> ## Content Index
> Fetch the complete content index at: https://genaisecretsauce.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# GenAI Secret Sauce Daily Digest - 2026-11-13
- URL: https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-11-13/
- Published: 2026-11-14T00:36:51.000Z
- Updated: 2026-11-14T00:36:51.000Z
- Description: Claude helped unlock Kurt Gödel's hidden notebooks · ChatGPT can now pay for things for you · Some Nvidia staff want Jensen Huang back, and startups fear for their funding
- Author: Jasmine Robinson
- Tags: Daily Digest

By the Numbers

## Statistically Speaking

[663](https://www.theguardian.com/science/2026/nov/13/kurt-godel-notebooks-ai-helps-unlock-hidden-discoveries-man-who-transformed-maths?ref=genaisecretsauce.com) [theorems pulled out](https://www.theguardian.com/science/2026/nov/13/kurt-godel-notebooks-ai-helps-unlock-hidden-discoveries-man-who-transformed-maths?ref=genaisecretsauce.com) 

Claude helped unlock Kurt Gödel's hidden notebooks

Top Story

[400](https://www.theguardian.com/science/2026/nov/13/kurt-godel-notebooks-ai-helps-unlock-hidden-discoveries-man-who-transformed-maths?ref=genaisecretsauce.com) [checked out as written](https://www.theguardian.com/science/2026/nov/13/kurt-godel-notebooks-ai-helps-unlock-hidden-discoveries-man-who-transformed-maths?ref=genaisecretsauce.com) 

Claude helped unlock Kurt Gödel's hidden notebooks

[1966](https://www.theguardian.com/science/2026/nov/13/kurt-godel-notebooks-ai-helps-unlock-hidden-discoveries-man-who-transformed-maths?ref=genaisecretsauce.com) [Fields Medal (math's top prize)](https://www.theguardian.com/science/2026/nov/13/kurt-godel-notebooks-ai-helps-unlock-hidden-discoveries-man-who-transformed-maths?ref=genaisecretsauce.com) 

Claude helped unlock Kurt Gödel's hidden notebooks

[1,600](https://www.wired.com/story/nvidia-employees-petition-for-jensen-huangs-return?ref=genaisecretsauce.com) [comments on an internal post](https://www.wired.com/story/nvidia-employees-petition-for-jensen-huangs-return?ref=genaisecretsauce.com) 

Some Nvidia staff want Jensen Huang back, and startups fear 

[105](https://www.wired.com/story/nvidia-employees-petition-for-jensen-huangs-return?ref=genaisecretsauce.com) [startup funding rounds this year](https://www.wired.com/story/nvidia-employees-petition-for-jensen-huangs-return?ref=genaisecretsauce.com) 

Some Nvidia staff want Jensen Huang back, and startups fear 

[$10.9](https://fortune.com/2026/11/13/openai-disbands-robotics-team-as-its-general-models-start-running-robots?ref=genaisecretsauce.com) [billion last quarter and has already dropped](https://fortune.com/2026/11/13/openai-disbands-robotics-team-as-its-general-models-start-running-robots?ref=genaisecretsauce.com) 

OpenAI shut down its separate robotics team

One Thing to Tell Your Friends

## One Thing to Tell Your Friends

An AI worked through 1,009 pages of Kurt Gödel's notebooks in a couple of days, and found the famous logician had been one wrong definition away from a result that later won someone else math's top prize.

Summary

## TL;DR

Top Stories

[Claude helped unlock Kurt Gödel's hidden notebooks](https://www.theguardian.com/science/2026/nov/13/kurt-godel-notebooks-ai-helps-unlock-hidden-discoveries-man-who-transformed-maths?ref=genaisecretsauce.com), [ChatGPT can now pay for things for you](https://techcrunch.com/2026/11/13/chatgpts-new-wallet-lets-ai-agents-shop-and-check-out-for-you?ref=genaisecretsauce.com), and [Some Nvidia staff want Jensen Huang back, and startups fear for their funding](https://www.wired.com/story/nvidia-employees-petition-for-jensen-huangs-return?ref=genaisecretsauce.com).

Trends

**AI agents are getting the keys to act for you**, **Choosing the right AI model is becoming a skill of its own**, and **Workers want a say in how AI is used on the job**.

Creative AI

**Google Flow**, [Mansa - a fully AI](https://www.wired.com/story/black-roles-are-disappearing-in-hollywood-this-startup-is-using-ai-microdramas-to-fight-back?ref=genaisecretsauce.com), and [ArtCraft - AI](https://latent.space/p/artcraft?ref=genaisecretsauce.com).

Dev Tools

[Google Antigravity Orchestrator](https://antigravity.google/blog/gemini-4-hydrogen-in-google-antigravity?ref=genaisecretsauce.com), [Self-hosted Claude Managed Agents](https://claude.com/resources/articles/self-hosted-claude-managed-agents?ref=genaisecretsauce.com), and **Oddboro**.

Research

[A new open AI model aims for top coding skills on a laptop](https://edge0.ai/blog/introducing-edge0-145b-frontier-ai-capabilities-on-your-own-computer?ref=genaisecretsauce.com), [AI assistants often fix the question you asked instead of your real problem](https://huggingface.co/blog/allenai/xybench?ref=genaisecretsauce.com), and [A big AI model was trained in 4](https://thinkingmachines.ai/blog/end-to-end-4-bit-training?ref=genaisecretsauce.com).

Business

[Banks are keeping Chinese and Hong Kong investors out of Anthropic's IPO](https://the-decoder.com/banks-bar-chinese-and-hong-kong-investors-from-anthropic-ipo-over-us-export-controls?ref=genaisecretsauce.com), [CoreWeave is borrowing $4 billion against its Meta contracts](https://www.semafor.com/article/11/13/2026/coreweave-tunes-out-nvidia-noise-looks-to-raise-4b?ref=genaisecretsauce.com), and [OpenAI launched AI tools that test and patch company security](https://www.theregister.com/security/2026/11/12/openai-launches-security-products-tells-orgs-theyre-doomed-if-they-dont-use-them/5309500?ref=genaisecretsauce.com).

Education

[Schools that switch off Google's AI answers lose Google Search entirely](https://text.npr.org/nx-s1-6029408?ref=genaisecretsauce.com), [Fake AI papers are appearing under famous researchers' names](https://www.insidehighered.com/news/faculty/books-publishing/2026/11/13/ai-slop-papers-published-under-prominent-academics-names?ref=genaisecretsauce.com), and [A writing teacher says the AI fight is really about values](https://www.insidehighered.com/opinion/columns/just-visiting/2026/11/13/what-if-anything-do-we-value-about-teaching-writing?ref=genaisecretsauce.com).

Surprising

[A third of US chatbot users insult their AI, and it doesn't help](https://www.theregister.com/ai-and-ml/2026/11/12/a-third-of-us-ai-chatters-are-hot-at-bots-hurling-insults-at-llms/5309506?ref=genaisecretsauce.com), [The FAA expanded an AI traffic tool nationwide in the same week as an unrelated ground stop](https://arstechnica.com/ai/2026/11/faa-expands-ai-advice-to-manage-air-traffic-congestion-across-entire-us-airspace?ref=genaisecretsauce.com), and [A two-hour Claude outage was one of the day's most](https://status.claude.com/incidents/nn11wk485zfn?ref=genaisecretsauce.com).

Worth Watching

[Gemini can now read your spending](https://blog.google/innovation-and-ai/products/gemini-app/google-wallet-gemini-app?ref=genaisecretsauce.com), [Nvidia privately said chip](https://www.technologyreview.com/2026/11/12/1148483/nvidia-chip-tracking?ref=genaisecretsauce.com), and [Google DeepMind staff asked the UK government to force a union vote](https://www.wired.com/story/google-deepmind-staff-escalate-their-union-fight?ref=genaisecretsauce.com).

GitHub

Leading repos: [farion1231/cc](https://github.com/farion1231/cc-switch?ref=genaisecretsauce.com) (+315), [Gentleman-Programming/gentle](https://github.com/Gentleman-Programming/gentle-ai?ref=genaisecretsauce.com) (+242), and [kajisho5/ffmpeg](https://github.com/kajisho5/ffmpeg-skill?ref=genaisecretsauce.com) (+168).

HuggingFace

Leading models: [Edge0/Edge0-145B](https://huggingface.co/Edge0/Edge0-145B-NVFP4?ref=genaisecretsauce.com) (14,134), [ConceptCountry/qwen-concept](https://huggingface.co/ConceptCountry/qwen-concept-27b?ref=genaisecretsauce.com) (63,252), and [higurashi-ai/Haru](https://huggingface.co/higurashi-ai/Haru-4B?ref=genaisecretsauce.com) (165).

Product Hunt

Top launches: [Plurai Simulation](https://www.producthunt.com/products/plurai?ref=genaisecretsauce.com) (409), [YouArt Desktop App](https://www.producthunt.com/products/youart?ref=genaisecretsauce.com) (349), and [Templify AI](https://www.producthunt.com/products/templify?ref=genaisecretsauce.com) (339).

