in Rust with AI
    hn-ai· 22-abr

    Article URL: https://malisper.me/pgrust-rebuilding-postgres-in-rust-with-ai/ Comments URL: https://news.ycombinator.com/item?id=47870335 Points: 3 # Comments: 0

  • Discord group says it accessed Claude Mythos by guessing location
    hn-ai· 23-abr

    Article URL: https://mashable.com/article/discord-group-accesses-claude-mythos-claims Comments URL: https://news.ycombinator.com/item?id=47870947 Points: 3 # Comments: 0

  • Americans Turning to AI to Supplement Healthcare Visits
    hn-ai· 22-abr

    Article URL: https://news.gallup.com/poll/707789/americans-turning-supplement-healthcare-visits.aspx Comments URL: https://news.ycombinator.com/item?id=47870211 Points: 7 # Comments: 3

  • Ask HN: Is modern AI like glasses for the mind?
    hn-ai· 22-abr

    And if so, wouldn't using modern AI make people's effective intelligence less varied between individuals than natural human intelligence? Comments URL: https://news.ycombinator.com/item?id=47870699 Points: 1 # Comments: 2

  • How are we supposed to use AI?
    hn-ai· 22-abr

    Article URL: https://petersmj.com/blog/technology-in-research/ Comments URL: https://news.ycombinator.com/item?id=47869447 Points: 1 # Comments: 1

  • In two years nobody will care if actors are AI or not–director Mathieu Kassovitz
    hn-ai· 22-abr

    Article URL: https://www.theguardian.com/film/2026/apr/22/actors-ai-la-haine-director-mathieu-kassovitz Comments URL: https://news.ycombinator.com/item?id=47870129 Points: 2 # Comments: 3

  • SAW-INT4: System-Aware 4-Bit KV-Cache Quantization for Real-World LLM Serving
    hn-ai· 22-abr

    Article URL: https://arxiv.org/abs/2604.19157 Comments URL: https://news.ycombinator.com/item?id=47870305 Points: 1 # Comments: 0

  • The AI Power Bottleneck: Data Centers Meet the Steel Monopoly
    hn-ai· 22-abr

    Article URL: https://blog.adafruit.com/2026/04/22/the-ai-power-bottleneck-data-centers-meet-the-steel-monopoly/ Comments URL: https://news.ycombinator.com/item?id=47869694 Points: 2 # Comments: 0

  • Bolt AI builder: Create apps and websites by chatting with AI
    hn-ai· 22-abr

    Article URL: https://bolt.new Comments URL: https://news.ycombinator.com/item?id=47869331 Points: 1 # Comments: 0

  • I'm Using Claude Code for Everything Else but Coding
    hn-ai· 22-abr

    Article URL: https://chandlernguyen.com/blog/2026/04/22/im-using-claude-code-for-everything-else-but-coding/ Comments URL: https://news.ycombinator.com/item?id=47869866 Points: 1 # Comments: 2

  • Sacrifice your job for the glorious AI future
    hn-ai· 22-abr

    Article URL: https://disconnect.blog/sacrifice-your-job-for-the-glorious-ai-future/ Comments URL: https://news.ycombinator.com/item?id=47869129 Points: 2 # Comments: 0

  • Z.ai phasing out original subscription plans
    hn-ai· 23-abr

    Article URL: https://docs.z.ai/devpack/transition Comments URL: https://news.ycombinator.com/item?id=47871339 Points: 1 # Comments: 0

  • Many anti-AI arguments are conservative arguments
    hn-ai· 22-abr

    Article URL: https://www.seangoedecke.com/many-anti-ai-arguments-are-conservative/ Comments URL: https://news.ycombinator.com/item?id=47870473 Points: 14 # Comments: 7

  • Ars Technica newsroom AI policy
    hn-ai· 22-abr

    Article URL: https://arstechnica.com/staff/2026/04/our-newsroom-ai-policy/ Comments URL: https://news.ycombinator.com/item?id=47869145 Points: 15 # Comments: 6

  • Fraud and the false optimism of AI for science
    hn-ai· 23-abr

    Article URL: https://statmodeling.stat.columbia.edu/2026/04/22/fraud-and-the-false-optimism-of-ai-for-science/ Comments URL: https://news.ycombinator.com/item?id=47871047 Points: 1 # Comments: 0

  • How to turn off Google AI overview
    hn-ai· 23-abr

    Article URL: https://substack.com/@grefiti/p-192810091 Comments URL: https://news.ycombinator.com/item?id=47871108 Points: 2 # Comments: 0

  • AI Tokens Optimization
    hn-ai· 23-abr

    Article URL: https://tokenpinch.com Comments URL: https://news.ycombinator.com/item?id=47871168 Points: 1 # Comments: 1

  • Show HN: DeckWeaver – Create AI-powered Google Slides presentations in minutes
    hn-ai· 22-abr

    As I frequently run bespoke training sessions and project-based workshops for clients, I create quite a lot of presentations and, naturally, to some extent rely on ChatGPT and similar tools to help me create the content. However, as I use Google Slides (rather than a Markdown-based presentation tool), I faced the significant problem of having to manually import the content into Google Slides presentations via copy and paste, which was very time-consuming. Hence, I built a solution called DeckWeaver that not only takes this tedious work off my hands, but also allows me to review generated content directly from within the application, which has the added benefit of fewer context switches and media breaks: https://usedeckweaver.com/ Comments URL: https://news.ycombinator.com/item?id=47870690…

  • What 81000 Claude users said about the economics of AI
    hn-ai· 22-abr

    Article URL: https://www.anthropic.com/research/81k-economics Comments URL: https://news.ycombinator.com/item?id=47869957 Points: 1 # Comments: 1

  • Claude Code for the Outer Loop: An AI SRE Playbook
    hn-ai· 22-abr

    Article URL: https://www.arcade.dev/blog/claude-code-ai-sre-oncall-workflows/ Comments URL: https://news.ycombinator.com/item?id=47869934 Points: 1 # Comments: 0

  • An AI-gile Manifesto (2026-2030)
    hn-ai· 22-abr

    Article URL: https://www.elitesoftwareengineer.com/blog/the-aigile-manifesto-2026-2030 Comments URL: https://news.ycombinator.com/item?id=47870512 Points: 1 # Comments: 1

  • Shopify’s AI Phase Transition: 2026 Usage Explosion, Unlimited Opus-4.6 Token Budget, Tangle, Tangent, SimGym — with Mikhail Parakhin, Shopify CTO
    latentspace· 22-abr

    A rare interview with Shopify's CTO on -everything- that Shopify is doing to maximize AI for their customers, with exclusive data on their own AI adoption.

  • Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
    simonw· 22-abr

    <p><strong><a href="https://qwen.ai/blog?id=qwen3.6-27b">Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model</a></strong></p> Big claims from Qwen about their latest open weight model:</p> <blockquote> <p>Qwen3.6-27B delivers flagship-level agentic coding performance, surpassing the previous-generation open-source flagship Qwen3.5-397B-A17B (397B total / 17B active MoE) across all major coding benchmarks.</p> </blockquote> <p>On Hugging Face <a href="https://huggingface.co/Qwen/Qwen3.5-397B-A17B/tree/main">Qwen3.5-397B-A17B</a> is 807GB, this new <a href="https://huggingface.co/Qwen/Qwen3.6-27B/tree/main">Qwen3.6-27B</a> is 55.6GB.</p> <p>I tried it out with the 16.8GB Unsloth <a href="https://huggingface.co/unsloth/Qwen3.6-27B-GGUF">Qwen3.6-27B-GGUF:Q4_K_M</a> quantized version and <…