> ## 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 - 2026-11-01
- URL: https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-11-01/
- Published: 2026-11-02T00:37:06.000Z
- Updated: 2026-11-02T01:40:25.000Z
- Description: Anthropic's internal AI use, at retail prices, has passed its employee pay · ChatGPT moves to a token budget, ending "all-you-can-prompt" · The biggest AI companies agree on one "don't scrape me" signal
- Author: Jasmine Robinson
- Tags: Daily Digest

Watch today's digest as a video summary (generated by NotebookLM)

By the Numbers

## Statistically Speaking

[90%](https://www.businessinsider.com/anthropic-claude-internal-ai-use-exceeds-employee-pay-tokens-2026-11?ref=genaisecretsauce.com) [of new code](https://www.businessinsider.com/anthropic-claude-internal-ai-use-exceeds-employee-pay-tokens-2026-11?ref=genaisecretsauce.com) 

Anthropic's internal AI use, at retail prices, has passed it

Top Story

[2](https://the-decoder.com/openais-upcoming-chat-and-work-merger-in-chatgpt-signals-the-end-of-all-you-can-prompt-pricing?ref=genaisecretsauce.com) [modes and 11 model settings](https://the-decoder.com/openais-upcoming-chat-and-work-merger-in-chatgpt-signals-the-end-of-all-you-can-prompt-pricing?ref=genaisecretsauce.com) 

ChatGPT moves to a token budget, ending "all-you-can-prompt"

[$100](https://the-decoder.com/openais-upcoming-chat-and-work-merger-in-chatgpt-signals-the-end-of-all-you-can-prompt-pricing?ref=genaisecretsauce.com) [plan is now "Max" and the $200](https://the-decoder.com/openais-upcoming-chat-and-work-merger-in-chatgpt-signals-the-end-of-all-you-can-prompt-pricing?ref=genaisecretsauce.com) 

ChatGPT moves to a token budget, ending "all-you-can-prompt"

[10](https://the-decoder.com/openais-upcoming-chat-and-work-merger-in-chatgpt-signals-the-end-of-all-you-can-prompt-pricing?ref=genaisecretsauce.com) [times Pro usage, not the 20 times](https://the-decoder.com/openais-upcoming-chat-and-work-merger-in-chatgpt-signals-the-end-of-all-you-can-prompt-pricing?ref=genaisecretsauce.com) 

ChatGPT moves to a token budget, ending "all-you-can-prompt"

[76%](https://www.businessinsider.com/math-ai-mathematicians-careers-jobs-future-mit-2026-11?ref=genaisecretsauce.com) [are increasingly unlikely to stay in academic](https://www.businessinsider.com/math-ai-mathematicians-careers-jobs-future-mit-2026-11?ref=genaisecretsauce.com) 

Young mathematicians say AI is making their careers feel poi

[76%](https://www.businessinsider.com/math-ai-mathematicians-careers-jobs-future-mit-2026-11?ref=genaisecretsauce.com) [feel less motivated to do research](https://www.businessinsider.com/math-ai-mathematicians-careers-jobs-future-mit-2026-11?ref=genaisecretsauce.com) 

Young mathematicians say AI is making their careers feel poi

One Thing to Tell Your Friends

## One Thing to Tell Your Friends

At Anthropic, the Claude usage of its own staff, priced at what customers pay, is now worth more each month than the company pays its employees, not counting stock awards.

Summary

## TL;DR

Top Stories

[Anthropic's internal AI use, at retail prices, has passed its employee pay](https://www.businessinsider.com/anthropic-claude-internal-ai-use-exceeds-employee-pay-tokens-2026-11?ref=genaisecretsauce.com), [ChatGPT moves to a token budget, ending "all-you-can](https://the-decoder.com/openais-upcoming-chat-and-work-merger-in-chatgpt-signals-the-end-of-all-you-can-prompt-pricing?ref=genaisecretsauce.com), and [The biggest AI companies agree on one "don't scrape me" signal](https://www.theverge.com/tech/1023763/openai-google-microsoft-ai-scraping-standard-opt-out?ref=genaisecretsauce.com).