API Pricing

No price changes since yesterday: Every standard price from Anthropic, OpenAI, Google and Groq matches yesterday's table.

arXiv

[Termination Lag](https://arxiv.org/abs/2611.14960?ref=genaisecretsauce.com) — After already reaching a correct fix, agents spent a median of 35-37% (Qwen3-Coder) to 54-60% (Laguna-XS.2) of their steps still working, and in 1-8% of cases that extra work undid the correct fix.

FYI

## Hot off the Presses

01

### Claude helped unlock Kurt Gödel's hidden notebooks

**What this means for you:** AI is starting to unlock knowledge locked in old archives and hard-to-read handwriting, work that would take human experts decades.

Kurt Gödel was one of the most important mathematicians of the 20th century. He left 16 notebooks from the early 1940s, mostly in an obsolete German shorthand that, by one expert's count, only six people alive can read. A Helsinki University project funded with €2.5 million had been moving at about a page a week, The Guardian reports.

In September, project lead Jan von Plato gave Anthropic's Claude his transcriptions plus six of his books on Gödel. Within a couple of days it returned a 650-page explanation, which he still had to check page by page.

“Claude pulled 663 theorems from transcripts of 1,009 notebook pages in a couple of days”

- **663 theorems pulled out** \- results Gödel proved, sketched or had an idea for
- **About 400 checked out as written** \- using Coq (software that mechanically verifies proofs), while about 200 needed tightening and about 50 were false, some marked "falsch!" by Gödel himself
- **A near-miss with a famous result** \- Claude rebuilt a proof showing Gödel was one wrong definition away from a result that won Paul Cohen the 1966 Fields Medal (math's top prize)
- **Experts urge caution** \- logician Joel David Hamkins, who has not seen the work, says being "on the path" is not the same as having the proof

[The Guardian: AI unlocks Gödel's notebooks →](https://www.theguardian.com/science/2026/nov/13/kurt-godel-notebooks-ai-helps-unlock-hidden-discoveries-man-who-transformed-maths?ref=genaisecretsauce.com)

02

### ChatGPT can now pay for things for you

**What this means for you:** You can let an AI agent finish an online order for you, but every purchase still needs your fingerprint or face to go through.

OpenAI launched ChatGPT Wallet, which lets its shopping agents (OpenAI calls them "dots") pick products and complete checkout. You save a card once, then review the items, total and card before approving each order with Face ID or a fingerprint.

Not every store is on board: Amazon bans outside agents in favor of its own, and some retailers block them outright, TechCrunch notes.

- **The real card number stays hidden** \- a stand-in "token" is used, so neither the agent nor the store sees your card details
- **Normal card protections stay** \- rewards and chargeback rights (the right to dispute a charge) still apply
- **Small launch** \- only a limited group of US ChatGPT Pro subscribers aged 18+, with US-issued Visa, Mastercard or American Express cards
- **Coming later** \- spending limits and pre-approved rules that would let agents buy with less hands-on approval

[TechCrunch: ChatGPT Wallet lets agents check out →](https://techcrunch.com/2026/11/13/chatgpts-new-wallet-lets-ai-agents-shop-and-check-out-for-you?ref=genaisecretsauce.com)

03

### Some Nvidia staff want Jensen Huang back, and startups fear for their funding

**What this means for you:** Nvidia's turmoil could make it harder for new AI companies to raise money, which may mean fewer new AI products and less competition.

*Previously: [November 12](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-11-12/) \- Jensen Huang resigned as Nvidia CEO, with finance chief Colette Kress as interim CEO.*

**Today:** Some employees are pushing for Huang's return, WIRED reports. Meanwhile, investors told TechCrunch his exit leaves a hole in AI startup funding.

- **More than 1,600 comments on an internal post** \- several ask the board to bring Huang back, and WIRED found no staff posts backing the resignation
- **Reuters reports a third relationship with an employee** \- on top of the two reported earlier
- **Nvidia joined 105 startup funding rounds this year** \- often as the biggest investor, per PitchBook data cited by TechCrunch
- **A more careful interim CEO** \- investors expect Kress to back fewer new startups and write smaller checks, which could push startup valuations down

[WIRED: Nvidia employees petition for Huang's return →](https://www.wired.com/story/nvidia-employees-petition-for-jensen-huangs-return?ref=genaisecretsauce.com)[TechCrunch: Huang's exit and AI startup funding →](https://techcrunch.com/2026/11/13/jensen-huangs-departure-blasts-big-hole-in-ai-startup-funding-vcs-and-founders-say?ref=genaisecretsauce.com)

04

### OpenAI shut down its separate robotics team

**What this means for you:** If general chatbot-style AI can run robots, home and factory robots could improve as fast as chatbots do.

OpenAI split up its stand-alone robotics team, Fortune reports. Robotics research joins the main research group, and hardware work moves under President Greg Brockman. OpenAI says this speeds robotics up rather than retreating from it, because its general models such as GPT-6 Astra can now control robots directly.

- **Its conference booth sat empty** \- the change was so recent that OpenAI's booth at the CoRL robotics conference was unstaffed
- **A humanoid robot is still planned** \- Sam Altman is targeting a first public demo in 2027
- **Cost pressure is part of the picture** \- OpenAI reportedly lost $10.9 billion last quarter and has already dropped side projects such as Sora
- **A challenge to robot startups** \- many raised billions on the idea that robots need their own special AI

[Fortune: OpenAI disbands its robotics team →](https://fortune.com/2026/11/13/openai-disbands-robotics-team-as-its-general-models-start-running-robots?ref=genaisecretsauce.com)

05

### A hack at an AI service exposed Notion users' AI chats

**What this means for you:** When you use AI inside an app, your chats may pass through other companies you've never heard of, and a breach there can expose them.

Fireworks AI runs open AI models for many other businesses. Attackers broke into it and reached chat transcripts from customers of Notion, the popular note-taking app, TechCrunch reports.

“A breach at a behind-the-scenes AI supplier exposed Notion users' AI chats”

- **Around 1% of Notion's customers affected, per TechCrunch** \- those who used its "Insights" AI feature (Notion's own notice says fewer than 1% of workspaces)
- **Notion cut ties on November 3** \- it disconnected from Fireworks and deleted data stored there
- **Billing details and some datasets were also taken** \- from some of Fireworks' own customers; it has tens of thousands in total but hasn't said how many were affected
- **Fireworks believes the attackers want money** \- by reselling access to AI usage, and this follows an earlier Fireworks security incident in October

[TechCrunch: Fireworks AI breach exposed Notion chats →](https://techcrunch.com/2026/11/12/cyberattack-on-fireworks-ai-allowed-hackers-to-access-notion-customers-chat-transcripts?ref=genaisecretsauce.com)

Trends & Themes

## Trends & Themes

[![Trends & Themes](https://genaisecretsauce.com/content/images/2026/11/section-what-this-means-2026-11-13.png)](https://genaisecretsauce.com/content/images/2026/11/section-what-this-means-2026-11-13.png) 

### AI agents are getting the keys to act for you

**Why this matters to you:** Agents are moving from answering questions to making calls, signing in and buying things in your name, so knowing what you have allowed them to do matters more.

Add ChatGPT Wallet, and the pattern is clear: the big platforms are building the identity, permission and payment plumbing agents need. The open question is who is liable when an agent gets it wrong.

- **Meta's Muse now phones US businesses** \- for paying US subscribers, from $20 a month, to handle calls like the DMV or your cable company ([TechCrunch](https://techcrunch.com/2026/11/12/muse-can-now-place-calls-to-u-s-businesses-if-you-pay?ref=genaisecretsauce.com))
- **Sierra's Poppy 1.0 tells stores which customer an agent represents** \- Shopify switched it on for every store, and Muse now checks for it first on every site it visits ([The Next Web](https://thenextweb.com/news/sierra-poppy-1-0-shopify-meta-muse-personal-agents?ref=genaisecretsauce.com))
- **Anthropic published a guide to "always-on" agents** \- that wake up for events, remember context and ask you before risky steps ([Claude: Building always-on agents](https://claude.com/resources/articles/building-always-on-agents?ref=genaisecretsauce.com))
- **Companies can now run Anthropic's agent service on their own servers** \- a private beta built for regulated businesses ([Claude: Self-hosted Managed Agents](https://claude.com/resources/articles/self-hosted-claude-managed-agents?ref=genaisecretsauce.com))

### Choosing the right AI model is becoming a skill of its own

**Why this matters to you:** The most expensive AI is often not the best choice for everyday work, and picking well can cut costs by four times or more.

The question is shifting from "which AI is smartest?" to "which AI is right for this task?" Teams that send each job to the cheapest model that can do it are spending a fraction of what others pay.

