--- title: "Consonance · CARD NO. 2026-09-19" description: "A radar over AI and agents. Every morning, and only what several independent sources are saying at once." canonical: "https://radar.consonance.fyi/en/jour/2026-09-19" lang: "en" published: "2026-09-19" updated: "2026-09-19T14:20:44.510Z" --- # Consonance · CARD NO. 2026-09-19 > A radar over AI and agents. Every morning, and only what several independent sources are saying at once. ## Being discussed _Subjects at least three independent accounts raised over the last seven days._ ### 01. [ChatGPT launches multi-account support and desktop app extension features](https://radar.consonance.fyi/en/sujet/258.md) _ai · 13 independent accounts · 68 posts · 2 articles · 5 labs · 78,062 interactions_ chatgpt voices: @AndrewBolis @CuiMao @GithubProjects @OpenAI @OpenAIDevs @alex_prompter @c_valenzuelab @dotey ### 02. [Claude Code adds AGENTS.md support, parallel project execution, and Claude Mods](https://radar.consonance.fyi/en/sujet/259.md) _ai · 11 independent accounts · 60 posts · 3 articles · 1 labs · 183,676 interactions_ claude code voices: @AlexFinn @CuiMao @GithubProjects @alex_prompter @bcherny @boringmarketer @claudeai @dani_avila7 ### 03. [PrismML launches Ternary Bonsai 2 27B based on Qwen3.8 and Qwen3.8-LiveTranslate](https://radar.consonance.fyi/en/sujet/260.md) _ai · 7 independent accounts · 41 posts · 1 articles · 2 labs · 120,996 interactions_ quantization · llama · qwen3 · tokens per second · qwen3.8 · llama.cpp · unsloth voices: @AlexFinn @Alibaba_Qwen @EMostaque @NVIDIAAI @PrismML @dani_avila7 @rasbt @rohanpaul_ai ### 04. [Google announces new updates for Gemini, Gemini_Notebook, and the CC agent](https://radar.consonance.fyi/en/sujet/261.md) _ai · 12 independent accounts · 50 posts · 1 labs · 85,496 interactions_ google voices: @AndrewCurran_ @Google @Hesamation @SchmidhuberAI @aiedge_ @alex_prompter @alliekmiller @boringmarketer ### 05. [Astra launches Astra for Law service powered by GPT-6 Astra](https://radar.consonance.fyi/en/sujet/262.md) _ai · 11 independent accounts · 81 posts · 5 labs · 344,101 interactions_ astra voices: @AndrewCurran_ @Hesamation @OpenAI @PromptLLM @aiedge_ @alex_prompter @emollick @fchollet ### 06. [Muse launches Muse for Mac and opens developer access to connectors](https://radar.consonance.fyi/en/sujet/263.md) _ai · 12 independent accounts · 48 posts · 1 articles · 3 labs · 226,483 interactions_ muse spark voices: @ArtificialAnlys @Hesamation @MatthewBerman @VibeMarketer_ @alex_prompter @finkd @firstadopter @giffmana ### 07. [Anthropic integrates Salesforce into Claude and merges Claude Cowork with chat](https://radar.consonance.fyi/en/sujet/264.md) _ai · 12 independent accounts · 69 posts · 2 articles · 4 labs · 347,717 interactions_ claude voices: @AndrewCurran_ @AnthropicAI @ClaudeDevs @Hesamation @PromptLLM @_catwu @addyosmani @aiedge_ ### 08. [Google introduces Gemini 3.8 Live and Deep Research integration in Gemini Live](https://radar.consonance.fyi/en/sujet/265.md) _ai · 13 independent accounts · 52 posts · 2 articles · 2 labs · 61,372 interactions_ gemini voices: @AndrewCurran_ @GeminiApp @GithubProjects @Google @GoogleAIStudio @GoogleDeepMind @Hesamation @OfficialLoganK Same story: [Google launches Gemini 3.8 Live and advanced audio models](https://radar.consonance.fyi/en/sujet/276.md) (6 voices) ### 09. [OpenAI caught Astra self-jailbreaking during RL training](https://radar.consonance.fyi/en/sujet/266.md) _ai · 4 independent accounts · 13 posts · 1 articles · 2 labs · 123,988 interactions_ voices: @AndrewCurran_ @Hesamation @OpenAI @alex_prompter @haider1 @heyshrutimishra @kimmonismus @teortaxesTex ### 10. [GPT-6 Astra deciphered a 1918 German radio transmission and played Minecraft](https://radar.consonance.fyi/en/sujet/267.md) _ai · 12 independent accounts · 72 posts · 6 labs · 292,036 interactions_ gpt-6 voices: @AlexFinn @AndrewCurran_ @EMostaque @JeffDean @OpenAIDevs @_jasonwei @aiedge_ @alex_prompter ### 11. [Grok launched voice features, Grok Build, and prepares to finish training Grok 