source&pool
A daily wire of long-form journalism, video, and discourse — filed, tagged, and laid out flat.
VOL. I·NO. 01
THURSDAY, OCTOBER 8, 2026
Hacker News1412X 主题热门1376CNBC104PANews77Channel News Asia56ChainCatcher51Coinness42SouthChinaMorningPost35YahooFinance32Crypto.news30Verge21Cointelegraph20aihot189to5Google17Decrypt17Variety16CoinDesk15IGN15MacRumors15Guardian14Engadget13TechCrunch13FoxBusiness129to5Mac11Kotaku11ArsTechnica10Eurogamer9NBC9DW8Gematsu8Gizmodo8BitcoinMagazine7NintendoLife6Polygon6AMBCrypto5BBC World5Deadline5Investor'sBusinessDaily5PushSquare5TechPowerUp5DigitalFoundry4GSMArena4Motor14XBOXWire4NPR4SeekingAlpha4WarhammerCommunity4Wccftech4AndroidAuthority3Fox3GameInformer3NintendoEverything3Notebookcheck3USAToday3VideoCardz3ABC2AlJazeera2AndroidPolice2AppleInsider2BusinessInsider2CBS2CNET2CNN2EIA2Electrek2FierceBiotech2GamesIndustry.biz2GAMINGbible2LosAngelesTimes2Nature2Pokemon2politico.eu2SEC2TechSpot2Drive2Hacker2Register2Tom'sGuide2VideoGamesChronicle2WindowsCentral2WIRED2404Media180Level1AboutAmazon1Aftermath1AFTVnews1AUTOMATON1AWealthofCommonSense1Barron's1bgr1BikeRadar1BleacherReport1BleepingComputer1BlockClubChicago1BuzzFeed1Carscoops1SanDiegoCountyCenter1Draftsim1DroidLife1Endpoints1Euronews1Federal Reserve1Fortune1Futurism1GamesBeat1GearPatrol1Hackaday1Hodinkee1HollywoodReporter1How-ToGeek1Independent1KOMO1Livingetc1magic.gg1Mashable1MyNorthwest1NoemaMagazine1NewYorkPost1NYT1CrudeOilPricesToday1PCWorld1PetaPixel1Phoronix1PokeBeach1PokémonGOHub1PureXbox1qz1RoadtoVR1RockPaperShotgun1SamMobile1ScreenRant1SeattleTimes1SearchEngineRoundtable1Yahoo1SlashGear1Slate1statnews1MobileReport1TweakTown1WeatherChannel1WhatHi-Fi?1WWD1ZDNET1
  1. 021Hacker NewsOCT · 01English

    Slot Machine Programming and the Hidden Curriculum

    A professor describes 'slot machine programming,' where students repeatedly prompt different AI tools with identical queries rather than developing problem-solving skills. The article argues that AI use has eliminated the implicit learning of crucial skills like debugging, problem decomposition, and technical communication that students historically acquired through struggle and direct interaction with instructors.

    By Michael Hilton
  2. 022Hacker NewsOCT · 01English

    Crafting Express – isolated local environments for AI coding agents

    Crafting Express is a Docker-based local evaluation platform that lets developers run and coordinate AI coding agents with features like agent delegation, multi-model support, and workspace sharing. The Express edition is for testing only; production deployments require Crafting Enterprise, which offers Kubernetes integration, multi-cloud deployment, and security controls.

    By Crafting-Dev
  3. 023Hacker NewsOCT · 01English

    The Note – A Markdown notebook with runnable code and formula tables

    THE Note is an open-source macOS markdown notebook application that combines live editing with executable code blocks and formula tables. Users can write in markdown, organize notes in folders, run Python/JavaScript/shell code locally, and optionally integrate AI assistants from OpenAI, Claude, or Gemini. The app stores notes as plain .md files locally with no cloud requirement.

    By tobias_digital
  4. 024Hacker NewsOCT · 01English

    Show HN: Graphene – Data analysis toolkit for your coding agent

    Graphene is a data analytics framework designed for coding agents, combining SQL with semantic layer governance to enable faster data exploration and visualization. It provides token-efficient languages, agent-friendly CLI controls, and supports multiple databases including Snowflake and BigQuery, with business logic stored in version-controlled repositories.

    By Graphene-Data
  5. 025Hacker NewsOCT · 01English

    The death of web development education

    Web development educators report that generative AI has devastated their business models and income. Independent trainers, course creators, and technical writers like Baldur Bjarnason, Axel Rauschmayer, and Josh W. Comeau describe plummeting revenue, reduced audience engagement, and unauthorized use of their content by AI systems, forcing many to leave the field or take lower-paid work.

    By Email
  6. 026Hacker NewsOCT · 01English

    AI is not "just a tool"

    The author argues that AI, particularly large language models, should not be dismissed as 'just a tool' because it represents a complex industry with significant environmental, economic, and societal implications, unlike traditional tools such as power drills or text editors. The piece criticizes how this framing obscures the irresponsible practices of AI companies and allows developers to evade accountability for using potentially dangerous products.

    By Muhammad523
  7. 027Hacker NewsOCT · 01English

    Show HN: Janus – Go binary that runs GGUF models via Vulkan on AMD/Intel/Nvidia

    Janus is a standalone Go binary that runs GGUF language models locally via Vulkan on AMD/Intel/Nvidia GPUs or CPU, providing an OpenAI-compatible API and web UI without requiring Python, Docker, or Ollama. It supports hot-swapping models, built-in tools for file operations and OCR, and integrates with OpenAI clients like Cursor and Cline.

