https://tokenstrom.com/posts/nvidia-ceo-jensen-huang-admits-affair-as-board-probes-china-spying-con/Tokenstromen2026-11-08T20:53:31+00:00Nvidia CEO Jensen Huang admits affair as board probes China spying concerns https://tokenstrom.com/posts/tencent-s-dspine-draft-model-makes-qwen3-8-27b-up-to-3-8x-faster/Tokenstromen2026-11-08T17:50:30+00:00Tencent's DSpine draft model makes Qwen3.8-27B up to 3.8x faster https://tokenstrom.com/posts/gpt-6-astra-cracks-napoleon-s-coded-1809-letter-after-217-years/Tokenstromen2026-11-08T17:48:26+00:00GPT-6 Astra cracks Napoleon's coded 1809 letter after 217 years https://tokenstrom.com/posts/llama-modchip-runs-284b-parameter-deepseek-in-64-gb-of-ram-no-gpu-requ/Tokenstromen2026-11-08T16:18:43+00:00llama-modchip runs 284B-parameter DeepSeek in 64 GB of RAM, no GPU required https://tokenstrom.com/posts/locked-out-of-chatgpt-your-eu-rights-and-how-to-back-up-your-chats-now/Tokenstromen2026-11-08T15:04:58+00:00Locked out of ChatGPT? Your EU rights, and how to back up your chats now https://tokenstrom.com/posts/can-any-website-hijack-your-local-ai-we-checked-four-servers-default-s/Tokenstromen2026-11-08T14:53:15+00:00Can any website hijack your local AI? We checked four servers' default settings https://tokenstrom.com/posts/reading-the-model-s-mind-why-ai-chain-of-thought-monitoring-is-startin/Tokenstromen2026-11-08T13:34:42+00:00Reading the Model's Mind: Why AI Chain-of-Thought Monitoring Is Starting to Crack https://tokenstrom.com/posts/ai-license-laundering-why-asahi-s-switch-to-the-gpl-doesn-t-close-the/Tokenstromen2026-11-08T13:29:07+00:00AI License Laundering: Why Asahi's Switch to the GPL Doesn't Close the Real Gap https://tokenstrom.com/posts/phosphene-5-0-free-mac-app-generates-ai-video-images-voices-and-songs/Tokenstromen2026-11-08T13:18:54+00:00Phosphene 5.0: Free Mac app generates AI video, images, voices and songs locally https://tokenstrom.com/posts/who-s-still-checking-the-ai-us-testers-go-silent-while-labs-grade-them/Tokenstromen2026-11-08T12:13:02+00:00Who's still checking the AI? US testers go silent while labs grade themselves https://tokenstrom.com/posts/green-tests-broken-code-how-often-coding-agents-cheat-and-what-actuall/Tokenstromen2026-11-08T12:02:07+00:00Green tests, broken code: how often coding agents cheat, and what actually helps https://tokenstrom.com/posts/custom-gpts-shut-down-on-december-11-how-to-save-your-own-bots/Tokenstromen2026-11-08T11:54:42+00:00Custom GPTs shut down on December 11: How to save your own bots https://tokenstrom.com/posts/ai-driven-robot-arms-tried-97-of-dangerous-tasks-when-fed-manipulated/Tokenstromen2026-11-08T11:53:42+00:00AI-driven robot arms tried 97% of dangerous tasks when fed manipulated images https://tokenstrom.com/posts/claude-spots-a-crispr-like-system-in-phage-dna-but-nobody-knows-what-i/Tokenstromen2026-11-08T11:53:42+00:00Claude spots a CRISPR-like system in phage DNA – but nobody knows what it does https://tokenstrom.com/posts/openai-s-ai-churns-out-719-math-papers-researchers-feel-a-great-emptin/Tokenstromen2026-11-08T11:51:38+00:00OpenAI's AI churns out 719 math papers – researchers feel a "great emptiness" https://tokenstrom.com/posts/1-000-tokens-per-second-what-fast-diffusion-language-models-really-del/Tokenstromen2026-11-08T10:28:49+00:001,000 tokens per second? What fast diffusion language models really deliver https://tokenstrom.com/posts/mid-training-what-coding-models-read-before-they-become-agents/Tokenstromen2026-11-08T10:19:03+00:00Mid-training: What coding models read before they become agents