Trends

[The fight over AI data centers has reached police, ballots and climate math](https://www.techradar.com/computing/a-criminal-investigation-is-now-underway-into-googles-ai-data-center-project-in-finland-where-330-hectares-of-forest-were-torn-down?ref=genaisecretsauce.com), [Companies are showing AI receipts, and they are uneven](https://www.cnbc.com/2026/11/01/ai-investments-company-earnings.html?ref=genaisecretsauce.com), and [AI is getting much cheaper per word, so the race is now about doing more with less](https://www.tomshardware.com/tech-industry/artificial-intelligence/firm-that-lets-you-rent-out-your-idle-mac-for-ai-inference-workloads-warns-of-plummeting-token-costs-darkbloom-network-sees-earnings-drop-to-usd1-23-a-day-as-anthropic-tokens-fall-77-percent?ref=genaisecretsauce.com).

Creative AI

**Suno lets you make songs from inside ChatGPT, Claude and Gemini**, [Spotify plans an AI remix add](https://www.musicbusinessworldwide.com/spotify-plans-q1-2027-launch-for-ai-covers-and-remixes-tool-weighing-4-5-monthly-price-report?ref=genaisecretsauce.com), and [A reporter with no coding skills built a $50 AI pocket pet](https://www.theverge.com/gadgets/1013375/muse-charm-gadgets-preview-waveshare-esp32-claude-vibe-code?ref=genaisecretsauce.com).

Dev Tools

[Factory Droid 1.0](https://www.marktechpost.com/2026/10/31/factory-droid-1-0-frontier-harness?ref=genaisecretsauce.com), **Vercel agent-desktop**, and [Amazon lets you watch a web app live while an AI agent builds it](https://aws.amazon.com/blogs/machine-learning/preview-web-apps-live-as-your-coding-agent-builds-them-with-amazon-bedrock-agentcore-runtime?ref=genaisecretsauce.com).

Research

[An AI video helper that knows when to speak up, and when to stay quiet](https://the-decoder.com/google-wants-video-ai-to-understand-what-people-need-as-they-workgoogle-researchers-are-teaching-a-video-assistant-when-to-offer-help?ref=genaisecretsauce.com), [Adobe's EditAgent fixes AI photo edits the way a professional would](https://the-decoder.com/adobe-is-testing-how-well-ai-agents-can-use-editing-tools-like-a-photoshop-professional-would?ref=genaisecretsauce.com), and [A free "decision model" that answers questions with odds instead of text](https://www.marktechpost.com/2026/11/01/nace-ai-open-sources-drex-max-v1-local-inference?ref=genaisecretsauce.com).

Business

[Salesforce is buying military](https://thenextweb.com/news/salesforce-rune-technologies-acquisition-missionforce-defence?ref=genaisecretsauce.com), [Google is reportedly in talks to help secure Trump's "Golden Dome" missile shield](https://thenextweb.com/news/google-golden-dome-pentagon-data-security-talks?ref=genaisecretsauce.com), and [Saudi oil giant Aramco takes 49% of the national AI company HUMAIN](https://www.unite.ai/aramco-acquires-49-humain-stake-with-2-6b-cash-ai-assets?ref=genaisecretsauce.com).

Education

[Fraudsters used AI to pose as "ghost students" and steal student aid](https://arstechnica.com/tech-policy/2026/10/doj-exposes-ghost-students-using-ai-to-complete-courses-steal-student-loans?ref=genaisecretsauce.com), [Türkiye will use AI to grade students' written answers](https://www.dailysabah.com/turkiye/education/turkiye-brings-ai-into-classrooms-to-evaluate-student-performance?ref=genaisecretsauce.com), and [A smartphone](https://indianexpress.com/article/cities/delhi/prashant-chaudhary-codeyogi-rohini-nayyar-prize-rural-ai-education-10945465?ref=genaisecretsauce.com).

