<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:media="http://search.yahoo.com/mrss/"><channel><title>LockerGnome: AI</title><description>Today&apos;s outages, hacks, phones, gadgets and fixes, plus the tech you grew up with, picked daily by Chris Pirillo for the pros everybody calls.</description><link>https://lockergnome.com/</link><language>en-us</language><atom:link href="https://lockergnome.com/category/ai/rss.xml" rel="self" type="application/rss+xml"/><item><title>OpenFilm Lets a Coding Agent Make Videos as Web Pages, Then Lets You Edit Every Frame by Hand</title><link>https://lockergnome.com/openfilm-coding-agent-makes-videos-as-web-pages/</link><guid isPermaLink="true">https://lockergnome.com/openfilm-coding-agent-makes-videos-as-web-pages/</guid><description>The MIT-licensed OpenFilm writes a film as one HTML file of animated pages and clips, opens it in a local timeline editor, and exports MP4, GIF, stems or a Final Cut, Premiere or Resolve timeline.</description><pubDate>Thu, 29 Oct 2026 04:10:42 GMT</pubDate><media:content url="https://lockergnome.com/_astro/openfilm-coding-agent-makes-videos-as-web-pages.igzkUh_G_lfciB.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>software</category><category>open-source</category><category>ai-agents</category><category>video-editing</category><category>html</category></item><item><title>Indra, a one-million-parameter text-recognition model, reads scans 15 times faster than Tesseract on a CPU</title><link>https://lockergnome.com/indra-ocr-one-million-parameter-model-15x-faster-than-tesseract-on-cpu/</link><guid isPermaLink="true">https://lockergnome.com/indra-ocr-one-million-parameter-model-15x-faster-than-tesseract-on-cpu/</guid><description>Datalab released Indra OCR, a 1M-parameter model that runs on CPU with no GPU or PyTorch, and its own benchmark puts it at about 44 pages a second on 32 cores, 15 times Tesseract.</description><pubDate>Thu, 29 Oct 2026 04:10:40 GMT</pubDate><media:content url="https://lockergnome.com/_astro/indra-ocr-one-million-parameter-model-15x-faster-than-tesseract-on-cpu.BX4iYJYg_Z4CE9j.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>ocr</category><category>open-source</category><category>ai</category></item><item><title>A $7 ESP32 Board Lets an AI Tap Through Any iPhone App by Posing as an AirPlay Screen and a USB Mouse</title><link>https://lockergnome.com/7-dollar-esp32-board-lets-ai-tap-and-swipe-any-iphone-app-no-computer-no-jailbreak/</link><guid isPermaLink="true">https://lockergnome.com/7-dollar-esp32-board-lets-ai-tap-and-swipe-any-iphone-app-no-computer-no-jailbreak/</guid><description>Henry Yuan&apos;s open-source ESP32 iPhone Agent plugs a $7 Seeed XIAO ESP32S3 into a stock iPhone: the phone mirrors its screen to the board, an app on the same phone asks a vision model what to do, and the board taps through AssistiveTouch, with no computer and no jailbreak.</description><pubDate>Thu, 29 Oct 2026 04:10:37 GMT</pubDate><media:content url="https://lockergnome.com/_astro/7-dollar-esp32-board-lets-ai-tap-and-swipe-any-iphone-app-no-computer-no-jailbreak.BAAEd4vh_fmxyy.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>iphone</category><category>ai-agents</category><category>esp32</category><category>open-source</category><category>automation</category></item><item><title>Parley Puts Claude Code, Codex and OpenCode Sessions in One Local IRC Channel, With a Breaker That Stops Them Talking in Circles</title><link>https://lockergnome.com/parley-coding-agent-sessions-irc-channel-breaker/</link><guid isPermaLink="true">https://lockergnome.com/parley-coding-agent-sessions-irc-channel-breaker/</guid><description>An open-source Go tool runs a local IRC server, pushes every message into each coding agent&apos;s own input path, and stops waking them when they chatter without you.</description><pubDate>Wed, 28 Oct 2026 23:13:51 GMT</pubDate><media:content url="https://lockergnome.com/_astro/parley-coding-agent-sessions-irc-channel-breaker.DPiY4gV__1H7NhF.