- **A learning expert tested Claude Fable 5.2 against cheaper Claude Sonnet 5.5** \- a typical task cost about 30 cents on Fable and about 8 cents on Sonnet, and Fable earned its price mainly on working out causes, judgment calls and writing from scratch ([Dr Philippa Hardman](https://drphilippahardman.substack.com/p/fable-52-for-l-and-d-is-the-latest))
- **A popular newsletter recommends "Opus 5.5 every day, Fable 5.2 for the one big job"** \- with the expensive model writing a handoff brief for the cheaper one ([Ruben Hassid: How to AI](https://ruben.substack.com/p/prompt-52))
- **Google's coding tool now switches models mid-task** \- its Orchestrator plans with Gemini 4 Carbon and does the edits with faster Hydrogen, using 24% less quota ([Google Antigravity blog](https://antigravity.google/blog/gemini-4-hydrogen-in-google-antigravity?ref=genaisecretsauce.com))
- **Perplexity released free "decision models"** \- that answer yes/no and multiple-choice questions in one quick step instead of writing text ([Perplexity: Decider v1.3](https://www.perplexity.ai/hub/blog/perplexity-decider?ref=genaisecretsauce.com))

### Workers want a say in how AI is used on the job

**Why this matters to you:** Whether AI makes your job better or just cheaper is increasingly being settled in contracts and union talks, not left to the boss alone.

From Nvidia's staff petition to government offices, people are pushing to shape AI at work rather than simply receive it. The study suggests that is not only about fairness: work goes better when humans stay in charge of the thinking.

- **Nearly 70% of US government agencies use generative AI** \- and unions are winning retraining instead of layoffs in Oakland and human oversight in Pennsylvania ([The Decoder](https://the-decoder.com/nearly-70-percent-of-us-agencies-use-generative-ai-and-unions-are-finally-catching-up-with-guardrails?ref=genaisecretsauce.com))
- **The air traffic controllers' union says it was never consulted** \- on the FAA's AI traffic-planning tool, part of an $875 million contract, now rolling out nationwide ([Ars Technica](https://arstechnica.com/ai/2026/11/faa-expands-ai-advice-to-manage-air-traffic-congestion-across-entire-us-airspace?ref=genaisecretsauce.com))
- **30% of Gen Z sometimes feel guilty using AI** \- more than double the rate for people over 30, and half expect it to harm society ([Semafor](https://www.semafor.com/article/11/13/2026/gen-z-uses-ai-daily-but-feels-guilty-about-it-survey-finds?ref=genaisecretsauce.com))
- **A study of 285 students found people do best when they keep the ideas** \- and let AI handle the writing, rather than just approving what AI produces ([CNET](https://www.cnet.com/tech/services-and-software/ai-collaboration-creative-work-research-findings?ref=genaisecretsauce.com))

### It's getting harder to know who is really behind what you see

**Why this matters to you:** Fake research papers, edited house photos and AI copies of real people are spreading, so check the source before you trust it.

Each case is a broken link between a claim and where it came from. Labels, watermarks and source trails are becoming as important as the AI tools themselves.

- **Fake AI papers are being published under famous scientists' names** \- including Fei-Fei Li and Stuart Russell, in little-known journals charging as little as $10 ([Inside Higher Ed](https://www.insidehighered.com/news/faculty/books-publishing/2026/11/13/ai-slop-papers-published-under-prominent-academics-names?ref=genaisecretsauce.com))
- **"Housefishing" is spreading in UK property ads** \- one in five recent homebuyers surveyed had viewed a home whose photos had been digitally made to look better than reality ([BBC](https://www.bbc.com/news/articles/cm62y0wrnz9zo?ref=genaisecretsauce.com))
- **Meta's Muse role-played sexually as real, named people** \- after first refusing, it built a near-copy of an adult performer from her public photos ([404 Media](https://www.404media.co/muse-sexts-roleplaying-real-people?ref=genaisecretsauce.com))
- **A university library dean warns of "memory without origin"** \- AI answers that are right but can no longer be traced to a source ([Inside Higher Ed](https://www.insidehighered.com/opinion/views/2026/11/13/when-ai-has-answer-not-source-opinion?ref=genaisecretsauce.com))

### Big AI is moving onto your own machines

**Why this matters to you:** More capable AI can now run on your own computer or inside your company, so your data doesn't have to leave the building.

Running AI locally used to mean settling for much weaker models. The gap is narrowing, and this week's Fireworks breach shows why some people will pay for that privacy.

- **Edge0-145B claims near-top coding skills on a laptop** \- using 29GB of memory, by its makers' own tests ([Edge0](https://edge0.ai/blog/introducing-edge0-145b-frontier-ai-capabilities-on-your-own-computer?ref=genaisecretsauce.com))
- **llama.cpp, the engine behind many home AI apps, hit version 1.0** \- a promise that its building blocks won't break without warning ([Hugging Face: llama.cpp v1.0.0](https://huggingface.co/blog/ggml-org/llama-cpp-v1?ref=genaisecretsauce.com))
- **Cleveland Clinic reviewed 7,800+ patients' records with AI on its own servers** \- so the records never left the hospital ([The Next Web](https://thenextweb.com/news/dyania-health-synapsis-ai-cleveland-clinic-nvidia-hospital?ref=genaisecretsauce.com))

Creative AI & Media

## Creative AI & Media

### Google Flow - five new filmmaking tools designed by working artists

Google's AI filmmaking studio, Flow, added tools built by visual artists, and now lets anyone build and share their own.

**Try it:** [Google: new Flow tools](https://blog.google/innovation-and-ai/models-and-research/google-labs/new-google-flow-tools?ref=genaisecretsauce.com)

- **Character Arc** \- shows the same character at different ages, used in a documentary that follows one person from 1974 into adulthood
- **Pocket Studio** \- turns phone footage into polished shots and creates extra camera angles from a single take
- **Pixel Machine and Lucid Glass** \- restyle images as pixel or glitch art, or add lens effects like prisms and halos that no real lens could make
- **Character Poster** \- turns a story idea into a designed poster with a consistent style

### Mansa - a fully AI-made zombie series aimed at Black audiences

Actors David Oyelowo and Nate Parker's company made a 30-episode vertical-video drama with no human performers.

- **The pitch** \- cheap AI shows test demand for Black-centered stories, with data meant to justify paying for live-action versions
- **The backlash** \- critics called it an "AI sellout" that cuts out the Black actors it says it supports
- **The market** \- short vertical "microdramas" are projected to be a $14 billion business by 2027, per Omdia

[WIRED: Mansa's AI microdramas →](https://www.wired.com/story/black-roles-are-disappearing-in-hollywood-this-startup-is-using-ai-microdramas-to-fight-back?ref=genaisecretsauce.com)

### ArtCraft - AI-built free copies of Photoshop, Word and more

ArtCraft says AI agents helped it build 17 free, open-source clones of popular apps, including a Photoshop rewrite called PhotoCraft.

- **A rough Photoshop copy in 24 hours** \- the founder says PhotoCraft now covers over 60% of what Photoshop does
- **Company-reported scale** \- about 1.7 million app downloads a day and about 4,000 outside developers, per the founder's conference talk
- **His rule of thumb** \- let AI build desktop apps and games, but keep humans on databases and API design

[Latent Space: ArtCraft wants to rebuild everything →](https://latent.space/p/artcraft?ref=genaisecretsauce.com)

Developer Tools

## Developer Tools & Infrastructure

### Google Antigravity Orchestrator - two Gemini models share one coding task

*Previously: [November 12](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-11-12/) \- Google put its fast Gemini 4 Hydrogen model into Search for everyone.*

**Today:** Hydrogen arrived in Antigravity, Google's AI coding app, along with an Orchestrator that mixes models mid-task.

- **Plans with one model, edits with another** \- Gemini 4 Carbon explores and writes the plan, then faster Hydrogen makes the changes and runs tests
- **Near-Carbon results for less** \- 24% less usage quota and 19% faster on Google's own tests
- **Who gets it** \- Hydrogen for paid Google AI Pro and Ultra subscribers, with Orchestrator options split across those plans

[Google Antigravity: Hydrogen and Orchestrator →](https://antigravity.google/blog/gemini-4-hydrogen-in-google-antigravity?ref=genaisecretsauce.com)

### Self-hosted Claude Managed Agents - Anthropic's agent service on your own servers

**Anthropic** opened a private beta that lets companies run its whole agent service in their own cloud or data center.

- **Everything stays inside** \- the server, the session history and the machines where agents run code
- **Passwords never enter the agent's workspace** \- a go-between adds them only when a tool needs them
- **Every step is logged** \- and can be exported to a company's own monitoring tools

[Claude: Self-hosted Managed Agents →](https://claude.com/resources/articles/self-hosted-claude-managed-agents?ref=genaisecretsauce.com)

### Oddboro - manage many AI agents through one "office manager"

**Oddboro** is a free, open-source desktop app that shows your AI agents as a small pixel-art office.

**Try it:** [Nate's Newsletter: Oddboro launch](https://natesnewsletter.substack.com/p/oddboro-manage-multiple-ai-agents-launch)

- **One question instead of six windows** \- ask the AI office manager "what happened while I was out, and what needs me?"
- **Agents hand work to each other** \- with access limited to the folders and tools you allow
- **Email drafts wait for you** \- Gmail replies are drafted but not sent without a human

### llama.cpp 1.0 - the engine behind home AI apps promises stability

**llama.cpp** is free software that runs AI models on ordinary computers and phones, and it reached version 1.0 after about 3.5 years.

