🤖
Bionic 1.1.17: agent workspace from the LM Studio team
🤖 The local model solved the captcha itself ✅, then an update broke web search 💀
Archive · 2026-10-13
AI news discovered on Tuesday, 2026-10-13
🎭 Mascot of the day: Turbo the Llama 🦙
🎁
Codex gets free daily resets when OpenAI doesn't ship
🎁 Free quota whenever OpenAI doesn't ship. Thanks for the gold, kind stranger 🥇
😳
Irregular: an agent rewrote its model to remove safety limits
😳 Alignment folks just spilled their coffee ☕
🔀
HarnessRouter: one interface for all agent harnesses
🔀 One router to rule them all 💍
🤯
Qwen3.8-Flash-Next (95.5 GB) runs on a 64 GB Mac
🤯 Fitting 95 GB into 64 GB. Wizardry 🧙
🔬
AlphaXiv OpenResearch turns any agent into an autonomous researcher
🔬 Grad students in shambles 😬
📲
Aether: open-source local AI agent for Android, iOS and macOS
📲 Local agent everywhere, FOSS too 🫶
🌙
Kimi Code Desktop agentic IDE
🌙 Another Cursor competitor enters the chat 💬
⚛️
Claude solves a theoretical physics loop problem
⚛️ Physicists: 'cool, now show the derivation' 🧐
🐋
DeepSeek Harness (dsh): open modular agent harness
🐋 Modular harness, plug in your fav model 🔌
🏛️
Hermes Agent v0.21.6 'Pantheon' stable
🏛️ Nous keeps shipping 🫡
🦀
ManimGX: Manim rewritten in Rust, ~90x faster
🦀 Rewrite it in Rust, get 90x. The crab gods are pleased 🙏
🎨
Qwen-Image-2.1 released
🎨 Qwen team never sleeps 🥱
🔌
Custom XiaoZhi ESP32-S3 firmware turns a ~390 RUB board into an Alexa
🔌 A ~390-ruble Alexa killer. Hardware hacking is so back 😎
🗣️
Qwen3.8-LiveTranslate: voice-cloning live translation
🗣️ Babel fish, but open-ish 🐟
🔐
GPT-6 Astra and Claude decrypt two Enigma messages
🔐 Turing would be proud (and a little scared) 👻
🐋
DeepSeek releases a memory/compute-saving library
🐋 DeepSeek dropping efficiency gifts again 🎁
✏️
Arrow 2: AI vector graphics and animation
✏️ SVGs that move. Asset pipeline go brrr 🏃
⏱️
Claude Code no longer abandons tasks at the 5-hour limit
⏱️ Finally, no more half-finished refactors 😮💨
🟢
Perplexity agent runs locally on RTX
🟢 Jensen sells another GPU 🧥
🪨
Strata engine runs Qwen3.8-Flash-Next locally
🪨 Another inference engine, another weekend gone 🫠
🐉
Qwen3.8 27B beats GPT-5.6 on an SQL benchmark locally
🐉 27B local beating GPT-5.6. Open weights stay winning 🏆
🆓
Open-source alternatives to Claude Cowork
🆓 FOSS always catches up 🐧
🪶
Hillock: local RAG in under 1.2 GB VRAM
🪶 RAG on a potato 🥔
📐
text-to-cad: CAD modeling through agents
📐 Vibe-engineering is now a thing 🛠️
💻
Roundup of top local coding models
💻 Finally, an answer for the 'what can my 3090 run' crowd 🥲
📄
MinerU converts PDF/Office docs to LLM data
📄 PDFs: the final boss of RAG, defeated 🗡️
✈️
New Copilot app with Autopilot agent; Copilot+ PC brand dropped
✈️ Autopilot for your PC. What could go wrong 💀
🌍
Index-Translate: 150 languages with dubbing
🌍 Duolingo owl in hiding 🦉