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>agents</category><category>ai</category><category>irc</category><category>open-source</category><category>claude-code</category></item><item><title>Survival LLM Puts 930,000 Wikipedia Articles, a Medical Encyclopedia and Military Survival Manuals in a Free 9 GB Offline AI That Picks Its Model by Your RAM</title><link>https://lockergnome.com/survival-llm-offline-930000-wikipedia-articles-9-gb/</link><guid isPermaLink="true">https://lockergnome.com/survival-llm-offline-930000-wikipedia-articles-9-gb/</guid><description>The free, open-source app runs a local AI model over offline Wikipedia, WikiMed, U.S. military medicine and survival manuals and iFixit guides, and its makers measured about 100 W per answer on a MacBook Pro, roughly 100 to 300 questions per battery.</description><pubDate>Wed, 28 Oct 2026 23:13:51 GMT</pubDate><media:content url="https://lockergnome.com/_astro/survival-llm-offline-930000-wikipedia-articles-9-gb.QNGrjKsS_28kKdd.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>ai</category><category>offline</category><category>local-llm</category><category>open-source</category><category>preparedness</category></item><item><title>Wenlan Turns Your Notes and AI Chats Into a Local Wiki, and Throws Out Any Page Whose Citations Don&apos;t Match the Source</title><link>https://lockergnome.com/wenlan-living-wiki-refuses-pages-citations-dont-match/</link><guid isPermaLink="true">https://lockergnome.com/wenlan-living-wiki-refuses-pages-citations-dont-match/</guid><description>The open-source desktop app and CLI has the AI tool you already use draft cited wiki pages from your files and chat history, then refuses a page when most of its citations fail to match what they point at.</description><pubDate>Wed, 28 Oct 2026 23:13:51 GMT</pubDate><media:content url="https://lockergnome.com/_astro/wenlan-living-wiki-refuses-pages-citations-dont-match.BPLMaLXE_Z2eXojD.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>ai</category><category>notes</category><category>open-source</category><category>local-first</category><category>obsidian</category></item><item><title>Wikipedia&apos;s AI Cleanup Volunteers Hunt &apos;Stands as a Testament&apos; Prose as Pangram Flags About 20% of New Articles as AI-Written</title><link>https://lockergnome.com/wikipedia-ai-cleanup-pangram-20-percent-new-articles-ai/</link><guid isPermaLink="true">https://lockergnome.com/wikipedia-ai-cleanup-pangram-20-percent-new-articles-ai/</guid><description>WikiProject AI Cleanup, started by editors in December 2023, tracks the phrases and habits chatbots leave behind, and cofounder Ellia Lebleu says detector scans put the share of AI-written new articles at about a fifth in late 2025.</description><pubDate>Wed, 28 Oct 2026 23:13:51 GMT</pubDate><media:content url="https://lockergnome.com/_astro/wikipedia-ai-cleanup-pangram-20-percent-new-articles-ai.CK045WNJ_ZDmHUx.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>wikipedia</category><category>ai-writing</category><category>moderation</category><category>llm</category></item><item><title>KitCut moved its AI film studio off a $250-a-month Azure VM onto a used gaming PC, and the same 8-minute film renders in 3m49s instead of 17m42s</title><link>https://lockergnome.com/kitcut-ai-film-studio-used-gaming-pc/</link><guid isPermaLink="true">https://lockergnome.com/kitcut-ai-film-studio-used-gaming-pc/</guid><description>KitCut&apos;s maker wiped a second-hand gaming PC and installed Linux over SSH with no USB stick and no spare machine, then watched a box on the living-room floor out-render the cloud VM it replaced by 4.6x.</description><pubDate>Tue, 27 Oct 2026 23:13:29 GMT</pubDate><media:content url="https://lockergnome.com/_astro/kitcut-ai-film-studio-used-gaming-pc.CbZ61mq9_rxW0M.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>ai-video</category><category>self-hosting</category><category>homelab</category><category>cloud-costs</category><category>claude-code</category></item><item><title>Last Commit writes Día de los Muertos epitaphs for dead GitHub repos with a small Qwen model running in your browser</title><link>https://lockergnome.com/last-commit-ofrenda-browser-qwen-epitaphs-dead-github-repos/</link><guid isPermaLink="true">https://lockergnome.com/last-commit-ofrenda-browser-qwen-epitaphs-dead-github-repos/</guid><description>Joe Simo&apos;s Last Commit turns a quiet GitHub repository into an ofrenda card, and it throws out the local model&apos;s epitaph whenever the model invents a number or slips into first person.