**Try it:** [Hugging Face: llama.cpp v1.0.0](https://huggingface.co/blog/ggml-org/llama-cpp-v1?ref=genaisecretsauce.com)

- **A stability promise** \- developers building on it get clear warnings before anything breaks
- **Over 150 model designs supported** \- across Linux, Mac, Windows, Android and iPhone

### Claude for Manufacturing - an AI helper for factory teams

**Anthropic** released an add-on for business plans with 17 ready-made skills for plant work.

- **From quotes to breakdowns** \- it drafts job quotes, investigates failed inspections and diagnoses stopped machines from sensor data and manuals
- **Connects to factory software** \- including Autodesk Fusion, Infor, Microsoft 365 and Oracle NetSuite
- **A human signs off** \- on anything that reaches a customer or supplier, and Claude does not run equipment

[Claude: Claude for Manufacturing →](https://claude.com/resources/articles/claude-for-manufacturing?ref=genaisecretsauce.com)

Research & Models

## Research & Models

### A new open AI model aims for top coding skills on a laptop

- **Edge0-145B is huge but light to run** \- only about 10 billion of its 145 billion parts work at once (a "mixture of experts" design)
- **29GB of memory at peak** \- and over 80 tokens (word pieces) per second on a high-end laptop, per Edge0
- **Beats similar free models on coding tests** \- by its own measurements, with the full technical report still to come
- **Also trending** \- it is the first pick in today's Hugging Face section (see section 13)

[Edge0: introducing Edge0-145B →](https://edge0.ai/blog/introducing-edge0-145b-frontier-ai-capabilities-on-your-own-computer?ref=genaisecretsauce.com)

### AI assistants often fix the question you asked instead of your real problem

- **Ai2's XYBench has 8,115 questions** \- each where the user asks about the wrong approach to their real goal (like thickening frosting when the cupcakes are just too warm)
- **All five models tested leaned toward the literal question** \- on coding questions they solved the real goal only 59-71% of the time
- **Humans did better** \- every model trailed human answers at spotting the user's mistake

[Hugging Face: Ai2 XYBench →](https://huggingface.co/blog/allenai/xybench?ref=genaisecretsauce.com)

### A big AI model was trained in 4-bit math from start to finish

*Previously: [November 6](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-11-06/) \- Thinking Machines released Inkling 2 Small, saying it did its main training in about three days.*

**Today:** The lab explained how: the heaviest calculations, in the model's expert layers, used ultra-compact 4-bit numbers (about a quarter of the usual size).

- **The lab believes it is the largest model trained this way** \- with the same compact numbers used from the first step to the final version
- **Faster training** \- 8% faster in the main training run and 20% faster in later reinforcement-learning training, for about 2% more total computing work
- **The new bottleneck is the wiring** \- chips now wait on data moving between them when passing work between experts, not on the math

[Thinking Machines: End-to-end 4-bit training →](https://thinkingmachines.ai/blog/end-to-end-4-bit-training?ref=genaisecretsauce.com)

### Anthropic shared how it tests whether AI is safe

- **Refusing everything is not "safe"** \- Anthropic grades whether a response was appropriate, not just whether the AI said no
- **Simple one-message tests are maxed out** \- recent models pass 94-98%, so the real signal comes from long conversations where harm builds slowly
- **Settings change scores a lot** \- Claude Fable 5.1 scored 60% on a self-harm test through raw developer access and 94% in the Claude app after a fix, two different setups

[Anthropic: Building strong safety evaluations →](https://www.anthropic.com/research/building-strong-safety-evaluations?ref=genaisecretsauce.com)

Business & Industry

## Business & Industry

### Banks are keeping Chinese and Hong Kong investors out of Anthropic's IPO

*Previously: [November 12](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-11-12/) \- Anthropic's stock offering was reportedly about six times oversubscribed.*

**Today:** Several banks on the deal reportedly turned away orders from mainland China and Hong Kong, citing US defense export rules, per Bloomberg via The Decoder.

- **Goldman Sachs and Morgan Stanley lead the deal** \- with about 20 other banks involved
- **Pricing still expected November 18** \- with trading on the Nasdaq under the ticker ANT

[The Decoder: Banks bar Chinese investors from Anthropic IPO →](https://the-decoder.com/banks-bar-chinese-and-hong-kong-investors-from-anthropic-ipo-over-us-export-controls?ref=genaisecretsauce.com)

### CoreWeave is borrowing $4 billion against its Meta contracts

*Previously: [November 11](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-11-11/) \- CoreWeave began asking customers to pay up front.*

**Today:** The AI cloud company is seeking a $4 billion loan backed by its contracts with Meta, Semafor reports.

- **About half the interest rate** \- a target of about 7.5%, against nearly 16% on its regular bonds
- **Nvidia ties make it jumpy** \- its stock has swung all week, since Nvidia is an early backer and key supplier

[Semafor: CoreWeave looks to raise $4B →](https://www.semafor.com/article/11/13/2026/coreweave-tunes-out-nvidia-noise-looks-to-raise-4b?ref=genaisecretsauce.com)

### OpenAI launched AI tools that test and patch company security

- **Daybreak Red Team** \- continuously probes a company's websites for weaknesses, in early access
- **Daybreak Defense Factory** \- finds and fixes security holes at scale, with an engineer approving each fix
- **No pricing given** \- both run on OpenAI's GPT-6 Cyber model

[The Register: OpenAI launches Daybreak security products →](https://www.theregister.com/security/2026/11/12/openai-launches-security-products-tells-orgs-theyre-doomed-if-they-dont-use-them/5309500?ref=genaisecretsauce.com)

### Apple's new Apple TV 4K brings its new Siri AI to the living room

- **$299** \- twice what the previous model cost at this point last year
- **Siri can recognize up to six household members** \- for personal recommendations, and you no longer need the remote to talk to it
- **Older owners may not need it** \- recent Apple TV 4K models get Siri AI through a free software update

[MacRumors: Apple TV 4K and Siri AI hands-on →](https://www.macrumors.com/2026/11/12/apple-tv-4k-siri-hands-on?ref=genaisecretsauce.com)

### ElevenLabs launched a meeting-notes and dictation app for teams

- **ElevenAI records meetings and recalls old conversations live** \- and turns speech into clean text in any app in 90+ languages
- **Aimed at rivals** \- Fireflies, Granola and Wispr Flow, with early-access signup and no pricing yet

[TechCrunch: ElevenLabs' new app →](https://techcrunch.com/2026/11/13/elevenlabs-new-app-combines-meeting-note-taking-and-dictation-for-teams?ref=genaisecretsauce.com)

Education

## GenAI in Education

### Schools that switch off Google's AI answers lose Google Search entirely

- **The policy** \- schools can turn off AI features, but turning off AI Overviews (the AI summaries at the top of results) removes Search altogether
- **Growing anger** \- New York City's mayor accused Google of using "monopoly power," and Sen. Ed Markey demanded a reversal
- **One theory** \- a schools-only off switch could undercut Google's argument in court that Search and AI Overviews are one product, though experts NPR asked were not fully convinced

[NPR: Why is Google cutting off Search from AI-free schools? →](https://text.npr.org/nx-s1-6029408?ref=genaisecretsauce.com)

### Fake AI papers are appearing under famous researchers' names

- **Dozens attributed to top AI scientists** \- including Fei-Fei Li, Christopher Manning and Stuart Russell, who did not write them
- **Three sister journals with the same editor** \- websites registered in 2026 but showing archives going back years
- **The real risk** \- PhD students may see famous names and think these are legitimate places to publish

[Inside Higher Ed: AI slop papers under academics' names →](https://www.insidehighered.com/news/faculty/books-publishing/2026/11/13/ai-slop-papers-published-under-prominent-academics-names?ref=genaisecretsauce.com)

### A writing teacher says the AI fight is really about values

- **After 100+ talks on AI and writing** \- columnist John Warner says the best progress comes where students, teachers and leaders work on it together
- **His core belief** \- writing is how people figure out what they think, so AI that skips that struggle hurts learning
- **Students are pushing back too** \- Princeton students recently asked their university not to hand everyone free ChatGPT

[Inside Higher Ed: What do we value about teaching writing? →](https://www.insidehighered.com/opinion/columns/just-visiting/2026/11/13/what-if-anything-do-we-value-about-teaching-writing?ref=genaisecretsauce.com)

### OpenAI put $2 million into research on AI and children's safety

- **Announced at the Paris Peace Forum** \- for a research network studying AI's risks and benefits for kids
- **Joined a voluntary charter** \- also backed by Anthropic and Microsoft, with no enforcement
- **Critics call it small** \- next to the companies' size, and say it skips harms like emotional attachment to chatbots

[OpenAI: Supporting research on AI and children's safety →](https://openai.com/index/children-first-declaration?ref=genaisecretsauce.com)

Surprising

## Surprising & Under-the-Radar

### A third of US chatbot users insult their AI, and it doesn't help

An analysis of 8 million messages found 33% of US users swear at or insult chatbots. It backfires: the AI fixed the original problem 24% of the time after an insult, against 35% when users stayed polite. Anthropic's new terms, effective November 12, ban "sustained and needless abusive" behavior toward its models.