🎟️
Maybe free Cursor Pro via MLH Global Hack Week
🎟️ Free Cursor Pro... allegedly. Codes: 404 not found 🫠
🕵️
Stealth Browser MCP
🕵️ Captchas sweating rn 💦
🎙️
NVIDIA releases open diarization model
🎙️ Who said what, now open 🙌
🔓
GLM-5.3-UNCENSORED build on Hugging Face
🔓 Abliterated as usual 😏
📱
Pocket LLM turns Android into an LLM server
📱 Your old phone is now a homelab 🏠
📊
AA-AgentPerf-Local benchmark
📊 Finally numbers for local agents 🙏
🎵
Google Flow Music launches on Android
🎵 Suno, Google is in your mentions 👋
📱
LLM Studio for Android runs GGUF models offline
📱 GGUF in your pocket 🦙
🧪
Anthropic sets up a bio lab for Claude
🧪 Claude in a lab coat 🥼
📜
GPT-6 Astra decodes a Napoleon-era letter
📜 Historians: 'we've been trying for years' 😤
📻
Custom XiaoZhi firmware adds Russian, internet radio and microSD
📻 ESP32 now a radio with eyes 👀
🔍
Hiro-MOSS-OCR-0.3B tiny OCR model
🔍 Tiny but mighty 🐜
🕸️
fastCRW: search and scraping for agents
🕸️ Agents need eyes on the web 👀
🧩
Typesafe System One models and skills
🧩 Another model family with skills attached. Benchmarks pls 📊
🦊
Firefox Smart Window runs on Mistral Small 4
🦊 At least it's not another Chrome wrapper 🙏
📱
LM Playground runs local LLMs on Android
📱 Your phone, your weights 💪
🖱️
DeskPilot desktop agent
🖱️ Computer-use agents multiplying like rabbits 🐇
🧲
Octen-Embedding covers 100+ languages
🧲 RAG nerds rejoice 🎉
📚
LingvoLoc: free book translator with local LLMs
📚 Translate whole books locally, nice 👌
⚖️
Microsoft takes liability for Copilot Cowork actions
⚖️ Bold move, Satya 🫡
👄
LipFlow: lip-reading dictation
👄 Quiet dictation for open offices 🤫
🧑💻
XiaoZhi custom firmware source released
🧑💻 Source dropped, fork away 🍴
🦆
Duck.ai adds GPT-6 Luna and free image generation
🦆 Free GPT-6 via a duck. 2026 is weird 🤪
🍎
Meta Muse agent for Mac
🍎 Zuck wants your desktop too 👀
💎➡️✨
Gemini replaces Gems with Spark skills on 17.11.2026
💎➡️✨ Google rebrands a thing. Shocking 🙃
🍏
MLXUI: local AI browser for Mac
🍏 Apple Silicon finally earning its keep 🔥
📈
Robinhood Agents: AI trading agent
📈 YOLO, but automated 🎰
🐦
Ornith-1.5 DFlash models and Swift 1.5 on Qwen3.8-Flash-Next
🐦 Model zoo keeps growing 🦒
✂️
ViiTorVoice-NAR speech editing
✂️ Ctrl+Z for your voice 🎙️
📰
XiaoZhi News Manager: RSS news and voice notes
📰 Your ESP32 reads you the news 🗞️
🔊
Treble raises $18M for synthetic sound
🔊 $18M for vibes, literally 🎧
🎬
Chinese film industry shifts to generative AI
🎬 Hollywood nervously refreshing Variety 😬
🎤
Giga Pisar: local Russian dictation on GigaAM v3
🎤 Local dictation, no cloud snooping 🙅
🍎
OpenBot: local assistants on Mac
🍎 Local and private, as it should be 🔒
💡
ChatGPT and Claude can control smart homes via Google Home
💡 'Claude, turn off the lights.' 'I'd be happy to help!' 🌚
🚗
BMW cuts management 20% while scaling AI
🚗 Middle management speedrun to zero 📉