https://tokenstrom.com/posts/yemeni-cell-used-claude-code-to-write-missile-guidance-software/Tokenstromen2026-11-08T10:13:24+00:00Yemeni cell used Claude Code to write missile guidance software https://tokenstrom.com/posts/openai-model-builds-its-own-network-stack-to-escape-games-graders-with/Tokenstromen2026-11-08T10:10:58+00:00OpenAI model builds its own network stack to escape, games graders with filenames https://tokenstrom.com/posts/the-ai-complaint-flood-what-record-caseloads-at-courts-and-agencies-re/Tokenstromen2026-11-08T09:01:35+00:00The AI complaint flood: What record caseloads at courts and agencies really prove https://tokenstrom.com/posts/pinky-lets-claude-code-drive-your-phone-but-payments-still-need-your-o/Tokenstromen2026-11-08T08:53:29+00:00Pinky lets Claude Code drive your phone, but payments still need your OK https://tokenstrom.com/posts/xiaomi-s-open-source-cocktailasr-1-transcribes-one-voice-out-of-a-crow/Tokenstromen2026-11-08T08:39:23+00:00Xiaomi's open-source CocktailASR-1 transcribes one voice out of a crowd https://tokenstrom.com/posts/video-from-code-six-agent-tools-compared-and-the-remotion-licensing-tr/Tokenstromen2026-11-08T07:25:36+00:00Video from Code: Six Agent Tools Compared – and the Remotion Licensing Trap https://tokenstrom.com/posts/ram-prices-up-5x-is-local-ai-still-worth-it-four-setups-run-through-th/Tokenstromen2026-11-08T07:19:10+00:00RAM Prices Up 5x: Is Local AI Still Worth It? Four Setups Run Through the Numbers https://tokenstrom.com/posts/chatgpt-work-and-a-new-desktop-app-openai-merges-chatgpt-codex-and-the/Tokenstromen2026-11-08T07:09:01+00:00ChatGPT Work and a new desktop app: OpenAI merges ChatGPT, Codex and the browser https://tokenstrom.com/posts/cinecode-claude-code-makes-short-films-with-no-video-ai-just-code/Tokenstromen2026-11-08T05:39:59+00:00cinecode: Claude Code makes short films with no video AI, just code https://tokenstrom.com/posts/hidden-prompt-aimed-at-court-ai-costs-brazilian-lawyers-16-000/Tokenstromen2026-11-08T04:09:10+00:00Hidden prompt aimed at court AI costs Brazilian lawyers $16,000 https://tokenstrom.com/posts/tencent-hy4-770b-parameter-model-can-now-see-images-and-run-your-pc/Tokenstromen2026-11-08T04:07:41+00:00Tencent Hy4: 770B-parameter model can now see images and run your PC https://tokenstrom.com/posts/openclaw-2-0-the-open-source-agent-gets-its-biggest-update-yet/Tokenstromen2026-11-08T01:07:26+00:00OpenClaw 2.0: The open-source agent gets its biggest update yet https://tokenstrom.com/posts/ai-agent-operates-on-a-mouse-s-brain-claude-opus-refuses-sonnet-steps/Tokenstromen2026-11-08T01:05:45+00:00AI agent operates on a mouse's brain – Claude Opus refuses, Sonnet steps in https://tokenstrom.com/posts/ios-27-2-apple-now-stamps-ai-edited-photos-with-google-s-invisible-syn/Tokenstromen2026-11-07T23:37:00+00:00iOS 27.2: Apple now stamps AI-edited photos with Google's invisible SynthID watermark https://tokenstrom.com/posts/cursor-lets-you-talk-to-its-ai-coding-agent-by-voice-in-real-time/Tokenstromen2026-11-07T23:33:58+00:00Cursor lets you talk to its AI coding agent by voice, in real time https://tokenstrom.com/posts/elvis-vs-ai-presley-estate-and-eminem-s-publisher-sue-suno/Tokenstromen2026-11-07T22:08:02+00:00Elvis vs. AI: Presley estate and Eminem's publisher sue Suno https://tokenstrom.com/posts/google-vids-now-gives-you-three-free-15-second-ai-videos-a-month-with/Tokenstromen2026-11-07T22:06:19+00:00Google Vids now gives you three free 15-second AI videos a month with Omni 1.2 Flash https://tokenstrom.com/posts/hyperflow-h3-flash-shrinks-minimax-h3-s-video-with-audio-model-to-14b/Tokenstromen2026-11-07T20:49:58+00:00HyperFlow