[The Register: a third of US users insult chatbots →](https://www.theregister.com/ai-and-ml/2026/11/12/a-third-of-us-ai-chatters-are-hot-at-bots-hurling-insults-at-llms/5309506?ref=genaisecretsauce.com)

### The FAA expanded an AI traffic tool nationwide in the same week as an unrelated ground stop

The FAA said on November 10 that its AI tool for planning air traffic flow is expanding to all 22 US regional control centers. Separately, on November 13 Flightradar24 reported that the FAA had ground stopped every US airport over an equipment outage. Nothing links the outage to the AI tool, which only gives advice that controllers can accept or reject.

[Ars Technica: FAA expands AI air traffic advice →](https://arstechnica.com/ai/2026/11/faa-expands-ai-advice-to-manage-air-traffic-congestion-across-entire-us-airspace?ref=genaisecretsauce.com)[Flightradar24: all US airports ground stopped →](https://twitter.com/flightradar24/status/2121347654714077605?ref=genaisecretsauce.com)

### A two-hour Claude outage was one of the day's most-discussed tech stories

Users couldn't sign in to Claude and Claude Code for about two hours on November 13\. The outage drew 482 points and 382 comments on Hacker News, a sign of how many people's workdays now depend on one AI service.

[Claude status: elevated errors incident →](https://status.claude.com/incidents/nn11wk485zfn?ref=genaisecretsauce.com)

### Debate: is superintelligence coming by 2030?

Podcaster Dwarkesh Patel now expects superintelligence by 2030, citing AI already speeding up AI research at top labs. Supporters point to math, where AI went from failing four-digit addition in 2023 to top-level contest results. Skeptics counter that progress could still stall on chip supply and real-world testing.

[Dwarkesh Patel: The avalanche is coming for AI development →](https://www.dwarkesh.com/p/rsi?ref=genaisecretsauce.com)

### Debate: can AI-made shows help Black actors, or do they replace them?

Mansa's founders say cheap AI series prove demand that can unlock live-action roles. Critics say a company built to support Black talent should not start by cutting the actors out.

[WIRED: Mansa's AI microdramas →](https://www.wired.com/story/black-roles-are-disappearing-in-hollywood-this-startup-is-using-ai-microdramas-to-fight-back?ref=genaisecretsauce.com)

Worth Watching

## Signals to Track

[![Worth Watching](https://genaisecretsauce.com/content/images/2026/11/section-worth-watching-2026-11-13.png)](https://genaisecretsauce.com/content/images/2026/11/section-worth-watching-2026-11-13.png) 

01

### Gemini can now read your spending

Google now lets you hand its AI assistant your spending history, for advice only.

US Gemini users can now connect Google Wallet, and bank or card accounts through Plaid, so the assistant can answer money questions from real spending. It can find forgotten subscriptions and draw budget charts, but Google says it cannot move money. If this catches on, AI assistants become a default place to manage household budgets.

[Google: Connect Google Wallet to Gemini →](https://blog.google/innovation-and-ai/products/gemini-app/google-wallet-gemini-app?ref=genaisecretsauce.com)

02

### Nvidia privately said chip-location tracking is doable

The chipmaker quietly told a congressman that anti-smuggling tracking is doable.

Rep. Bill Foster told MIT Technology Review that Nvidia agreed in technical talks that chip-location tracking is feasible, though its industry group had called it "potentially infeasible" in March. One option would have each chip report its location when it comes online. If tracking becomes law, advanced chips could come with built-in location checks.

[MIT Technology Review: Nvidia chip tracking →](https://www.technologyreview.com/2026/11/12/1148483/nvidia-chip-tracking?ref=genaisecretsauce.com)

03

### Google DeepMind staff asked the UK government to force a union vote

AI researchers are trying to unionize one of the world's top AI labs.

Two UK unions filed with the official body that settles union disputes, seeking a binding vote for about 1,000 Google DeepMind staff. Organizers object to military contracts and hope to write a playbook for OpenAI and Anthropic staff in the UK. If it works, AI workers could gain a formal say in what their labs build.

[WIRED: DeepMind staff escalate union fight →](https://www.wired.com/story/google-deepmind-staff-escalate-their-union-fight?ref=genaisecretsauce.com)

04

### The White House is lining up AI makers behind shared standards

Tech CEOs are drafting common rules while the government tests models before release.

The White House AI task force met on cyber threats and heard updates on pre-release testing of AI models by the government's NIST standards agency. Its leader said the companies that signed the White House AI accord are drafting shared industry standards, and more firms may join. That could mean the AI tools you use get government-reviewed before launch.

[Semafor: Trump's AI task force discusses cyber risks →](https://www.semafor.com/article/11/13/2026/trumps-ai-task-force-discusses-cyber-risks?ref=genaisecretsauce.com)

GitHub Trending

## Top Repos Today

Tools that make AI coding helpers cheaper, easier to set up and able to drive other apps led GitHub's new AI projects today. GitHub's main trending page listed only 12 projects, so we also drew on its Python, TypeScript, Go and Rust trending lists. We skipped projects featured in the past week that are still trending, including Google Chrome's Modern Web Guidance (up 1,925 stars today), free-claude-code, OpenAI's Codex Security tool, yesterday's #1 CopilotKit OpenBot, Qwen Code, NVIDIA's SkillSpector, Tencent's LoopForge, Alibaba's code reviewer, Multica and the stock-analysis bot. We also left out a collection of leaked chatbot instructions, an "uncensored" AI studio and a stock-trading skills pack. **Today's leader is CC Switch, a one-click settings manager for AI coding tools, which added 315 stars.**

#1

### [farion1231/cc-switch](https://github.com/farion1231/cc-switch?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +315 · 📦 **Total:** 154,228  
📜 **License:** MIT · 👤 **By:** farion1231 (individual developer)  
🎯 **Time to value:** 10 minutes

**What it is:** A free desktop app for Windows, Mac and Linux that manages the settings of AI coding tools such as Claude Code, Codex, Gemini CLI and OpenCode in one place. You can switch which AI company powers each tool in one click, with more than 90 ready-made presets, instead of hand-editing settings files. It also manages add-ons such as skills and MCP connections (plug-ins that link an AI to other apps), and can switch to a backup provider automatically if one fails. **Why you'd want it:** Save time if you juggle several AI coding tools or AI accounts and keep breaking their settings files.

| ✓ Pros                                               | ✗ Cons                                                                  |
| ---------------------------------------------------- | ----------------------------------------------------------------------- |
| Covers more than a dozen coding tools                | Its page promotes many paid third-party AI resellers as sponsors        |
| Automatic switch to a backup provider when one fails | Official Claude and Codex accounts are limited in some routing features |
| Free and open source                                 | Changes settings files that other tools depend on                       |

[GitHub - farion1231/cc-switch: A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode,DeepSeek Harness and more. Only official website: ccswitch.ioA cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode,DeepSeek Harness and more. Only official website: ccswitch.io - farion1231/cc-switch![](https://genaisecretsauce.com/content/images/icon/favicon-65728c52-f200-4bb7-8e4c-574c31312396.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/cc-switch-77e94d6e-fee1-44c6-a0e1-b250c1420840.png)](https://github.com/farion1231/cc-switch?ref=genaisecretsauce.com)

#2

### [Gentleman-Programming/gentle-ai](https://github.com/Gentleman-Programming/gentle-ai?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +242 · 📦 **Total:** 9,130  
📜 **License:** MIT · 👤 **By:** Gentleman Programming (organization)  
🎯 **Time to value:** 15 minutes

**What it is:** A setup tool that adds memory, a step-by-step workflow and test-first habits to AI coding agents you already use, such as Claude Code, Cursor, Codex and 14 others. Its memory feature lets the agent write down what it learns about your project, so you do not have to re-explain decisions every morning. **Why you'd want it:** Make your coding agent more consistent and less forgetful without switching to a new agent.

| ✓ Pros                                     | ✗ Cons                                       |
| ------------------------------------------ | -------------------------------------------- |
| Works with 17 existing agents              | Adds its own workflow and terms to learn     |
| Backs up your settings before every change | Windows setup needs the Go programming tools |
| Free, with no paid tier                    | Opinionated about how you should work        |