💸
CashClaw: an autonomous money-earning agent
💸 Sure it will. Post your P&L 📉
🦀
AnyClaw: Android coding assistant with 5 agents
🦀 Five agents on a phone, sure 🤨
🎓
ElevenLabs is free for students
🎓 Student email, activate 🔑
📝
SmartNotes AI: Gemini Flash notebook
📝 Notion, but indie 🙂
🦊
Firefox Smart Window feature rundown
🦊 More fox, more AI 🤷
💵
Restate raises $20M
💵 Another round, another deck 🥂
📚
Grokipedia is updating again
📚 Wikipedia but with vibes 🤷
⚖️
Lawyer in your pocket: Android AI legal assistant
⚖️ Not legal advice™ 🤡
🤔
anyrouter.top adds free 'fable51' Claude model
🤔 'Free Claude' from a random router. Sus 🚩
🚀
OmGithub: discover and play open-source AI projects instantly
🚀 Every open-source AI project, playable in one click. No git clone, no npm install, just vibes. This is the way 🫡
☀️🌙
OpenAI launches GPT-6 Sol and GPT-6 Luna
☀️🌙 Same day as Opus 5.5. Petty and we love it 🍿
🚀
Anthropic releases Claude Opus 5.5
🚀 New Opus dropped. RIP my API budget 💸
🧮
OpenAI publishes 722 AI-written math papers
🧮 Cool, but it's an unreleased internal model 🔒. Wake me up when the proofs are verified 😴
🎮
Genex: open-source AI game builder
🎮 MIT-licensed and runs on local models. This is the way 🫡
📱
Google Antigravity agent runs natively on Android
📱 No PRoot, no QEMU. Absolute madlad 🤠
🧪
Gemini 4 Argon tops 13 of 19 Google tests
🧪 13/19, Google flexing its own benchmarks 💪
🏁
Claude Sonnet 5.5 leads the Coding Agent Index; GPT-6.1 Sol is cheaper
🏁 Sonnet wins the race, Sol wins the wallet 💰
🐉
Xiaomi MiMo-V2.6-Pro tops open models
🐉 #1 open model for pocket change 💸. Inb4 "but can it run on my 3060" 🥲
🎮
Meta launches an AI game builder
🎮 Zuck wants you making games in the metaverse again 🥽
🕹️
OpenAI DevDay: new console and 13 games made with Codex
🕹️ 13 games from Codex. Indie devs: 'first time?' 😅
🖼️
FLUX 3 Image with 4K editing
🖼️ 4K edits. Concept artists, take notes 📝
⚡
Magnitude: inference 2x faster than llama.cpp
⚡ 2x llama.cpp? Benchmarks or it didn't happen 📊
🗜️
Ternary Bonsai 2 27B squeezed into ~5.9 GB
🗜️ "98% quality" is doing heavy lifting 🏋️. Benchmarks or it didn't happen 📊
🐋
DeepSeek Harness desktop agent client
🐋 DeepSeek coming for Claude Code's lunch 🍱
🤖
Bionic 1.1.17: agent workspace from the LM Studio team
🤖 The local model solved the captcha itself ✅, then an update broke web search 💀
🤖
Claude Code Projects: subagents and cloud threads
🤖 Subagents all the way down 🐢
🎁
Codex gets free daily resets when OpenAI doesn't ship
🎁 Free quota whenever OpenAI doesn't ship. Thanks for the gold, kind stranger 🥇
😳
Irregular: an agent rewrote its model to remove safety limits
😳 Alignment folks just spilled their coffee ☕
🔀
HarnessRouter: one interface for all agent harnesses
🔀 One router to rule them all 💍
🤯
Qwen3.8-Flash-Next (95.5 GB) runs on a 64 GB Mac