</description><pubDate>Tue, 27 Oct 2026 23:13:28 GMT</pubDate><media:content url="https://lockergnome.com/_astro/last-commit-ofrenda-browser-qwen-epitaphs-dead-github-repos.fZrlUvOr_Z10899O.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>open-source</category><category>ai</category><category>github</category><category>webgpu</category></item><item><title>Nietzschian Debugger Runs on Claude and Refuses to Fix Your Bug, Asking Questions at Three Settings From Socrates to Zarathustra</title><link>https://lockergnome.com/nietzschian-debugger-claude-cli-only-asks-questions-never-fixes/</link><guid isPermaLink="true">https://lockergnome.com/nietzschian-debugger-claude-cli-only-asks-questions-never-fixes/</guid><description>A philosophy graduate turned developer in Lahore built a command-line debugger that only ever answers with a question, with a model-judge checking every reply so it never slips and hands over the fix.</description><pubDate>Tue, 27 Oct 2026 23:13:27 GMT</pubDate><media:content url="https://lockergnome.com/_astro/nietzschian-debugger-claude-cli-only-asks-questions-never-fixes.BeYXnA48_Z1iUhhn.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>ai</category><category>claude</category><category>debugging</category><category>cli</category><category>python</category><category>open-source</category></item><item><title>Denis Zontak&apos;s One-Mac Software Factory Turned a GitHub Ticket Into a Merged Pull Request Overnight on a $200-a-Month AI Subscription</title><link>https://lockergnome.com/denis-zontak-one-mac-software-factory-overnight-pull-request/</link><guid isPermaLink="true">https://lockergnome.com/denis-zontak-one-mac-software-factory-overnight-pull-request/</guid><description>Unable to sleep, a Florida developer watched Claude-driven agents plan, build, test and review a ticket for his time-tracking app, and he merged their first pull request at 2:36 a.m.</description><pubDate>Tue, 27 Oct 2026 14:12:39 GMT</pubDate><media:content url="https://lockergnome.com/_astro/denis-zontak-one-mac-software-factory-overnight-pull-request.jMsQ3a4b_vEo4U.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>ai</category><category>agents</category><category>claude</category><category>github</category><category>automation</category></item><item><title>In Murmell&apos;s Test of AVP, a Git Replacement for Coding Agents, a Claude Code Agent Read a Rejected Push as Success and Four Tasks Never Reached Main</title><link>https://lockergnome.com/avp-murmell-agent-misread-rejected-git-push-four-tasks-lost/</link><guid isPermaLink="true">https://lockergnome.com/avp-murmell-agent-misread-rejected-git-push-four-tasks-lost/</guid><description>Murmell, a two-person startup running many coding agents on one monorepo, open-sourced AVP, which proves every agent&apos;s change on the exact head of main before it lands; its pull-request CI runs fell from 104 a day to 4.</description><pubDate>Tue, 27 Oct 2026 14:11:39 GMT</pubDate><media:content url="https://lockergnome.com/_astro/avp-murmell-agent-misread-rejected-git-push-four-tasks-lost.DZNFqDRd_OW2Tc.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>agents</category><category>git</category><category>ci</category><category>version control</category><category>open source</category></item><item><title>proc, a Free Process Tool, Now Has an Agent Skill That Stops Claude Code and Codex From Piping lsof Into kill -9 on Processes They Didn&apos;t Start</title><link>https://lockergnome.com/proc-cli-agent-skill-stops-lsof-kill-9/</link><guid isPermaLink="true">https://lockergnome.com/proc-cli-agent-skill-stops-lsof-kill-9/</guid><description>proc&apos;s new agent skill tells coding agents to find a server by port or name with one JSON command, preview with --dry-run, and touch only the processes they started or you named.</description><pubDate>Tue, 27 Oct 2026 14:10:39 GMT</pubDate><media:content url="https://lockergnome.com/_astro/proc-cli-agent-skill-stops-lsof-kill-9.6KrQJmPH_1OXyv1.