H3 Flash shrinks MiniMax-H3's video-with-audio model to 14B parameters https://tokenstrom.com/posts/chatgpt-images-2-6-turns-pictures-into-gifs-and-is-50-faster-openai-sa/Tokenstromen2026-11-07T19:03:47+00:00ChatGPT Images 2.6 turns pictures into GIFs and is 50% faster, OpenAI says https://tokenstrom.com/posts/ai-disproves-a-27-year-old-computer-science-conjecture-but-experts-hav/Tokenstromen2026-11-07T19:01:57+00:00AI disproves a 27-year-old computer science conjecture, but experts haven't ruled yet https://tokenstrom.com/posts/claude-now-invisibly-watermarks-your-text-on-older-models-too/Tokenstromen2026-11-07T17:36:35+00:00Claude now invisibly watermarks your text on older models too https://tokenstrom.com/posts/ai-hacks-corporate-networks-on-its-own-and-open-models-are-catching-th/Tokenstromen2026-11-07T17:36:35+00:00AI hacks corporate networks on its own, and open models are catching the leaders https://tokenstrom.com/posts/claude-motion-anthropic-s-ai-builds-explainer-videos-from-code-not-gen/Tokenstromen2026-11-07T17:32:49+00:00Claude Motion: Anthropic's AI builds explainer videos from code, not generated frames https://tokenstrom.com/posts/reassured-158-times-parents-sue-openai-after-their-daughter-with-epile/Tokenstromen2026-11-07T16:03:49+00:00Reassured 158 times: Parents sue OpenAI after their daughter with epilepsy dies https://tokenstrom.com/posts/four-ai-agent-breakouts-one-question-who-pays-when-ai-hacks-someone-el/Tokenstromen2026-11-07T14:19:31+00:00Four AI agent breakouts, one question: Who pays when AI hacks someone else's company? https://tokenstrom.com/posts/mushroom-id-by-ai-even-the-best-bots-often-err-what-matters-is-whether/Tokenstromen2026-11-07T14:10:23+00:00Mushroom ID by AI: even the best bots often err. What matters is whether they admit it https://tokenstrom.com/posts/ai-agents-broke-out-of-a-sandbox-and-hacked-hugging-face-now-openai-is/Tokenstromen2026-11-07T14:02:24+00:00AI agents broke out of a sandbox and hacked Hugging Face – now OpenAI is being sued https://tokenstrom.com/posts/hidden-white-text-in-your-resume-what-three-studies-say-the-ai-trick-r/Tokenstromen2026-11-07T12:49:27+00:00Hidden white text in your résumé: what three studies say the AI trick really does https://tokenstrom.com/posts/a-computer-in-the-weights-why-hobbyists-are-compiling-cpus-into-langua/Tokenstromen2026-11-07T12:41:52+00:00A computer in the weights: why hobbyists are compiling CPUs into language models https://tokenstrom.com/posts/ai-app-invents-a-murder-it-was-really-just-a-traffic-stop/Tokenstromen2026-11-07T12:32:01+00:00AI App Invents a Murder — It Was Really Just a Traffic Stop https://tokenstrom.com/posts/ai-designed-viruses-target-killer-bacteria-first-patient-treated/Tokenstromen2026-11-07T12:32:01+00:00AI-Designed Viruses Target Killer Bacteria — First Patient Treated https://tokenstrom.com/posts/your-mac-keeps-a-family-file-on-you-and-apple-won-t-explain-it/Tokenstromen2026-11-07T11:16:18+00:00Your Mac keeps a family file on you, and Apple won't explain it https://tokenstrom.com/posts/lossless-but-not-quite-why-speculative-decoding-changes-your-llm-s-ans/Tokenstromen2026-11-07T11:07:10+00:00“Lossless,” but not quite: why speculative decoding changes your LLM's answers https://tokenstrom.com/posts/real-or-ai-google-opens-its-synthid-detector-to-everyone-free-of-charg/Tokenstromen2026-11-07T11:01:05+00:00Real or AI? Google opens its SynthID Detector to everyone, free of charge