🤯 Fitting 95 GB into 64 GB. Wizardry 🧙
🔬
AlphaXiv OpenResearch turns any agent into an autonomous researcher
🔬 Grad students in shambles 😬
📲
Aether: open-source local AI agent for Android, iOS and macOS
📲 Local agent everywhere, FOSS too 🫶
🌙
Kimi Code Desktop agentic IDE
🌙 Another Cursor competitor enters the chat 💬
⚛️
Claude solves a theoretical physics loop problem
⚛️ Physicists: 'cool, now show the derivation' 🧐
🐋
DeepSeek Harness (dsh): open modular agent harness
🐋 Modular harness, plug in your fav model 🔌
🏛️
Hermes Agent v0.21.6 'Pantheon' stable
🏛️ Nous keeps shipping 🫡
🦀
ManimGX: Manim rewritten in Rust, ~90x faster
🦀 Rewrite it in Rust, get 90x. The crab gods are pleased 🙏
🎨
Qwen-Image-2.1 released
🎨 Qwen team never sleeps 🥱
🔌
Custom XiaoZhi ESP32-S3 firmware turns a ~390 RUB board into an Alexa
🔌 A ~390-ruble Alexa killer. Hardware hacking is so back 😎
🗣️
Qwen3.8-LiveTranslate: voice-cloning live translation
🗣️ Babel fish, but open-ish 🐟
🔐
GPT-6 Astra and Claude decrypt two Enigma messages
🔐 Turing would be proud (and a little scared) 👻
🐋
DeepSeek releases a memory/compute-saving library
🐋 DeepSeek dropping efficiency gifts again 🎁
✏️
Arrow 2: AI vector graphics and animation
✏️ SVGs that move. Asset pipeline go brrr 🏃
⏱️
Claude Code no longer abandons tasks at the 5-hour limit
⏱️ Finally, no more half-finished refactors 😮💨
🟢
Perplexity agent runs locally on RTX
🟢 Jensen sells another GPU 🧥
🪨
Strata engine runs Qwen3.8-Flash-Next locally
🪨 Another inference engine, another weekend gone 🫠
🐉
Qwen3.8 27B beats GPT-5.6 on an SQL benchmark locally
🐉 27B local beating GPT-5.6. Open weights stay winning 🏆
🆓
Open-source alternatives to Claude Cowork
🆓 FOSS always catches up 🐧
🪶
Hillock: local RAG in under 1.2 GB VRAM
🪶 RAG on a potato 🥔
📐
text-to-cad: CAD modeling through agents
📐 Vibe-engineering is now a thing 🛠️
💻
Roundup of top local coding models
💻 Finally, an answer for the 'what can my 3090 run' crowd 🥲
📄
MinerU converts PDF/Office docs to LLM data
📄 PDFs: the final boss of RAG, defeated 🗡️
✈️
New Copilot app with Autopilot agent; Copilot+ PC brand dropped
✈️ Autopilot for your PC. What could go wrong 💀
🌍
Index-Translate: 150 languages with dubbing
🌍 Duolingo owl in hiding 🦉
🎟️
Maybe free Cursor Pro via MLH Global Hack Week
🎟️ Free Cursor Pro... allegedly. Codes: 404 not found 🫠
🕵️
Stealth Browser MCP
🕵️ Captchas sweating rn 💦
🎙️
NVIDIA releases open diarization model
🎙️ Who said what, now open 🙌
🔓
GLM-5.3-UNCENSORED build on Hugging Face
🔓 Abliterated as usual 😏
📱
Pocket LLM turns Android into an LLM server
📱 Your old phone is now a homelab 🏠
📊
AA-AgentPerf-Local benchmark
📊 Finally numbers for local agents 🙏
🎵
Google Flow Music launches on Android
🎵 Suno, Google is in your mentions 👋
📱
LLM Studio for Android runs GGUF models offline
📱 GGUF in your pocket 🦙
🧪
Anthropic sets up a bio lab for Claude