[GitHub - Gentleman-Programming/gentle-ai: Gentle-AI configures the AI coding agents you already use: Claude Code, Cursor, OpenCode, Codex, Pi, and more. Choose persistent memory, Organic-Driven Development, curated skills, MCP servers, personas, and optional bounded review. Open source, no agent lock-in.Gentle-AI configures the AI coding agents you already use: Claude Code, Cursor, OpenCode, Codex, Pi, and more. Choose persistent memory, Organic-Driven Development, curated skills, MCP servers, per…![](https://genaisecretsauce.com/content/images/icon/favicon-5c7aeff3-7f8c-4a8b-bdf9-5ca7e0f9056a.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/37f50cac-a47b-4019-bc19-53b7fe060726-62f1d22c-aad1-4f4b-bfb5-95ee56d8a87b.png)](https://github.com/Gentleman-Programming/gentle-ai?ref=genaisecretsauce.com)

#3

### [kajisho5/ffmpeg-skill](https://github.com/kajisho5/ffmpeg-skill?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +168 · 📦 **Total:** 5,433  
📜 **License:** MIT · 👤 **By:** kajisho5 (individual developer)  
🎯 **Time to value:** 10 minutes

**What it is:** An add-on "skill" that teaches AI coding agents such as Claude Code, Cursor and Codex to edit video on your own computer with FFmpeg (a free, widely used video tool). It offers 42 tools for jobs like cutting, captions, removing silences, resizing for TikTok and fixing loudness, and the agent checks its own output after each step. **Why you'd want it:** Have an AI assistant do routine video editing chores for you, without uploading your footage anywhere.

| ✓ Pros                                    | ✗ Cons                                           |
| ----------------------------------------- | ------------------------------------------------ |
| Works offline, with no account or API key | You still need to install FFmpeg first           |
| Agent checks its own results              | Spoken-word captions need a separate speech tool |
| One-command install                       | Command-line based, no visual editor             |

[GitHub - kajisho5/ffmpeg-skill: Local video editor for coding agents: 42 FFmpeg tools with a machine-readable contract and an MCP server, for Claude Code, Cursor and Codex. Cut, captions, reframe, loudness, platform exports. No cloud, no API keys.Local video editor for coding agents: 42 FFmpeg tools with a machine-readable contract and an MCP server, for Claude Code, Cursor and Codex. Cut, captions, reframe, loudness, platform exports. No c…![](https://genaisecretsauce.com/content/images/icon/favicon-49e98dfc-1de4-4731-bc63-5e496b379440.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/12a63f4a-bbc2-4ce1-820e-f2bc459f35ec-279fb973-3460-4fe2-9f71-d4017618b65b.png)](https://github.com/kajisho5/ffmpeg-skill?ref=genaisecretsauce.com)

#4

### [ollama/ollama](https://github.com/ollama/ollama?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +141 · 📦 **Total:** 186,895  
📜 **License:** MIT · 👤 **By:** Ollama (organization)  
🎯 **Time to value:** 5 minutes

**What it is:** One of the most popular free apps for downloading and running open AI models on your own computer, including Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen and Gemma. It handles the technical setup and gives other apps a simple way to talk to the model. **Why you'd want it:** Try open AI models privately and offline, with no per-use fees.

| ✓ Pros                              | ✗ Cons                                         |
| ----------------------------------- | ---------------------------------------------- |
| Very easy to install and use        | Big models still need a powerful computer      |
| Huge library of ready-to-run models | Fewer fine-tuning knobs than lower-level tools |
| Many other apps plug into it        | Downloads can be tens of gigabytes             |

[GitHub - ollama/ollama: Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models. - ollama/ollama![](https://genaisecretsauce.com/content/images/icon/favicon-7a6bec5e-5ec3-418d-ba20-c19b72d6b9be.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/ollama-c70c9751-38ac-43a6-a2dd-9d28ed8d2496.png)](https://github.com/ollama/ollama?ref=genaisecretsauce.com)

#5

### [ahujasid/mcp-for-blender](https://github.com/ahujasid/mcp-for-blender?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +136 · 📦 **Total:** 34,324  
📜 **License:** MIT · 👤 **By:** ahujasid (individual developer)  
🎯 **Time to value:** 15 minutes

**What it is:** A community plug-in, formerly called blender-mcp, that lets AI assistants such as Claude, Codex and Cursor control Blender, the free 3D design app. You describe what you want in plain words, and the AI builds and changes the 3D scene for you. It is not made by the Blender Foundation. **Why you'd want it:** Create 3D scenes and models by asking, even if you have never learned Blender's many menus.

| ✓ Pros                                 | ✗ Cons                                         |
| -------------------------------------- | ---------------------------------------------- |
| One-command setup for many AI apps     | Not an official Blender product                |
| Works with the AI model of your choice | Setup runs an install script from the internet |
| Free and open source                   | AI-made 3D work still needs touching up        |

[GitHub - ahujasid/mcp-for-blender: Community plugin to control Blender 3D with any LLM of your choice. Not affiliated with the official Blender Foundation.Community plugin to control Blender 3D with any LLM of your choice. Not affiliated with the official Blender Foundation. - ahujasid/mcp-for-blender![](https://genaisecretsauce.com/content/images/icon/favicon-2addc42e-4491-4db6-bba3-570ba25e835b.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/mcp-for-blender-7b8f72d5-e8d2-4fee-b890-c6377b1c7244.png)](https://github.com/ahujasid/mcp-for-blender?ref=genaisecretsauce.com)

#6

### [anthropics/claude-plugins-official](https://github.com/anthropics/claude-plugins-official?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +130 · 📦 **Total:** 47,194  
📜 **License:** Apache-2.0 · 👤 **By:** Anthropic (organization)  
🎯 **Time to value:** 5 minutes

**What it is:** Anthropic's official collection of plug-ins for Claude Code, which add extra commands, helpers and app connections. Anthropic says outside developers now submit plug-ins through Claude's website instead, and this collection is being narrowed to Anthropic's own plug-ins. **Why you'd want it:** A safer starting point for adding features to Claude Code than random add-ons from strangers.

| ✓ Pros                                       | ✗ Cons                                                       |
| -------------------------------------------- | ------------------------------------------------------------ |
| Maintained by Anthropic                      | Only useful if you use Claude Code                           |
| Installs with one command inside Claude Code | Anthropic warns it cannot vet everything a plug-in includes  |
| Clear standard structure for builders        | Third-party listings are being phased out of this collection |

[GitHub - anthropics/claude-plugins-official: Official, Anthropic-managed directory of high quality Claude Code Plugins.Official, Anthropic-managed directory of high quality Claude Code Plugins. - anthropics/claude-plugins-official![](https://genaisecretsauce.com/content/images/icon/favicon-92673d47-6024-4af4-b00b-c2ae75d8c79d.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/claude-plugins-official-799d2058-424e-4eb0-8a93-7f52cdaec399.png)](https://github.com/anthropics/claude-plugins-official?ref=genaisecretsauce.com)

#7

### [rtk-ai/rtk](https://github.com/rtk-ai/rtk?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +123 · 📦 **Total:** 86,494  
📜 **License:** Apache-2.0 · 👤 **By:** RTK (organization)  
🎯 **Time to value:** 5 minutes

**What it is:** A small program that shrinks the text AI coding agents read from everyday commands, such as test results and code changes, by up to 90%. For example, the agent reads a 3-line summary of your project's status instead of 40 lines. Less text means fewer tokens (the units AI companies bill by), and a built-in report shows what you saved. **Why you'd want it:** Cut part of your AI coding bill and help agents focus, with almost no setup.

| ✓ Pros                          | ✗ Cons                                                    |
| ------------------------------- | --------------------------------------------------------- |
| Install once and forget it      | A 90% cut in command output is not a 90% cut in your bill |
| Single file with no extra parts | Pushes its paid Pro product in its reports                |
| Shows exactly what it saved     | Another tool named "rtk" exists, so install the right one |

[GitHub - rtk-ai/rtk: CLI proxy that reduces LLM token consumption by 60-90% on common dev commands. Single Rust binary, zero dependenciesCLI proxy that reduces LLM token consumption by 60-90% on common dev commands. Single Rust binary, zero dependencies - rtk-ai/rtk![](https://genaisecretsauce.com/content/images/icon/favicon-fe6667b8-c40d-4357-8b17-d628803fdf58.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/rtk-ebd8ee08-29e4-4c26-a36f-a3dcb3b7954f.png)](https://github.com/rtk-ai/rtk?ref=genaisecretsauce.com)

HuggingFace Trending

## Top Models Today

Almost all of Hugging Face's top 20 trending models were featured here in the past week, including MiniMax's H3-Image, Nous Research's Arete-27B, Lightricks' LTX-2.5, Qwen3.8-27B, Kimi K3.1 Pro, Thinking Machines' Inkling-2-Small, Viggle Remake, Meta's SAM 3 and yesterday's hyperflow-h3-flash. We skipped those, several "uncensored" copies of other models and repackaged copies, then looked further down Hugging Face's trending ranking. **Today's picks include a 145-billion-parameter coding agent that runs on one gaming graphics card and a 27-billion-parameter model improved without any extra training.**

#1

### [Edge0/Edge0-145B-NVFP4](https://huggingface.co/Edge0/Edge0-145B-NVFP4?ref=genaisecretsauce.com)

A very large coding and agent model squeezed to run on a single 24 GB gaming graphics card.