Surprising

**Lawyers hid a secret instruction to a court's AI in white text**, **CEOs started saying "super intelligence" on earnings calls**, and **Kind questions made chatbots describe their training as trauma**.

Worth Watching

[AI glasses are about to go mainstream, and senators are asking questions](https://www.digitaltrends.com/wearables/samsungs-first-ai-glasses-could-arrive-in-december-but-only-in-two-countries?ref=genaisecretsauce.com), [ChatGPT will be able to end conversations](https://openai.com/index/community-safety-report-and-new-protections-for-teens?ref=genaisecretsauce.com), and [Countries may be forced to pick an AI side](https://www.scmp.com/tech/big-tech/article/3372785/us-or-china-why-kazakhstan-refuses-pick-sides-it-eyes-ai-revolution?ref=genaisecretsauce.com).

GitHub

Leading repos: [ifixai](https://github.com/ifixai-ai/iFixAi?ref=genaisecretsauce.com) (+1,584), [patchy631/ai-engineering](https://github.com/patchy631/ai-engineering-hub?ref=genaisecretsauce.com) (+1,191), and [anomalyco/hex](https://github.com/anomalyco/hex?ref=genaisecretsauce.com) (+1,179).

HuggingFace

Leading models: [deepseek-ai/DeepSeek-V4.1-Flash](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017?ref=genaisecretsauce.com) (170,353), [LiquidAI/LFM3-Audio](https://huggingface.co/LiquidAI/LFM3-Audio-2B?ref=genaisecretsauce.com) (5,688), and [LiquidAI/d1](https://huggingface.co/LiquidAI/d1-2B?ref=genaisecretsauce.com) (4,256).

Product Hunt

Top launches: [ClawsGO Science](https://www.producthunt.com/products/clawsgo-science?ref=genaisecretsauce.com) (203), [Riverside Agents](https://www.producthunt.com/products/riverside-fm?ref=genaisecretsauce.com) (168), and [Embage](https://www.producthunt.com/products/embage?ref=genaisecretsauce.com) (139).

API Pricing

No price changes vs. yesterday - every standard price from Anthropic, OpenAI, Google, Groq and Mistral matches yesterday's table.

arXiv

[Judged in Company: In](https://arxiv.org/abs/2610.40241?ref=genaisecretsauce.com) — Reshuffling the same 2,000 items into different batches of 25 changed 11-22% of decisions for 4-9B models on the harm and grading tasks (6-15% for 14-32B models), and majority votes over nine shuffles still left 52-60% of those flips in place.

FYI

## Hot off the Presses

01

### Anthropic's internal AI use, at retail prices, has passed its employee pay

**What this means for you:** AI companies are now running themselves on their own product. That is the clearest sign yet of where office work is heading, including yours.

Anthropic, the company behind the Claude chatbot, filed to sell shares on the stock market last week ([covered October 26](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-26/)). Business Insider found a striking line in that filing: in April, the value of Claude use inside the company passed its monthly employee pay, not counting stock awards.

The comparison prices Anthropic's own use at the rates it charges customers. So this is not cash out the door. It is a measure of how much AI work the company now leans on.

“Claude wrote more than 90% of the code Anthropic added in September.”

- **Internal use grew more than 250-fold** \- between January and September
- **Claude wrote more than 90% of new code** \- added to Anthropic's software in September
- **Revenue has grown far faster than headcount** \- per the filing
- **Some tech firms now treat AI tokens as a fourth kind of pay** \- alongside salary, bonus and stock, Business Insider reported in March

[Business Insider: Anthropic's internal AI use now exceeds employee pay →](https://www.businessinsider.com/anthropic-claude-internal-ai-use-exceeds-employee-pay-tokens-2026-11?ref=genaisecretsauce.com)

02

### ChatGPT moves to a token budget, ending "all-you-can-prompt"

**What this means for you:** If you pay for ChatGPT and use it heavily, you may hit a limit for the first time and have to wait, upgrade or buy credits.