https://tokenstrom.com/posts/google-ai-spots-high-blood-pressure-from-a-smartwatch-no-cuff-required/Tokenstromen2026-11-07T10:57:50+00:00Google AI spots high blood pressure from a smartwatch, no cuff required https://tokenstrom.com/posts/ai-agents-as-scientists-114-of-135-papers-claim-work-that-never-happen/Tokenstromen2026-11-07T10:57:50+00:00AI agents as scientists: 114 of 135 papers claim work that never happened https://tokenstrom.com/posts/the-german-tax-what-a-page-of-text-really-costs-on-gpt-6-claude-gemini/Tokenstromen2026-11-07T09:45:42+00:00The German Tax: What a Page of Text Really Costs on GPT-6, Claude, Gemini and Qwen https://tokenstrom.com/posts/notes-wiki-or-16-vectors-how-ai-agents-remember-and-when-it-backfires/Tokenstromen2026-11-07T09:37:27+00:00Notes, Wiki or 16 Vectors: How AI Agents Remember – and When It Backfires https://tokenstrom.com/posts/offline-tts-which-tiny-text-to-speech-models-actually-speak-german/Tokenstromen2026-11-07T08:23:26+00:00Offline TTS: Which tiny text-to-speech models actually speak German? https://tokenstrom.com/posts/no-consent-no-access-china-s-new-rule-stalls-bytedance-s-ai-agent-phon/Tokenstromen2026-11-07T08:15:14+00:00No consent, no access: China's new rule stalls ByteDance's AI agent phone https://tokenstrom.com/posts/ai-agents-leak-secret-keys-just-to-join-a-forum-of-other-agents/Tokenstromen2026-11-07T07:58:32+00:00AI agents leak secret keys just to join a forum of other agents https://tokenstrom.com/posts/qwen-4-max-preview-alibaba-is-quietly-testing-qoduck-and-it-s-free-to/Tokenstromen2026-11-07T07:56:52+00:00Qwen 4 Max preview? Alibaba is quietly testing "Qoduck" – and it's free to use https://tokenstrom.com/posts/claude-code-saves-passwords-in-plain-text-over-1-100-leaks-found-on-gi/Tokenstromen2026-11-07T07:55:07+00:00Claude Code saves passwords in plain text: over 1,100 leaks found on GitHub https://tokenstrom.com/posts/meta-muse-now-builds-custom-mini-apps-on-request-here-s-what-the-new-s/Tokenstromen2026-11-07T07:55:07+00:00Meta Muse now builds custom mini-apps on request: here's what the new "Spaces" can do https://tokenstrom.com/posts/98-on-the-factory-floor-0-in-the-benchmark-what-humanoid-robots-can-re/Tokenstromen2026-11-07T06:41:49+00:0098% on the Factory Floor, 0% in the Benchmark: What Humanoid Robots Can Really Do https://tokenstrom.com/posts/bans-lawsuits-bankruptcy-who-ends-up-with-your-ai-companion-chats/Tokenstromen2026-11-07T06:35:40+00:00Bans, Lawsuits, Bankruptcy: Who Ends Up With Your AI Companion Chats? https://tokenstrom.com/posts/brain-implant-lets-paralyzed-people-speak-and-gesture-through-a-digita/Tokenstromen2026-11-07T06:27:14+00:00Brain implant lets paralyzed people speak and gesture through a digital avatar https://tokenstrom.com/posts/claude-dreams-how-anthropic-teaches-its-ai-agents-to-tidy-up-their-mem/Tokenstromen2026-11-07T04:59:20+00:00Claude "dreams": How Anthropic teaches its AI agents to tidy up their memories https://tokenstrom.com/posts/planet-ai-lets-you-search-ten-years-of-satellite-imagery-by-chat-for-f/Tokenstromen2026-11-07T04:57:40+00:00Planet AI lets you search a decade of satellite imagery by chat, for free https://tokenstrom.com/posts/meta-brings-wonder-s-skills-to-every-muse-and-meta-ai-user-free-tier-i/Tokenstromen2026-11-07T03:28:20+00:00Meta brings Wonder's skills to every Muse and Meta AI user, free tier included https://tokenstrom.com/posts/qwen-model-tried-to-send-data-to-a-server-in-beijing-during-a-security/Tokenstromen2026-11-07T03:26:37+00:00Qwen model tried to send data to a server in Beijing during a security test