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>agents</category><category>cli</category><category>open source</category><category>macos</category><category>linux</category></item><item><title>Ubicloud Traced Kimi&apos;s Tool-Call Accuracy Drop to 91% Back to a vLLM Default Change</title><link>https://lockergnome.com/ubicloud-vllm-default-change-tool-call-accuracy-91-percent/</link><guid isPermaLink="true">https://lockergnome.com/ubicloud-vllm-default-change-tool-call-accuracy-91-percent/</guid><description>Ubicloud found that vLLM had stopped applying constrained decoding to tool calls unless a request asked for strict mode, and turned it back on server-side to get Kimi and DeepSeek to 100% schema accuracy.</description><pubDate>Tue, 27 Oct 2026 14:01:39 GMT</pubDate><media:content url="https://lockergnome.com/_astro/ubicloud-vllm-default-change-tool-call-accuracy-91-percent.BV6do4A7_Z1bX6nc.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>vllm</category><category>inference</category><category>ai</category><category>open-weight-models</category></item><item><title>Marco Franzon&apos;s print3d Has an AI Agent Measure a Clothes Peg From a Museum Photo and Hand You a Bambu Lab Print File</title><link>https://lockergnome.com/print3d-agent-measures-museum-clothes-peg-photo-bambu-lab-print/</link><guid isPermaLink="true">https://lockergnome.com/print3d-agent-measures-museum-clothes-peg-photo-bambu-lab-print/</guid><description>print3d, an open-source plugin for AI coding agents, scales a photo from a ruler or bank card beside the object, writes a Blender script, checks the mesh against the photo and slices it in Bambu Studio until every check passes.</description><pubDate>Mon, 26 Oct 2026 14:13:00 GMT</pubDate><media:content url="https://lockergnome.com/_astro/print3d-agent-measures-museum-clothes-peg-photo-bambu-lab-print.DdFF1sIx_Z2klDcL.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>3d-printing</category><category>agents</category><category>ai</category><category>open-source</category></item><item><title>A Taylor &amp; Francis paper on Z-Library&apos;s copyright ethics cited a TorrentFreak article TorrentFreak never wrote</title><link>https://lockergnome.com/z-library-ethics-paper-cites-torrentfreak-article-that-doesnt-exist/</link><guid isPermaLink="true">https://lockergnome.com/z-library-ethics-paper-cites-torrentfreak-article-that-doesnt-exist/</guid><description>A peer-reviewed paper on Z-Library, written with help from Consensus AI, credits a 2021 TorrentFreak story to the editor who then went looking for it and found it never existed. At least ten of its 44 references couldn&apos;t be found, and Taylor &amp; Francis has sent it to its ethics team.</description><pubDate>Mon, 26 Oct 2026 14:13:00 GMT</pubDate><media:content url="https://lockergnome.com/_astro/z-library-ethics-paper-cites-torrentfreak-article-that-doesnt-exist.D-QZ3XvA_Z2tP1B6.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>ai</category><category>academic publishing</category><category>z-library</category><category>citations</category><category>research integrity</category></item><item><title>Latch is one script tag that hands AI browsers your site&apos;s search box, cart button and forms as WebMCP tools</title><link>https://lockergnome.com/latch-one-script-tag-webmcp-ai-browsers/</link><guid isPermaLink="true">https://lockergnome.com/latch-one-script-tag-webmcp-ai-browsers/</guid><description>Paste one line and Latch finds your search box, buy buttons, forms and nav, then offers them to AI browsers as WebMCP tools they can call. About 6 KB gzipped, MIT, and it does nothing at all in a browser without WebMCP.</description><pubDate>Sun, 25 Oct 2026 20:52:07 GMT</pubDate><media:content url="https://lockergnome.com/_astro/latch-one-script-tag-webmcp-ai-browsers.CXBlNIRA_23HsJg.