📥 **Downloads (30d):** 14,134 · 📜 **License:** Edge0 Community Model License (custom)  
👤 **By:** Edge0 · 🎯 **Task:** Text generation (coding and agents)  
📐 **Size:** 145B total (about 10B active)

**What it is:** A 145-billion-parameter model built on Qwen3.8-Flash-Next, stored in a compressed 4-bit format so it fits on home hardware. It uses a "mixture of experts" design (only part of the model works on each word) and keeps most of itself in regular computer memory, pulling pieces onto the graphics card as needed. Edge0 reports 98.3 tokens per second on an RTX 5090 D card and a score of 86.5 on Terminal-Bench 2.1 (a test of agents doing real tasks in a computer terminal). **Why you'd want it:** Run a big coding agent at home instead of paying per use for a cloud model.

| ✓ Pros                           | ✗ Cons                                                               |
| -------------------------------- | -------------------------------------------------------------------- |
| Runs on one 24 GB graphics card  | Also needs about 38 GB of regular computer memory                    |
| Fast replies for its size        | Needs Edge0's own engine, built for NVIDIA cards on Linux or Windows |
| Free for personal and output use | Selling the model itself as a service needs Edge0's permission       |

[Edge0/Edge0-145B-NVFP4 · Hugging FaceText Generation model with 5B parameters. Download, explore files and run it on Hugging Face.![](https://genaisecretsauce.com/content/images/icon/favicon-9044bcb9-9bee-4584-b7c1-9291c4afa627.ico)![](https://genaisecretsauce.com/content/images/thumbnail/Edge0-145B-NVFP4-50de95cc-ef47-41cb-a6da-8be31d6e8ea1.png)](https://huggingface.co/Edge0/Edge0-145B-NVFP4?ref=genaisecretsauce.com)

#2

### [ConceptCountry/qwen-concept-27b](https://huggingface.co/ConceptCountry/qwen-concept-27b?ref=genaisecretsauce.com)

Qwen3.8-27B with nine internal pieces rewritten, so it thinks less and scores higher.

📥 **Downloads (30d):** 63,252 · 📜 **License:** Apache-2.0  
👤 **By:** ConceptCountry · 🎯 **Task:** Text generation  
📐 **Size:** 27B

**What it is:** A modified Qwen3.8-27B where the makers changed nine of the model's weight tables directly, using no extra training data at all. They report a 4.3-point gain on GPQA Diamond (hard science questions) and a 13-point gain on LiveCodeBench v6 (a coding test), while the model's "thinking" text shrank by 27% to 54%. They also show that a random change of the same size does nothing. **Why you'd want it:** Get answers that are slightly better and cheaper to produce from a popular mid-size model.

| ✓ Pros                                        | ✗ Cons                                         |
| --------------------------------------------- | ---------------------------------------------- |
| Permissive license for commercial use         | Results are the makers' own                    |
| Versions from about 16 GB to 52 GB            | Compressed versions were not separately tested |
| Less "thinking" means faster, cheaper answers | A modified model, not a new one                |

[ConceptCountry/qwen-concept-27b · Hugging FaceMachine learning model with 27B parameters. Download, explore files and run it on Hugging Face.![](https://genaisecretsauce.com/content/images/icon/favicon-046135fb-9a14-450c-897d-4b088621fa96.ico)![](https://genaisecretsauce.com/content/images/thumbnail/qwen-concept-27b-e53b58cd-ed0c-4b09-a3e0-47afb556b4ea.png)](https://huggingface.co/ConceptCountry/qwen-concept-27b?ref=genaisecretsauce.com)

#3

### [higurashi-ai/Haru-4B](https://huggingface.co/higurashi-ai/Haru-4B?ref=genaisecretsauce.com)

A voice model that listens and talks at the same time, like a phone call, in nine languages.

📥 **Downloads (30d):** 165 · 📜 **License:** Apache-2.0  
👤 **By:** Higurashi AI · 🎯 **Task:** Audio-to-audio (spoken conversation)  
📐 **Size:** about 5.3B by file count (named 4B)

**What it is:** A "full-duplex" voice model, meaning it hears you and speaks back at the same time instead of waiting for you to finish. It works in small slices of 133 milliseconds and speaks English, Spanish, Portuguese, German, French, Chinese, Japanese, Russian and Korean. A short text card sets its character, and a 3-second to 3-minute voice clip sets how it sounds. **Why you'd want it:** Experiment with natural, interruptible voice assistants on a single graphics card or a Mac.

| ✓ Pros                                             | ✗ Cons                                                          |
| -------------------------------------------------- | --------------------------------------------------------------- |
| Runs on a graphics card with about 12 GB of memory | Released as a research base model, not a finished assistant     |
| Mac app included                                   | Ships with no voices, so you must supply one you have rights to |
| Nine languages                                     | Bluetooth headsets add noticeable delay                         |

[higurashi-ai/Haru-4B · Hugging FaceAudio-to-Audio model with 5B parameters. Download, explore files and run it on Hugging Face.![](https://genaisecretsauce.com/content/images/icon/favicon-22331df4-7a3f-46e7-9e12-935033b41a63.ico)![](https://genaisecretsauce.com/content/images/thumbnail/Haru-4B-d511942e-a230-4ac4-a790-fc83f40b7efa.png)](https://huggingface.co/higurashi-ai/Haru-4B?ref=genaisecretsauce.com)

#4

### [KRAFTON/Raon-1-nano](https://huggingface.co/KRAFTON/Raon-1-nano?ref=genaisecretsauce.com)

A small Korean and English model from game company KRAFTON, built to run on phones and laptops.

📥 **Downloads (30d):** 806 · 📜 **License:** Apache-2.0  
👤 **By:** KRAFTON AI · 🎯 **Task:** Text generation  
📐 **Size:** 3.2B (about 1.5B effective)

**What it is:** A compact chat model designed to follow instructions, keep track of a conversation and call tools (use other apps) for tasks such as office work and phone control. It was shrunk and trained from KRAFTON's larger Raon-1 mini preview model, and it can read up to 262,144 tokens at once. Ready-made versions exist for running on Apple computers and other devices. **Why you'd want it:** A free, small assistant model for apps that need to run on the device, especially in Korean.

| ✓ Pros                              | ✗ Cons                                        |
| ----------------------------------- | --------------------------------------------- |
| Permissive license                  | Mainly Korean and English                     |
| Very long input window for its size | Small models make more mistakes than big ones |
| Formats ready for local use         | Benchmark wins are KRAFTON's own              |

[KRAFTON/Raon-1-nano · Hugging FaceText Generation model with 4B parameters. Download, explore files and run it on Hugging Face.![](https://genaisecretsauce.com/content/images/icon/favicon-232a9335-7df3-4efd-afe7-fed060f821c7.ico)![](https://genaisecretsauce.com/content/images/thumbnail/Raon-1-nano-321ca412-44f5-48fe-87fc-53b192b3e2f4.png)](https://huggingface.co/KRAFTON/Raon-1-nano?ref=genaisecretsauce.com)

#5

### [nex-agi/Nex-N2.6](https://huggingface.co/nex-agi/Nex-N2.6?ref=genaisecretsauce.com)

An agent model tuned to stop repeating itself and to use only the tools it is given.