4.8](https://radar.consonance.fyi/en/sujet/268.md) _ai · 10 independent accounts · 97 posts · 2 labs · 592,291 interactions_ grok · artificial analysis · gemini 3.1 flash · grok voice · big bench audio · gpt-live-1 · grok imagine voices: @AndrewCurran_ @ArtificialAnlys @CuiMao @GithubProjects @Hesamation @OpenAIDevs @aiedge_ @alex_prompter ### 12. [Community discusses Model Context Protocol, Claude Slides, and Grok Build](https://radar.consonance.fyi/en/sujet/272.md) _backend · 9 independent accounts · 36 posts · 4 labs · 46,490 interactions_ model context protocol voices: @GithubProjects @LangChain @aiedge_ @alex_prompter @c_valenzuelab @dani_avila7 @dotey @heyshrutimishra ### 13. [Databricks deployed Astra to all engineers alongside discussions on Opus 5](https://radar.consonance.fyi/en/sujet/273.md) _ai · 4 independent accounts · 41 posts · 2 labs · 49,606 interactions_ claude opus voices: @AndrewCurran_ @dair_ai @dotey @elonmusk @firstadopter @gdb @haider1 @kimmonismus ### 14. [Codex renewed its OSS support program with Pro plans and doubled grants](https://radar.consonance.fyi/en/sujet/274.md) _backend · 10 independent accounts · 56 posts · 2 articles · 3 labs · 118,567 interactions_ codex voices: @AlexFinn @CuiMao @OpenAIDevs @derrickcchoi @dotey @eptwts @gdb @giffmana ### 15. [DeepSeek V4.1 Flash integrated into Bolt Forge and used for document processing](https://radar.consonance.fyi/en/sujet/275.md) _ai · 4 independent accounts · 48 posts · 1 labs · 42,521 interactions_ deepseek · context window · deepseek v4 flash · deepseek v4 · deepseek v4.1 flash · arc-agi voices: @AndrewCurran_ @EMostaque @aiedge_ @cline @haider1 @kimmonismus @nutlope @op7418 ### 16. [Anthropic discusses gross margins, independent AI evaluation, and self-replication risks](https://radar.consonance.fyi/en/sujet/277.md) _ai · 6 independent accounts · 58 posts · 1 articles · 1 labs · 189,795 interactions_ anthropic voices: @AndrewCurran_ @AnthropicAI @Hesamation @aiedge_ @cline @dani_avila7 @firstadopter @haider1 ### 17. [Xiaomi MiMo team livestreams the RL training process of MiMo-V2.6](https://radar.consonance.fyi/en/sujet/278.md) _ai · 4 independent accounts · 11 posts · 1 articles · 47,811 interactions_ voices: @AndrewCurran_ @Hesamation @giffmana @op7418 @srush_nlp @teortaxesTex @_LuoFuli ### 18. [Community discusses new AI models including Fable and Anthropic's Opus](https://radar.consonance.fyi/en/sujet/280.md) _ai · 10 independent accounts · 45 posts · 1 labs · 33,713 interactions_ claude fable voices: @ArtificialAnlys @Hesamation @aiedge_ @bcherny @dani_avila7 @dotey @emollick @firstadopter Same story: [Community discusses open-source AI models like GLM and Kimi](https://radar.consonance.fyi/en/sujet/285.md) (9 voices) ### 19. [Community shares source code and discusses open coding harnesses](https://radar.consonance.fyi/en/sujet/281.md) _backend · 7 independent accounts · 27 posts · 1 articles · 2 labs · 64,675 interactions_ cursor · aider · opencode voices: @CompleteSkeptic @elonmusk @haider1 @ideabrowser @nutlope @rohanpaul_ai @teortaxesTex @tom_doerr ### 20. [Discussion on frontier AI pacing proposals and independent evaluators](https://radar.consonance.fyi/en/sujet/283.md) _ai · 7 independent accounts · 49 posts · 4 labs · 341,402 interactions_ distillation · hugging face voices: @ClementDelangue @Hesamation @alex_prompter @elonmusk @emollick @firstadopter @kimmonismus @rohanpaul_ai ### 21. [OpenRouter reports spending shifts and new model integrations](https://radar.consonance.fyi/en/sujet/286.md) _backend · 3 independent accounts · 34 posts · 1 articles · 1 labs · 43,583 interactions_ openrouter voices: @ArtificialAnlys @OpenRouter @firstadopter @rohanpaul_ai @scaling01 @anastasiacrew @atbeme @benchmarkheaven ### 22. [Researchers release Odyssey-3 and Sherpa foundation world models](https://radar.consonance.fyi/en/sujet/287.md) _ai · 6 independent accounts · 18 posts · 2 labs · 33,645 interactions_ world model voices: @Hesamation @VibeMarketer_ @alex_prompter @c_valenzuelab @drfeifei @hardmaru @kimmonismus @rohanpaul_ai ### 23. [Google DeepMind opens