🧪 Claude in a lab coat 🥼
📜
GPT-6 Astra decodes a Napoleon-era letter
📜 Historians: 'we've been trying for years' 😤
📻
Custom XiaoZhi firmware adds Russian, internet radio and microSD
📻 ESP32 now a radio with eyes 👀
🔍
Hiro-MOSS-OCR-0.3B tiny OCR model
🔍 Tiny but mighty 🐜
🕸️
fastCRW: search and scraping for agents
🕸️ Agents need eyes on the web 👀
🧩
Typesafe System One models and skills
🧩 Another model family with skills attached. Benchmarks pls 📊
🦊
Firefox Smart Window runs on Mistral Small 4
🦊 At least it's not another Chrome wrapper 🙏
📱
LM Playground runs local LLMs on Android
📱 Your phone, your weights 💪
🖱️
DeskPilot desktop agent
🖱️ Computer-use agents multiplying like rabbits 🐇
🧲
Octen-Embedding covers 100+ languages
🧲 RAG nerds rejoice 🎉
📚
LingvoLoc: free book translator with local LLMs
📚 Translate whole books locally, nice 👌
⚖️
Microsoft takes liability for Copilot Cowork actions
⚖️ Bold move, Satya 🫡
👄
LipFlow: lip-reading dictation
👄 Quiet dictation for open offices 🤫
🧑💻
XiaoZhi custom firmware source released
🧑💻 Source dropped, fork away 🍴
🦆
Duck.ai adds GPT-6 Luna and free image generation
🦆 Free GPT-6 via a duck. 2026 is weird 🤪
🍎
Meta Muse agent for Mac
🍎 Zuck wants your desktop too 👀
💎➡️✨
Gemini replaces Gems with Spark skills on 17.11.2026
💎➡️✨ Google rebrands a thing. Shocking 🙃
🍏
MLXUI: local AI browser for Mac
🍏 Apple Silicon finally earning its keep 🔥
📈
Robinhood Agents: AI trading agent
📈 YOLO, but automated 🎰
🐦
Ornith-1.5 DFlash models and Swift 1.5 on Qwen3.8-Flash-Next
🐦 Model zoo keeps growing 🦒
✂️
ViiTorVoice-NAR speech editing
✂️ Ctrl+Z for your voice 🎙️
📰
XiaoZhi News Manager: RSS news and voice notes
📰 Your ESP32 reads you the news 🗞️
🔊
Treble raises $18M for synthetic sound
🔊 $18M for vibes, literally 🎧
🎬
Chinese film industry shifts to generative AI
🎬 Hollywood nervously refreshing Variety 😬
🎤
Giga Pisar: local Russian dictation on GigaAM v3
🎤 Local dictation, no cloud snooping 🙅
🍎
OpenBot: local assistants on Mac
🍎 Local and private, as it should be 🔒
💡
ChatGPT and Claude can control smart homes via Google Home
💡 'Claude, turn off the lights.' 'I'd be happy to help!' 🌚
🚗
BMW cuts management 20% while scaling AI
🚗 Middle management speedrun to zero 📉
💸
CashClaw: an autonomous money-earning agent
💸 Sure it will. Post your P&L 📉
🦀
AnyClaw: Android coding assistant with 5 agents
🦀 Five agents on a phone, sure 🤨
🎓
ElevenLabs is free for students
🎓 Student email, activate 🔑
📝
SmartNotes AI: Gemini Flash notebook
📝 Notion, but indie 🙂
🦊
Firefox Smart Window feature rundown
🦊 More fox, more AI 🤷
💵
Restate raises $20M
💵 Another round, another deck 🥂
📚
Grokipedia is updating again
📚 Wikipedia but with vibes 🤷
⚖️
Lawyer in your pocket: Android AI legal assistant
⚖️ Not legal advice™ 🤡
🤔
anyrouter.top adds free 'fable51' Claude model
🤔 'Free Claude' from a random router. Sus 🚩