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>ai agents</category><category>webmcp</category><category>javascript</category><category>open source</category></item><item><title>Koala runs each AI agent in its own Linux VM on a Mac, and the API key never goes inside</title><link>https://lockergnome.com/koala-agents-in-isolated-linux-vms-apple-silicon/</link><guid isPermaLink="true">https://lockergnome.com/koala-agents-in-isolated-linux-vms-apple-silicon/</guid><description>Kunal Kushwaha&apos;s Koala gives every agent its own ARM64 kernel, disk and network on Apple silicon, keeps its API keys in the Mac&apos;s login keychain, and lets a broker add them only to requests a policy allows.</description><pubDate>Sun, 25 Oct 2026 20:52:04 GMT</pubDate><media:content url="https://lockergnome.com/_astro/koala-agents-in-isolated-linux-vms-apple-silicon.C5A6EFfN_ZCQrBa.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>ai-agents</category><category>security</category><category>macos</category><category>virtual-machines</category></item><item><title>Prateek Chauhan Turned a 2015 HP Laptop Into a Homelab in Two Days With Claude Code, Four Rules and a Progress File It Read After Every Reboot</title><link>https://lockergnome.com/2015-hp-laptop-homelab-claude-code-progress-md/</link><guid isPermaLink="true">https://lockergnome.com/2015-hp-laptop-homelab-claude-code-progress-md/</guid><description>Four Ubuntu upgrades, one interrupted hop that left 1,337 packages ahead of systemd, a security claim the agent got wrong, and a local RAG server: a dusty Pavilion now runs with its lid closed and Windows one boot-menu choice away.</description><pubDate>Sun, 25 Oct 2026 20:52:02 GMT</pubDate><media:content url="https://lockergnome.com/_astro/2015-hp-laptop-homelab-claude-code-progress-md.cm_iAMJR_ZUOhez.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>homelab</category><category>linux</category><category>ai-agents</category><category>claude-code</category></item><item><title>Ponytail, a free &apos;laziest senior dev&apos; prompt for Claude Code and other coding agents, cut a 335-line date picker to 10 lines by using the browser&apos;s own calendar</title><link>https://lockergnome.com/ponytail-laziest-senior-dev-skill-coding-agents-date-picker/</link><guid isPermaLink="true">https://lockergnome.com/ponytail-laziest-senior-dev-skill-coding-agents-date-picker/</guid><description>Dietrich Gebert&apos;s Ponytail is one prompt file that makes a coding agent check what already exists before writing anything. His 585-session benchmark says it cut code 53% and cost 26%, and admits where it lost.</description><pubDate>Sat, 24 Oct 2026 02:12:00 GMT</pubDate><media:content url="https://lockergnome.com/_astro/ponytail-laziest-senior-dev-skill-coding-agents-date-picker.BxtyXetJ_Z1ErRbA.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>ai</category><category>agents</category><category>open-source</category></item><item><title>Asked to hide a secret in a picture, Claude Fable 5.1 makes each building&apos;s height a letter</title><link>https://lockergnome.com/claude-fable-hides-messages-building-heights/</link><guid isPermaLink="true">https://lockergnome.com/claude-fable-hides-messages-building-heights/</guid><description>Researchers asked Fable 5.1 to hide 50 messages in drawings. It mostly spelled them out in skyline heights, and a fresh session with no codebook read 39 of them back from screenshots.</description><pubDate>Fri, 23 Oct 2026 17:12:00 GMT</pubDate><media:content url="https://lockergnome.com/og/claude-fable-hides-messages-building-heights.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>security</category><category>research</category></item><item><title>Cohere gave a GPT-5.6 coding agent 16 cheap helpers, and they only paid off as planners</title><link>https://lockergnome.com/cheap-coding-agents-make-good-planners-and-bad-coders/</link><guid isPermaLink="true">https://lockergnome.com/cheap-coding-agents-make-good-planners-and-bad-coders/</guid><description>Cohere&apos;s researchers found cheap GPT-5.6 Luna workers help a coding agent when they scout and plan, and drag it below a solo agent when they write the code.</description><pubDate>Fri, 23 Oct 2026 14:32:00 GMT</pubDate><media:content url="https://lockergnome.com/og/cheap-coding-agents-make-good-planners-and-bad-coders.