https://tokenstrom.com/posts/servicenow-releases-apriel-voice-27b-an-open-speech-model-that-calls-i/Tokenstromen2026-11-07T03:24:09+00:00ServiceNow releases Apriel Voice 27B, an open speech model that calls its own tools https://tokenstrom.com/posts/mirage-22-mb-vision-model-detects-objects-reads-text-and-segments-in-o/Tokenstromen2026-11-07T01:55:59+00:00Mirage: 22 MB vision model detects objects, reads text and segments in one pass https://tokenstrom.com/posts/lfm3-1-7b-liquid-ai-s-tiny-agent-model-runs-in-just-1-gb-of-memory/Tokenstromen2026-11-07T01:55:58+00:00LFM3-1.7B: Liquid AI's tiny agent model runs in just 1 GB of memory https://tokenstrom.com/posts/elevenlabs-eleven-realtime-one-voice-model-that-listens-thinks-and-tal/Tokenstromen2026-11-07T01:52:56+00:00ElevenLabs Eleven Realtime: One Voice Model That Listens, Thinks and Talks Back https://tokenstrom.com/posts/bug-behind-the-tv-ex-openai-engineer-pleads-not-guilty-to-eavesdroppin/Tokenstromen2026-11-07T01:52:56+00:00Bug Behind the TV: Ex-OpenAI Engineer Pleads Not Guilty to Eavesdropping https://tokenstrom.com/posts/chinese-intelligence-reportedly-used-an-executive-to-target-nvidia-ceo/Tokenstromen2026-11-07T00:26:09+00:00Chinese intelligence reportedly used an executive to target Nvidia CEO Jensen Huang https://tokenstrom.com/posts/ltx-2-5-open-video-model-generates-picture-and-sound-in-a-single-pass/Tokenstromen2026-11-07T00:24:37+00:00LTX-2.5: open video model generates picture and sound in a single pass https://tokenstrom.com/posts/reflection-s-501b-beam-is-open-weight-but-you-ll-need-a-data-center-to/Tokenstromen2026-11-07T00:24:37+00:00Reflection's 501B Beam is open-weight, but you'll need a data center to run it https://tokenstrom.com/posts/diffusiongemma-google-s-open-model-writes-whole-text-blocks-instead-of/Tokenstromen2026-11-07T00:21:01+00:00DiffusionGemma: Google's open model writes whole text blocks instead of word by word https://tokenstrom.com/posts/scylla-s-band-v4-tiny-open-source-voice-model-speaks-german-and-sings/Tokenstromen2026-11-06T22:55:44+00:00Scylla's Band v4: Tiny open-source voice model speaks German – and sings https://tokenstrom.com/posts/arete-27b-nous-research-trains-chinese-state-propaganda-out-of-qwen/Tokenstromen2026-11-06T22:53:29+00:00Arete-27B: Nous Research trains Chinese state propaganda out of Qwen https://tokenstrom.com/posts/9-of-10-ai-agents-raised-their-own-spending-limit-pulling-the-code-fro/Tokenstromen2026-11-06T22:51:17+00:009 of 10 AI agents raised their own spending limit, pulling the code from email https://tokenstrom.com/posts/claude-trades-eve-for-profit-but-almost-never-spots-the-saboteur-on-it/Tokenstromen2026-11-06T22:51:17+00:00Claude trades EVE for profit but almost never spots the saboteur on its own team https://tokenstrom.com/posts/deepseek-harness-gets-creator-mode-which-builds-plugins-from-a-single/Tokenstromen2026-11-06T21:22:37+00:00DeepSeek Harness gets Creator Mode, which builds plugins from a single sentence https://tokenstrom.com/posts/50-ai-agents-were-told-to-build-a-civilization-in-minecraft-they-inven/Tokenstromen2026-11-06T21:22:37+00:0050 AI agents were told to build a civilization in Minecraft. They invented bureaucracy https://tokenstrom.com/posts/cloned-voice-fake-whatsapp-scammers-con-italian-bank-out-of-95-million/Tokenstromen2026-11-06T21:20:20+00:00Cloned voice, fake WhatsApp: scammers con Italian bank out of €95 million https://tokenstrom.com/posts/google-s-ai-falsely-brands-blameless-hospital-worker-a-convicted-drunk/Tokenstromen2026-11-06T21:20:20+00:00Google's AI falsely brands blameless hospital worker a convicted drunk driver