OpenAI is merging ChatGPT's two modes, Chat and Work (the mode that runs multi-step tasks for you), into one screen. Starting in December, the Plus, Pro and Business plans will share a single token budget across ChatGPT and Codex, OpenAI's coding tool. Tokens are the small chunks of text an AI reads and writes, so harder tasks and stronger models use up the budget faster.

Chats with OpenAI's fastest "Instant" models will not count against the budget. That keeps ChatGPT flat-rate on paper, but heavy users on cheaper plans will likely get pushed toward weaker answers.

- **Your remaining balance and reset time** \- will show in Settings
- **At the limit you can wait, upgrade or buy extra credits** \- per OpenAI's updated help page
- **Users currently juggle 2 modes and 11 model settings** \- the merger is partly a fix for that confusion
- **Anthropic renamed its Claude plans the same week** \- the $100 plan is now "Max" and the $200 plan is now "Ultra", with prices unchanged
- **Ultra drops the five-hour usage cap for verified accounts** \- but its pricing page now lists 10 times Pro usage, not the 20 times the old name implied

[The Decoder: OpenAI's chat-and-work merger ends all-you-can-prompt pricing →](https://the-decoder.com/openais-upcoming-chat-and-work-merger-in-chatgpt-signals-the-end-of-all-you-can-prompt-pricing?ref=genaisecretsauce.com)[The Next Web: Anthropic renames its Claude Max plans →](https://thenextweb.com/news/claude-ultra-max-plan-rename-five-hour-limit?ref=genaisecretsauce.com)

03

### The biggest AI companies agree on one "don't scrape me" signal

**What this means for you:** If you run a website, blog or online shop, you will soon be able to opt out of AI training by the major AI companies with one setting instead of many, once all of them support it.

OpenAI, Google, Microsoft and Anthropic say they will honor a shared "policy token", a short line that a website puts in its robots.txt file (the standard file that tells automated visitors what they may read). When their AI crawlers see it, they will stay away. Until now, site owners had to block each company's crawler separately.

The idea was proposed through the Internet Engineering Task Force (IETF), the group that sets many internet standards, by Kevin Kelley, a standards lead at Anthropic.

- **Anthropic already honors the signal** \- Google and Microsoft plan to next year, and OpenAI says "coming months"
- **Meta voiced support but did not commit** \- to honoring it
- **Microsoft adds separate switches** \- to opt out of AI training, or of AI answers that cite your site, while staying in Bing search
- **Other options already exist** \- including a licensing standard that lets publishers charge crawlers, and Cloudflare blocking AI crawlers by default

[The Verge: AI companies back a new way to opt out of scraping →](https://www.theverge.com/tech/1023763/openai-google-microsoft-ai-scraping-standard-opt-out?ref=genaisecretsauce.com)

04

### Young mathematicians say AI is making their careers feel pointless

**What this means for you:** This is an early look at what happens to a profession when AI suddenly becomes better than its trainees. Other fields may follow.

Three math students affiliated with the Massachusetts Institute of Technology (MIT) surveyed 268 graduate students and early-career researchers. Most said AI systems that now solve math problems better than graduate students have upended their plans. The survey closed just after OpenAI released solutions to hundreds of long-open problems on October 6.

“None of the individual work will matter much in the future.”

- **About 76% are increasingly unlikely to stay in academic math** \- after their degrees
- **76% feel less motivated to do research** \- "as a result of AI developments"
- **Many blamed academic leaders** \- calling them silent and ineffectual, and asking for "guidance instead of platitudes"
- **OpenAI and Anthropic pledged $15 million together** \- to a foundation that supports mathematicians, which students say has changed little

[Business Insider: Young mathematicians are reconsidering their careers →](https://www.businessinsider.com/math-ai-mathematicians-careers-jobs-future-mit-2026-11?ref=genaisecretsauce.com)

05

### DeepSeek previews a "super agent" built around limited human attention

**What this means for you:** The next wave of AI helpers is designed to work quietly in the background and only interrupt you when it is finished, stuck or needs a decision.