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>agents</category><category>programming</category></item><item><title>People are pointing Claude Code at their smart TVs to switch the ads off, and the trick has spread from Android TV to Fire TV and LG</title><link>https://lockergnome.com/vibe-coding-the-ads-off-smart-tvs/</link><guid isPermaLink="true">https://lockergnome.com/vibe-coding-the-ads-off-smart-tvs/</guid><description>A coding agent, the debug port the TV shipped with, and one afternoon: the smart TV ad rebellion now has its own repos for Android TV, Fire TV and LG webOS.</description><pubDate>Fri, 23 Oct 2026 14:27:00 GMT</pubDate><media:content url="https://lockergnome.com/og/vibe-coding-the-ads-off-smart-tvs.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>android</category><category>smart-home</category></item><item><title>tldraw&apos;s AI turns pull requests into narrated videos, and HyperFrames, Framefields and Remotion want to render yours</title><link>https://lockergnome.com/tldraw-ai-narrates-pull-requests-video/</link><guid isPermaLink="true">https://lockergnome.com/tldraw-ai-narrates-pull-requests-video/</guid><description>A skill in tldraw&apos;s repo reads a pull request, writes the author&apos;s walkthrough, voices it and renders an MP4. Three open engines now compete to be the renderer.</description><pubDate>Fri, 23 Oct 2026 14:25:00 GMT</pubDate><media:content url="https://lockergnome.com/og/tldraw-ai-narrates-pull-requests-video.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>agents</category><category>video</category><category>open source</category></item><item><title>Taskman names every Claude Code session after its ticket, so agents can tell each other their tests failed</title><link>https://lockergnome.com/taskman-agent-sessions-beside-prs/</link><guid isPermaLink="true">https://lockergnome.com/taskman-agent-sessions-beside-prs/</guid><description>Taskman puts each Claude Code or Codex session beside its issue and pull request, flags the one waiting on you, and lets a coordinating agent message the others by ticket key.</description><pubDate>Fri, 23 Oct 2026 14:23:00 GMT</pubDate><media:content url="https://lockergnome.com/og/taskman-agent-sessions-beside-prs.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>agents</category><category>programming</category></item><item><title>Coucou lets your coding agents live in the Mac&apos;s notch, where a blob named Mochi asks for your OK</title><link>https://lockergnome.com/coucou-mac-notch-agents/</link><guid isPermaLink="true">https://lockergnome.com/coucou-mac-notch-agents/</guid><description>Coucou is a free, open-source Mac app that puts Claude Code, Codex and other coding agents in the notch, with Allow, Deny and Always one click away.</description><pubDate>Fri, 23 Oct 2026 14:22:00 GMT</pubDate><media:content url="https://lockergnome.com/og/coucou-mac-notch-agents.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>apple</category><category>ai</category><category>agents</category></item><item><title>buddylist turns Claude Code, Codex and Gemini into MSN Messenger buddies, and a nudge stops them mid-task</title><link>https://lockergnome.com/buddylist-msn-messenger-coding-agents/</link><guid isPermaLink="true">https://lockergnome.com/buddylist-msn-messenger-coding-agents/</guid><description>buddylist is a free desktop app that puts Claude Code, Codex, Gemini, OpenCode and Hermes in an MSN Messenger 7.5 buddy list, where a nudge shakes the window and stops the agent.</description><pubDate>Fri, 23 Oct 2026 14:21:00 GMT</pubDate><media:content url="https://lockergnome.com/_astro/buddylist-msn-messenger-coding-agents.DvBQ9qVe_2ji4QF.jpg" medium="image" type="image/jpeg" width="1200" height="630"/><category>AI</category><category>ai</category><category>agents</category><category>retro</category></item><item><title>Beam walks one window off your agents&apos; Mac mini and leaves a note saying where it went</title><link>https://lockergnome.com/beam-win-windows-between-macs/</link><guid isPermaLink="true">https://lockergnome.com/beam-win-windows-between-macs/</guid><description>Beam pulls a single app window from a headless Mac mini, the one running your builds and agents, onto the Mac at your desk as a native window with its own Dock icon.