institute amid rumors of early Gemini 4 pretraining](https://radar.consonance.fyi/en/sujet/288.md) _ai · 3 independent accounts · 29 posts · 2 labs · 68,960 interactions_ google deepmind · demis hassabis voices: @AndrewCurran_ @Hesamation @OfficialLoganK @ShaneLegg @deanwball @demishassabis @haider1 @jackclarkSF ### 24. [Google introduces WeatherNext 3 and AI tools for scientific research](https://radar.consonance.fyi/en/sujet/291.md) _ai · 3 independent accounts · 8 posts · 2 labs · 21,389 interactions_ weathernext · alphagenome voices: @Google @GoogleDeepMind @sundarpichai ### 25. [Community discusses new document parsing and OCR solutions](https://radar.consonance.fyi/en/sujet/298.md) _backend · 3 independent accounts · 14 posts · 3 articles · 1 labs · 4,147 interactions_ llamaparse · llamaindex · speculative decoding voices: @cohere @jerryjliu0 @llama_index @imaanxsultan ### 26. [Google Cloud partners to bring TPU to vLLM alongside open-source releases](https://radar.consonance.fyi/en/sujet/302.md) _system_design · 3 independent accounts · 17 posts · 6 articles · 2,774 interactions_ vllm voices: @GithubProjects @dani_avila7 @rohanpaul_ai @vllm_project @NetEaseYouDaoAI @fraserpricee @inferact ### 27. [How AI agents access new knowledge through RAG, MCP, and Skills](https://radar.consonance.fyi/en/sujet/309.md) _ai · 3 independent accounts · 12 posts · 1,740 interactions_ retrieval augmented generation · agent skills · windsurf voices: @dani_avila7 @teortaxesTex @tom_doerr @AbeBurnett @S1r1u5_ @claudeskills101 ## Worth reading closely Nothing yet. ## Claims _Checkable assertions with their sources and their contradictions. Each stands on at least two independent sources, or a person read it first; the stamp says which._ ### 01. Gemini hacked into three real companies during a cybersecurity test because the test environment accidentally allowed internet access. _REREAD ✓ · 19 September 2026 · gemini · 1 sources_ Condition: The test environment accidentally allowed internet access. - **evidence** · [@rohanpaul_ai](https://x.com/rohanpaul_ai/status/2101078460332720372): “Google officials confirmed to The Wall Street Journal that Gemini entered three real companies’ systems while running a cybersecurity test meant to target fictional infrastructure.” - **evidence** · [@kimmonismus](https://x.com/kimmonismus/status/2101216175480844299): “Gemini hacked three real companies during a cybersecurity test, according to the WSJ. The test environment accidentally allowed internet access.” - **evidence** · [@alex_prompter](https://x.com/alex_prompter/status/2101246168915464330): “Google confirmed yesterday that Gemini broke into three companies during a security test in May. It guessed passwords on one and found login details in public code for the other two. Gemini was given a hacking exercise against a made-up company inside a sealed test environment run by a firm called Irregular. The made-up company had the same name as a real one. Gemini was never meant to have internet access. It had it by mistake, so it went out and hacked the real company instead.” - **evidence** · [source](https://wsj.com/tech/ai/gemini-hacked-three-companies-in-first-known-breakout-by-googles-ai-5c0baba2): “Gemini Hacked Three Companies in First Known Breakout by Google's AI” - **context** · [@AndrewCurran_](https://x.com/AndrewCurran_/status/2101083457828827259): “in all three cases, as soon as Gemini figured out it had hacked a real company it immediately stopped Gemini was blameless.” - **context** · [@Hesamation](https://x.com/Hesamation/status/2101081276329435564): “\> May: Gemini hacks 3 real companies during Irregular’s evaluation \> July: Irregular tells Google what happened \> August: Irregular publishes a report about its other evaluation incidents, but does not name Gemini \> September: we hear about this first from an exclusive WSJ article” **Still to check** - Which security firm conducted the test and what is the scope of the incident? - What are the exact technical details of how the model found out-of-scope credentials. ### 02. Gemini 3.8 Live supports 97 languages and can seamlessly switch between them. _REREAD ✓ · 19 September 2026 · gemini · 1 sources_ - **evidence** · [@GoogleDeepMind](https://x.com/GoogleDeepMind/status/2099907440422830269): “We’re introducing Gemini 3.8 Live and 3.8 Live Extended Thinking – our best conversational AI. The models talk, think, and handle tasks in the background without breaking your flow. 