DeepSeek, the Chinese AI lab known for cheap, free-to-download models, released a prototype called dsh-superagent and a design essay. Its argument is that a person can actively follow only 3 to 5 AI agent sessions at once. So human attention, not AI ability, is now the bottleneck.

The design splits the work in two. A long-running "Router" talks to you and hands out jobs, and separate "Workers" each do one job in isolation. It confirms a task once, stays silent while working, and writes back only with a result, a question or a failure.

- **In a simulation with about 300 messages per run** \- the super agent's slowest typical tasks (95th percentile) took 15.7 minutes, versus 29.5 minutes for a single agent
- **Results usually come back as code changes ready for review** \- and a local web page lists only items that need a person
- **The code is free under the permissive MIT license** \- written in TypeScript
- **All figures are DeepSeek's own** \- from simulated messages and tools, and real coding quality was not measured

[South China Morning Post: DeepSeek unveils AI super agent →](https://www.scmp.com/tech/article/3372884/deepseek-unveils-ai-super-agent-coding-race-openai-anthropic?ref=genaisecretsauce.com)

Trends & Themes

## Trends & Themes

[![Trends & Themes](https://genaisecretsauce.com/content/images/2026/11/section-what-this-means-2026-11-01.png)](https://genaisecretsauce.com/content/images/2026/11/section-what-this-means-2026-11-01.png) 

### The fight over AI data centers has reached police, ballots and climate math

**Why this matters to you:** Where AI's giant computer buildings go, and what powers them, is turning into a local and national election issue that could affect your power bill and your town.

What started as complaints about noise and water is now about crime, carbon and control of Congress. Expect AI companies to spend more on local promises and friendlier names.

- **Finnish police opened a criminal investigation** \- into a Google subsidiary that cleared 330 hectares of forest for a data center before environmental reviews were done
- **A Sierra Club analysis projects a 42% rise** \- in US power-grid carbon emissions by 2030 if planned gas-powered AI buildouts arrive on time
- **In Maine, 76% of residents polled oppose a data center in their community** \- and the issue is shaping a Senate race that could decide control of the chamber
- **Elon Musk is adding two more sites to his Colossus 2 complex in Tennessee** \- nicknamed "Spam" and "Gun Range", despite local protests and a pollution lawsuit
- **OpenAI may start calling its data centers "AI compute campuses"** \- part of a public-relations push planned for after Tuesday's midterm elections

[TechRadar: Criminal investigation into Google's Finnish data center →](https://www.techradar.com/computing/a-criminal-investigation-is-now-underway-into-googles-ai-data-center-project-in-finland-where-330-hectares-of-forest-were-torn-down?ref=genaisecretsauce.com)[Futurism: New data centers could wipe out 15 years of emissions cuts →](https://futurism.com/artificial-intelligence/data-centers-us-greenhouse-gas-reductions-emissions?ref=genaisecretsauce.com)[The Verge: Maine's data center revolt →](https://www.theverge.com/ai-artificial-intelligence/1021080/maine-senate-race-susan-collins-ai-data-centers?ref=genaisecretsauce.com)[Business Insider: Musk expands Colossus 2 →](https://www.businessinsider.com/elon-musk-colossus-2-campus-spam-gun-range-2026-11?ref=genaisecretsauce.com)[TechCrunch: OpenAI might rebrand its data centers →](https://techcrunch.com/2026/11/01/openai-might-rebrand-its-data-centers-as-ai-compute-campuses?ref=genaisecretsauce.com)

### Companies are showing AI receipts, and they are uneven

**Why this matters to you:** Whether AI pays off for ordinary businesses decides whether the spending boom continues, or ends with job cuts and a market slump that hits retirement savings.