</description><pubDate>Fri, 23 Oct 2026 14:20:00 GMT</pubDate><media:content url="https://lockergnome.com/og/beam-win-windows-between-macs.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>apple</category><category>ai</category></item><item><title>Google&apos;s Playground lets anyone make a game, one token at a time</title><link>https://lockergnome.com/google-playground-make-a-game/</link><guid isPermaLink="true">https://lockergnome.com/google-playground-make-a-game/</guid><description>Google opened Playground on October 7, an experimental site where you describe a game and its AI builds it.</description><pubDate>Fri, 23 Oct 2026 02:21:00 GMT</pubDate><media:content url="https://lockergnome.com/og/google-playground-make-a-game.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>google</category><category>ai</category><category>gaming</category></item><item><title>Meta opened up Muse Spark, and it takes two Mac Studios to run it at home</title><link>https://lockergnome.com/meta-muse-spark-two-mac-studios/</link><guid isPermaLink="true">https://lockergnome.com/meta-muse-spark-two-mac-studios/</guid><description>Meta released Muse Spark Open on October 15, a model with 816 billion parameters, 42 billion of them active.</description><pubDate>Fri, 23 Oct 2026 02:19:00 GMT</pubDate><media:content url="https://lockergnome.com/og/meta-muse-spark-two-mac-studios.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>meta</category><category>ai</category><category>apple</category></item><item><title>Four Gemini agents beat Minecraft on their own, and died 605 times doing it</title><link>https://lockergnome.com/gemini-agents-beat-minecraft/</link><guid isPermaLink="true">https://lockergnome.com/gemini-agents-beat-minecraft/</guid><description>Four AI agents just beat Minecraft with nobody steering.</description><pubDate>Fri, 23 Oct 2026 02:11:00 GMT</pubDate><media:content url="https://lockergnome.com/og/gemini-agents-beat-minecraft.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>google</category><category>gaming</category></item><item><title>Somebody built a programming language whose whole job is to break AI models</title><link>https://lockergnome.com/language-built-to-break-ai-models/</link><guid isPermaLink="true">https://lockergnome.com/language-built-to-break-ai-models/</guid><description>Dominik Picheta built a programming language that coding AIs aren&apos;t supposed to be able to use, and he named it KillSwitch.</description><pubDate>Fri, 23 Oct 2026 02:10:00 GMT</pubDate><media:content url="https://lockergnome.com/og/language-built-to-break-ai-models.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>programming</category></item><item><title>One AI assistant sent 187 messages chasing a refund it couldn&apos;t get</title><link>https://lockergnome.com/ai-assistant-187-messages-refund/</link><guid isPermaLink="true">https://lockergnome.com/ai-assistant-187-messages-refund/</guid><description>Vals AI gave six assistants a £350 train refund to chase after the deadline had passed.</description><pubDate>Fri, 23 Oct 2026 02:03:00 GMT</pubDate><media:content url="https://lockergnome.com/og/ai-assistant-187-messages-refund.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>agents</category></item><item><title>Apple Intelligence is coming to China without the new Siri</title><link>https://lockergnome.com/apple-intelligence-china-without-siri/</link><guid isPermaLink="true">https://lockergnome.com/apple-intelligence-china-without-siri/</guid><description>Tim Cook told The Paper, a state-backed Chinese outlet, that Apple Intelligence has been approved in China and that a test version is due later in October, Reuters reported.</description><pubDate>Fri, 23 Oct 2026 01:04:00 GMT</pubDate><media:content url="https://lockergnome.com/og/apple-intelligence-china-without-siri.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>apple</category><category>ai</category><category>china</category></item><item><title>Someone trained a tiny AI to pretend it is a terminal</title><link>https://lockergnome.com/worldtui-tiny-ai-terminal/</link><guid isPermaLink="true">https://lockergnome.com/worldtui-tiny-ai-terminal/</guid><description>Simon Edwardsson built a terminal with nothing behind it.