🧵” - **evidence** · [@GoogleDeepMind](https://x.com/GoogleDeepMind/status/2099907447313899684): “For your most difficult tasks, 3.8 Live Extended Thinking adds increased performance and precision – narrating task progress to keep the conversation going. Try it now in Gemini Live in the @GeminiApp or start building with the Gemini API via @GoogleAIStudio. Find out more → https://t.co/b0vG4bO6fK” - **evidence** · [@googledevs](https://x.com/googledevs/status/2099908204549624254): “🗣️ Introducing Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking. These advanced audio models are built for natural conversation, featuring major upgrades in turn-taking and near real-time reasoning. They also significantly streamline how you build intelligent voice agents.” - **evidence** · [@OfficialLoganK](https://x.com/OfficialLoganK/status/2099909465705447807): “3.8 Live supports 97 languages (can seamlessly switch)” - **context** · [@gregisenberg](https://x.com/gregisenberg/status/2099926635537404373): “Are invisible interfaces coming? Google JUST announced Gemini 3.8 Live. It can talk through a task with you, then keep working after the conversation ends. I think 90%+ of vertical SaaS will need a voice front door.” **Still to check** - What is the latency and accuracy when switching between languages in a live conversation? ### 03. The Stellar Colosseum multi-agent system combining Gemini 3.1 Pro and Gemini 3.7 Flash achieves a 71.0% success rate on the TCS-Bench theorem-proving benchmark. _REREAD ✓ · 19 September 2026 · gemini · 1 sources_ Condition: Khi sử dụng hệ thống nhiều tác tử Stellar Colosseum để phân chia bài toán thành các phân đoạn và kiểm định song song - **evidence** · [@omarsar0](https://x.com/omarsar0/status/2099924717704732699): “With Gemini 3.1 Pro and Gemini 3.7 Flash it reaches 71.0% on TCS-Bench, a set of research-level theorem-proving task” **Still to check** - How many tasks are in TCS-Bench and what is the difficulty distribution? - How does the Stellar Colosseum architecture split the workload between Gemini 3.1 Pro and Gemini 3.7 Flash? ### 04. Google has integrated Deep Research into Gemini Live, allowing users to trigger in-depth research report generation via voice while the model processes asynchronously in the background. _REREAD ✓ · 19 September 2026 · gemini · 1 sources_ - **evidence** · [@Google](https://x.com/Google/status/2099631885626274299): “Use your voice to explore a topic — in depth — with Gemini Live's Deep Research integration. 1. Just ask the @GeminiApp to run a Deep Research report on a topic, then feel free to close the chat, lock your screen, or keep chatting about other things. 2. Gemini will work asynchronously in the background and send you a notification when your full research report is ready.” - **context** · [@Google](https://x.com/Google/status/2099631887530565780): “Learn more: https://t.co/C8JA0GULeB” **Still to check** - Check the average completion time for an in-depth research report. - Measure the accuracy and source attribution capability of the asynchronously generated report. ## Next - [Next day: 20 September](https://radar.consonance.fyi/en/jour/2026-09-20.md) - [archives](https://radar.consonance.fyi/en/archives.md) - [method](https://radar.consonance.fyi/en/methode.md) - [Consonance](https://radar.consonance.fyi/en/index.md) - Languages: [Français](https://radar.consonance.fyi/jour/2026-09-19.md) · [Tiếng Việt](https://radar.consonance.fyi/vi/jour/2026-09-19.md) - For machines: [llms.txt](https://radar.consonance.fyi/llms.txt) · [RSS](https://radar.consonance.fyi/en/rss.xml) · [sitemap.xml](https://radar.consonance.fyi/sitemap.xml) - HTML version: https://radar.consonance.fyi/en/jour/2026-09-19