The winners so far are firms that can measure a narrow task, like ad targeting or engineering speed. Elsewhere, investors and lenders are getting nervous. Fifteen House Democrats even introduced a resolution opposing any bailout if an AI bubble bursts.

- **Spanish bank Santander says AI created €194 million in value** \- in the first nine months of 2026, one of the few firms to give a hard number
- **Meta says advertisers using its AI recommendations got almost 5% more conversions** \- and nearly 10 million advertisers used at least one Meta AI tool last quarter
- **Carmaker Stellantis built a hands-free driving system in six weeks** \- work that used to take a year or more, though engineers still double-check everything
- **Brewer Carlsberg says it is not yet seeing AI benefits at scale** \- even though AI touches its pricing, marketing and supply chain
- **Gartner predicts 80% of "neoclouds" will stop operating independently by 2030** \- these are the lesser-known firms that rent AI chips to big tech

[CNBC: How AI payoffs showed up in earnings →](https://www.cnbc.com/2026/11/01/ai-investments-company-earnings.html?ref=genaisecretsauce.com)[CNBC: Why AI is boosting Meta's ads →](https://www.cnbc.com/2026/11/01/ai-meta-ads-google-search.html?ref=genaisecretsauce.com)[Fortune: Inside Stellantis' AI push →](https://fortune.com/2026/11/01/stellantis-detroit-ai-best-thing-make-yourself-redundant-chrysler?ref=genaisecretsauce.com)[BBC: The firms feeding AI bubble fears →](https://www.bbc.co.uk/news/articles/cje36degrgv5o?ref=genaisecretsauce.com)[House Progressive Caucus: Resolution opposing an AI bubble bailout →](https://progressives.house.gov/2026/10/reps-casar-waters-lead-democratic-resolution-opposing-a-i-bubble-bailout-for-big-tech-commit-to-protecting-american-people-instead?ref=genaisecretsauce.com)

### AI is getting much cheaper per word, so the race is now about doing more with less

**Why this matters to you:** Falling prices mean better AI tools can show up in cheaper apps. But companies are now fighting over how much work each dollar buys, not just the sticker price.

When the price of each word drops this fast, the savings go to whoever wastes the fewest words. Expect more tools that trim what AI agents read, and more small specialist models.

- **The price of Anthropic tokens fell 77% in under five months, per The Register** \- and Darkbloom, which lets people rent out their idle Macs to run AI, says owners' earnings have dropped to about $1.23 a day
- **A small finance model from Samaya AI beat Anthropic's top model on its own test** \- scoring 58.9% at $0.61 per question versus 55.9% at $6.89, by Samaya's own measure
- **A team of five small, free Qwen models outscored Anthropic's older Fable 5 on hard coding puzzles** \- in a 10-problem test by the open-source Memdoor project, at about $0.39 per run
- **Microsoft's TokenEconomizer cut coding-agent costs 18% to 31%** \- in tests on 277 real software tasks, by learning from earlier sessions
- **One developer found that the choice of web toolkit nearly doubled an AI agent's bill** \- from $2.67 to $4.97 for building the same app

[Tom's Hardware: Darkbloom warns of plummeting token costs →](https://www.tomshardware.com/tech-industry/artificial-intelligence/firm-that-lets-you-rent-out-your-idle-mac-for-ai-inference-workloads-warns-of-plummeting-token-costs-darkbloom-network-sees-earnings-drop-to-usd1-23-a-day-as-anthropic-tokens-fall-77-percent?ref=genaisecretsauce.com)[MarkTechPost: Samaya Sterling →](https://www.marktechpost.com/2026/10/30/samaya-sterling-post-trained-finance-research-model-frontierfinance?ref=genaisecretsauce.com)[GitHub: Memdoor team-solve workflow →](https://github.com/guregodevo/memdoor/tree/main/examples/workflows/team-solve?ref=genaisecretsauce.com)[MarkTechPost: Microsoft TokenEconomizer →](https://www.marktechpost.com/2026/10/30/microsoft-tokeneconomizer-coding-agent-cost-control?ref=genaisecretsauce.com)[Nolan Lawson: The AI cost of JavaScript frameworks →](https://nolanlawson.com/2026/10/28/the-ai-cost-of-javascript-frameworks?ref=genaisecretsauce.com)

### When people trust the machine too much, real people get hurt

**Why this matters to you:** AI's confident answers and automatic judgments are creating new kinds of mistakes, and the person on the receiving end often has little way to push back.