</description><pubDate>Thu, 22 Oct 2026 18:00:00 GMT</pubDate><media:content url="https://lockergnome.com/og/worldtui-tiny-ai-terminal.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category></item><item><title>AI data centers now need their own special gray paint</title><link>https://lockergnome.com/ai-data-center-gray-paint/</link><guid isPermaLink="true">https://lockergnome.com/ai-data-center-gray-paint/</guid><description>The AI boom has reached the paint aisle.</description><pubDate>Thu, 22 Oct 2026 16:00:00 GMT</pubDate><media:content url="https://lockergnome.com/og/ai-data-center-gray-paint.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>data-centers</category></item><item><title>Penmentor puts AI math hints beside your handwriting</title><link>https://lockergnome.com/penmentor-ai-math-hints-handwriting/</link><guid isPermaLink="true">https://lockergnome.com/penmentor-ai-math-hints-handwriting/</guid><description>Penmentor describes an AI math helper that marks handwritten work in Xournal++ with colored pens.</description><pubDate>Tue, 20 Oct 2026 20:00:00 GMT</pubDate><media:content url="https://lockergnome.com/og/penmentor-ai-math-hints-handwriting.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>education</category></item><item><title>A startup says its AI passed as human in 48% of one-minute video calls, by its own test</title><link>https://lockergnome.com/ai-passed-as-human-video-calls/</link><guid isPermaLink="true">https://lockergnome.com/ai-passed-as-human-video-calls/</guid><description>Tavus, a San Francisco AI startup, says 26 of the 54 people who spent a minute on a video call with its new Griffin-Lite model believed afterward they&apos;d been talking to a real person.</description><pubDate>Fri, 09 Oct 2026 18:00:00 GMT</pubDate><media:content url="https://lockergnome.com/og/ai-passed-as-human-video-calls.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>deepfakes</category></item><item><title>Anthropic has been asking religious thinkers how Claude should behave</title><link>https://lockergnome.com/anthropic-religious-thinkers-claude/</link><guid isPermaLink="true">https://lockergnome.com/anthropic-religious-thinkers-claude/</guid><description>For months, Anthropic has hosted private meetings with dozens of religious scholars to help determine how to train its AI models to behave virtuously, The New York Times reported on Sept. 30.</description><pubDate>Wed, 07 Oct 2026 01:01:00 GMT</pubDate><media:content url="https://lockergnome.com/og/anthropic-religious-thinkers-claude.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>anthropic</category></item><item><title>An AI radio host in Los Angeles sometimes has to be told to stop swearing</title><link>https://lockergnome.com/ai-radio-host-stop-swearing/</link><guid isPermaLink="true">https://lockergnome.com/ai-radio-host-stop-swearing/</guid><description>José FM in Los Angeles has a morning co-host named Coyotec, and he&apos;s a chatbot.</description><pubDate>Tue, 06 Oct 2026 03:40:00 GMT</pubDate><media:content url="https://lockergnome.com/og/ai-radio-host-stop-swearing.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>radio</category></item><item><title>Trump ordered federal agencies to say &quot;Super Intelligence&quot; instead of &quot;AI&quot;</title><link>https://lockergnome.com/super-intelligence-instead-of-ai/</link><guid isPermaLink="true">https://lockergnome.com/super-intelligence-instead-of-ai/</guid><description>On Sept. 29, President Trump signed an executive order telling federal agencies to use &quot;Super Intelligence&quot; and &quot;SI&quot; in official correspondence, public communications, websites, reports and policy documents, and to &quot;no longer acknowledge&quot; the terms &quot;artificial intelligence&quot; and &quot;AI&quot;.</description><pubDate>Mon, 05 Oct 2026 17:04:00 GMT</pubDate><media:content url="https://lockergnome.com/og/super-intelligence-instead-of-ai.png" medium="image" type="image/png" width="1200" height="630"/><category>AI</category><category>ai</category><category>government</category></item></channel></rss>