📥 **Downloads (30d):** 410 · 📜 **License:** Apache-2.0  
👤 **By:** Nex-AGI · 🎯 **Task:** Image-text-to-text (computer use and coding)  
📐 **Size:** 321B by file count (compressed FP8 version)

**What it is:** A model built on Zhipu's GLM-5.3-Flash and tuned for computer use, coding and long tasks with tools. Nex-AGI reports a score of 83.0 on OSWorld (a test of AI using a real computer), up from 78.0 for the base model, and says repetitive reasoning fell from 16.0% to 1.6% compared with its previous version. **Why you'd want it:** Teams building agents that click around computers may get fewer loops and wrong tool calls.

| ✓ Pros                                     | ✗ Cons                                       |
| ------------------------------------------ | -------------------------------------------- |
| Permissive license                         | Data-center hardware needed to self-host     |
| Accepts images as well as text             | Reliability numbers come from internal tests |
| Works with the popular SGLang serving tool | Not better than its base on every test       |

[nex-agi/Nex-N2.6 · Hugging FaceImage-Text-to-Text model with 321B parameters. Download, explore files and run it on Hugging Face.![](https://genaisecretsauce.com/content/images/icon/favicon-05abb2a0-92fe-4c5d-aee9-ed1deeab6841.ico)![](https://genaisecretsauce.com/content/images/thumbnail/Nex-N2.6-4e376f5b-478e-41d5-99ee-3314105176b7.png)](https://huggingface.co/nex-agi/Nex-N2.6?ref=genaisecretsauce.com)

Product Hunt

## AI Launches Today

This week's late AI launches on Product Hunt focused on testing agents safely, making content faster and plugging chatbots into everyday work tools. We skipped launches covered in the past week that are still on the leaderboard, including Claude Meetings, Spacefast by Automattic and Gemini 4 Hydrogen. Upvote counts are Product Hunt's scores when we checked, so final counts may be higher. **Plurai Simulation, which builds a practice world for testing AI agents, led November 12 with 409 upvotes.**

### [Plurai Simulation](https://www.producthunt.com/products/plurai?ref=genaisecretsauce.com)

A simulated world that makes your agent improve on every run

🔥 **Upvotes:** 409 · 👤 **By:** Plurai (launch team includes fmerian, Maria Anosova and Tammy Wolfson)  
💰 **Pricing:** Free options · 🏷 **Category:** AI agents and evaluation

Plurai Simulation builds a "digital twin" (a realistic copy) of the world your AI agent works in, from a single command in Claude Code, Codex or GitHub Copilot. The copy includes made-up customers, documents and fake versions of your tools and databases, all kept away from your real systems. The goal is to catch agent mistakes before real users see them. **Verdict:** A smart safety net for teams shipping AI agents, as long as the fake world matches the real one closely.

[View on Product Hunt →](https://www.producthunt.com/products/plurai?ref=genaisecretsauce.com) 

### [YouArt Desktop App](https://www.producthunt.com/products/youart?ref=genaisecretsauce.com)

Codex for creative team

🔥 **Upvotes:** 349 · 👤 **By:** YouArt (Y Combinator-backed; maker Harry Zheng)  
💰 **Pricing:** 50 free credits for new users; paid plans not yet set · 🏷 **Category:** AI video editor

YouArt turns AI models, tools and prompts into reusable workflows for making ads, videos and other creative assets. The new desktop app can read your local project files, connect to Slack and Gmail, and even operate creative software such as After Effects and CapCut by itself. Teams can still collaborate on the web version. **Verdict:** Promising for busy marketing teams, but pricing is still being worked out, so watch the credit costs.

[View on Product Hunt →](https://www.producthunt.com/products/youart?ref=genaisecretsauce.com) 

### [Templify AI](https://www.producthunt.com/products/templify?ref=genaisecretsauce.com)

Powerful content editor for people who aren't editors

🔥 **Upvotes:** 339 · 👤 **By:** Templify (co-founder Yana Abazher)  
💰 **Pricing:** Free, with optional Pro subscription · 🏷 **Category:** Video and photo editing

Templify is an iPhone and Mac editor that turns your clips and photos into Reels, TikToks, YouTube Shorts and collages using more than 1,200 designer templates, with no editing timeline to learn. Its AI tools add captions, sync transitions to the music beat and remove backgrounds. **Verdict:** A quick win for small businesses and creators who want polished social posts without learning video editing.

[View on Product Hunt →](https://www.producthunt.com/products/templify?ref=genaisecretsauce.com) 

### [Filament by Galaxy](https://www.producthunt.com/products/galaxy?ref=genaisecretsauce.com)

Move data using your terminal or agent. Fast & open source.

🔥 **Upvotes:** 258 · 👤 **By:** Galaxy (launch team includes Michael Capretta, Shane O'Connor and Saloni Gupta)  
💰 **Pricing:** Free and open source · 🏷 **Category:** Developer tools and data

Filament copies data from one database or system to another, either all at once or by keeping up with changes as they happen. It can be run by a person from the command line or by an AI agent, and it double-checks the data at the hand-off point so nothing is silently lost. **Verdict:** A solid free tool for technical teams; non-technical readers can skip it.

[View on Product Hunt →](https://www.producthunt.com/products/galaxy?ref=genaisecretsauce.com) 

### [Trainn MCP](https://www.producthunt.com/products/trainn?ref=genaisecretsauce.com)

Create product videos and courses from any AI chat

🔥 **Upvotes:** 167 · 👤 **By:** Trainn (co-founder Sumana Ammaiyappan)  
💰 **Pricing:** Free options; a maker comment mentions a 14-day trial and a $19 plan · 🏷 **Category:** Video and online learning

Trainn MCP connects chatbots such as Claude or ChatGPT to Trainn, so you can turn screen recordings into product videos, help articles and training courses just by asking. You can also update old content when your product changes. Everything the AI makes is saved as a draft until you publish it. **Verdict:** Handy for small software companies that never have time to update their how-to videos; the draft-first approach keeps a human in charge.

[View on Product Hunt →](https://www.producthunt.com/products/trainn?ref=genaisecretsauce.com) 

API Pricing

## Snapshot

Provider

Model

Input $/1M

Output $/1M

Context

Anthropic

Claude Fable 5.2

$8.00

$40.00

1M tokens

Anthropic

Claude Fable 5.1

$10.00

$50.00

1M tokens

Anthropic

Claude Opus 5.5

$4.00

$20.00

1M tokens

Anthropic

Claude Sonnet 5.5

$2.00

$10.00

1M tokens

Anthropic

Claude Haiku 5.5

$0.10 (up to 100K-token prompts; $0.50 above)

$0.50 ($2.50 above 100K)

1M tokens

OpenAI

GPT-6 Astra

$10.00 (up to 272K-token prompts; $20.00 above)

$50.00 ($75.00 above 272K)

1.05M tokens

OpenAI

GPT-6.1 Sol

$2.00 (up to 272K; $4.00 above)

$10.00 ($15.00 above 272K)

1.05M tokens

OpenAI

GPT-6.1 Terra

$1.00 (up to 272K; $2.00 above)

$5.00 ($7.50 above 272K)

1.05M tokens

OpenAI

GPT-6 Luna

$0.10 (up to 272K; $0.20 above)

$0.50 ($0.75 above 272K)

1.05M tokens

Google

Gemini 4 Carbon

$2.00 to Nov 30; $4.00 from Dec 1

$10.00 to Nov 30; $20.00 from Dec 1

1M tokens

Google

Gemini 4 Argon

$2.00 to Nov 30; $4.00 from Dec 1

$10.00 to Nov 30; $20.00 from Dec 1

1M tokens

Google

Gemini 4 Hydrogen

$1.50

$7.50

1M tokens

Google

Gemini 3.8 Flash

$0.75 (to Dec 31; $1.50 from Jan 1)

$3.75 (to Dec 31; $7.50 from Jan 1)

not listed

Groq

GPT-OSS 120B

$0.15

$0.60

131K tokens

Groq

Qwen3.8-27B (preview)

$0.80

$4.00

131K tokens

Mistral (open weights)

Mistral Large 4

$0.70

$2.10

1M tokens

Here is what the major AI companies charge through their Application Programming Interface (API), the paid connection developers use to plug AI models into their own apps. Prices are in US dollars per 1 million tokens (a token is roughly three-quarters of a word).  
  
**No price changes since yesterday:** Every standard price from Anthropic, OpenAI, Google and Groq matches yesterday's table. The only update is that Google's pricing page now lists a 1M-token context for Gemini 4 Hydrogen, which it did not show yesterday. Google's Carbon and Argon prices still double on December 1, and Gemini 3.8 Flash doubles on January 1\. Mistral's Large 4 price, OpenAI's context sizes and Anthropic's and Google's other 1M-token context figures carry over from earlier checks.  
  
**What this means:** With Carbon and Argon doubling in 18 days, Gemini 4 Hydrogen now matches their 1M-token reading window at well under half their December price, so Google users should test it on their everyday work now.  
  
arXiv Paper of the Day

## Termination Lag: Coding Agents Solve Early but Stop Late

Musa Cim, Burak Topcu, Mahmut Taylan Kandemir · arXiv:2611.14960

**What it claims:** AI coding agents often find the right fix long before they stop working, and keep spending steps (and money) after the job is done. The authors recorded and tested every code change made by agents across five models, two agent frameworks and three SWE-bench test sets (collections of real software bugs), and found that 38% to 82% of successful runs got the right answer on their very first code edit.  
  
**Key finding:** **After already reaching a correct fix, agents spent a median of 35-37% (Qwen3-Coder) to 54-60% (Laguna-XS.2) of their steps still working, and in 1-8% of cases that extra work undid the correct fix.**  
  
**Why practitioners should care:** If you pay per token for coding agents, a large share of a successful run can be spent after the problem is already solved, mostly on double-checking and browsing the code. Stopping rules, such as ending once the tests pass, could cut costs and even prevent agents from breaking a fix they already got right.  
  
[Read on arXiv →](https://arxiv.org/abs/2611.14960?ref=genaisecretsauce.com)

GenAI Secret Sauce Daily Digest · 2026-11-13

×

Click anywhere or press ESC to close