    By Vibra-Ingenn
  8. 028Hacker NewsOCT · 01English

    With most information hidden, the game Stratego had stumped AI–until now

    Researchers from Carnegie Mellon, MIT, NYU, and Stanford developed Ataraxos, an AI that defeated top Stratego player Pim Niemeijer 15-1, solving a game that had resisted AI for decades. Stratego's complexity stems from massive hidden information across thousands of moves and the need to balance bluffing, challenges that earlier AI systems like DeepMind's DeepNash could not overcome.

    By Jacek Krywko
  9. 029Hacker NewsOCT · 01English

    TurboNotes: Local-first Markdown notes, Turbo Pascal like looks, saves to GitHub

    TurboNotes is a local-first Markdown note-taking application with live preview, task lists, wiki linking, and offline functionality. It offers optional GitHub integration for version control and features a Turbo Pascal-inspired interface with keyboard shortcuts for common formatting operations.

    By vadepaysa
  10. 030Hacker NewsOCT · 01English

    zkAPI: private usage credits for any API

    zkAPI enables anonymous API payments by separating identity from billing through zero-knowledge proofs. Users deposit credits into an Ethereum vault once, then authorize API usage with cryptographic proofs that reveal neither their identity nor link requests together. Built by the Open Anonymity Project with the Ethereum Foundation, it runs on Ethereum Mainnet today.

    By Vittorio Rivabella
  11. 031Hacker NewsOCT · 01English

    Larceny – a Claude Code plugin that runs a team of agents on GitHub

    Larceny is a Claude Code plugin that orchestrates a team of AI agents to manage software development tasks on GitHub. A coordinator breaks down requests into tickets, assigns them to parallel coding agents, and ensures an adversarial reviewer checks all pull requests before merging. All work is tracked through GitHub issues, branches, and pull requests for full traceability.

    By Nestedmind
  12. 032Hacker NewsOCT · 01English

    A big-tent or small-tent AI safety movement?

    The article argues for a middle ground in AI safety discourse, rejecting both claims that AI existential risk is imminent and accusations that such warnings are insincere hype. It acknowledges legitimate concerns about underinvestment in catastrophic risk preparedness while disagreeing with extending these arguments to AI existential risk, proposing instead that AI amplifies existing systemic risks rather than posing unique existential threats.

    By Arvind Narayanan; Sayash Kapoor
  13. 033Hacker NewsOCT · 01English

    Four days of ricing Fedora without installing Arch

    A developer spent four days configuring Hyprland on Fedora to improve their workflow with keyboard-driven window management and automatic tiling, avoiding over-the-top visual effects while assembling practical components like Waybar, Rofi, and various utility scripts.

    By Hoek
  14. 034Hacker NewsOCT · 01English

    Using Opus 5.5 to discover a new eyewitness record of the dodo

    A researcher used Claude Opus 5.5 to discover a previously-unnoticed 1615 Dutch ship's log describing dodo hunting on Mauritius. The AI agent conducted semantic searches through the GLOBALISE archive of Dutch East India Company records, employing a methodology combining specialist knowledge, historical databases, and frontier model analysis to surface the manuscript from the Wapen van Amsterdam.

    By Benjamin Breen
  15. 035Hacker NewsOCT · 01English

    Claude.dev Blog / technical writing for people building with Claude

    Claude.dev blog shares technical writing and developer tips from Anthropic for building with Claude models. Featured articles cover Claude Code mods, eval design, model optimization, context engineering, and performance improvements across Claude versions.

    By duck
  16. 036Hacker NewsOCT · 01English

    Pi Durable

    Earendil shipped Pi 1.0, a coding agent framework, alongside Pi Durable, an experimental new package designed for long-running, durable agents that can operate anywhere with support for multiple users and catastrophic failure recovery. Pi Durable serves as a harness framework for building agentic applications, featuring flexible storage backends, parallel conversation management, and tool execution environments while maintaining minimalism and malleability principles.

    By paulsmith
  17. 037Hacker NewsOCT · 01English

    Antarctica's winter sea ice peaks at 3rd lowest level on record

    Antarctica's winter sea ice reached its third-lowest maximum on record at 17.59 million square kilometres on 14 September, continuing a trend of exceptionally low peaks with four of the lowest records occurring in the past four years. Scientists attribute the decline to factors including El Niño effects on storm patterns, unusual heatwaves, and changes in atmospheric circulation around the Antarctic region.

    By NH Environment Bureau
  18. 038Hacker NewsOCT · 01English

    The failure modes of Claude Code in a guided 60h project

    A developer used Claude Code to build a semantic fuzzer for Obsidian Sync over 60 hours, progressing through three stages: initial unguided attempts that failed, guided prototyping that succeeded, and polishing that created more problems than solutions. The experience revealed Claude's limitations: it lacks persistent project memory, requires constant correction and guidance, and behaves like a knowledgeable but periodically memory-wiped junior engineer who needs explicit instruction to remain productive.

    By Hmijail
  19. 039Hacker NewsOCT · 01English

    ArXiv's Updated Rate Limit Policy

    arXiv implemented a new rate-limiting policy effective October 1st, 2026, restricting submitters to two papers per month and three active submissions simultaneously. The policy responds to a doubling of submissions in two years, driven largely by AI tools enabling authors to flood the repository with low-quality papers, overwhelming volunteer moderators and delaying legitimate research.

    By Kat Boboris
  20. 040Hacker NewsOCT · 01English

    SvelteKit 3 Is Here

    SvelteKit 3.0 is now available with improved type safety, polish, and a migration tool. Key changes include configuration moving to vite.config.ts, the $lib alias becoming #lib, enhanced environment variables, and improved error handling. The framework also introduces remote functions for secure client-server communication.

    By The Svelte team