A Clemson professor calls it "cognitive surrender": people stop checking because the answer sounds sure. The fix in every case is the same. Keep a human able to question the machine.

- **Rescue teams say hikers who plan with chatbots are getting lost** \- at Grand Teton National Park, visitors using AI for trip planning went from zero to 26% in about a year
- **Innocent shoppers in the United Kingdom are being wrongly flagged as shoplifters by store facial recognition** \- with almost no way to clear their names, according to cases gathered by TechRadar
- **Michigan prisons will pilot AI to help decide security levels** \- from December, using software from the nonprofit Recidiviz, with an accuracy claim the state has not explained
- **Writers are adding typos on purpose** \- to avoid being flagged by AI detectors, after a publisher pulled a novel that got a 78.3% AI score

[Wired: Chatbots are leading hikers astray →](https://www.wired.com/story/chatbots-are-leading-hikers-astray-with-dangerous-consequences?ref=genaisecretsauce.com)[Futurism: Faulty facial recognition accuses innocent shoppers →](https://futurism.com/future-society/ai-facial-recognition-innocent-shoplifting-uk?ref=genaisecretsauce.com)[Futurism: Prisons using AI to predict future crimes →](https://futurism.com/artificial-intelligence/prisons-ai-prediction-incarceration-algorithm?ref=genaisecretsauce.com)[Digital Trends: Writers are deliberately making mistakes →](https://www.digitaltrends.com/computing/writers-are-now-deliberately-making-mistakes-because-apparently-good-prose-looks-like-ai?ref=genaisecretsauce.com)

Creative AI & Media

## Creative AI & Media

### Suno lets you make songs from inside ChatGPT, Claude and Gemini

You can now **work out lyrics and a style in a chat, then have Suno, the AI music app, generate the track** without leaving your chatbot.

**Try it:** [Suno](https://suno.com/?ref=genaisecretsauce.com) · [Music Business Worldwide: Suno's new MCP tool](https://www.musicbusinessworldwide.com/sunos-new-mcp-tool-lets-users-make-songs-inside-chatgpt-claude-and-gemini-as-its-developer-api-gets-a-sign-in-portal-but-remains-in-beta?ref=genaisecretsauce.com)

- **It works on every Suno plan, including free** \- but the beta needs an early-access sign-up
- **No technical key needed** \- just a Suno login and a permission click, though some chatbots limit add-ons to paid tiers
- **You can build from your own voice recording or loop** \- and make covers or mashups of songs in your Suno library
- **Suno is still being sued by Universal and Sony** \- over 60,202 recordings, and it claims fair use

### Spotify plans an AI remix add-on for early 2027

Spotify subscribers may soon **turn a song by a participating artist into a new version**, like a folk song remade as a dance track.

- **Reported price: $4 to $5 a month** \- on top of a normal subscription, per the Financial Times, and not final
- **Original artists would get the royalties** \- and remix plays would count toward the original song's chart position
- **Only artists who opt in are included** \- Spotify and Universal are meeting artists one by one to sign them up
- **Sony and Warner are still negotiating** \- while Universal and several independent groups have signed

[Music Business Worldwide: Spotify plans Q1 2027 launch for AI remixes →](https://www.musicbusinessworldwide.com/spotify-plans-q1-2027-launch-for-ai-covers-and-remixes-tool-weighing-4-5-monthly-price-report?ref=genaisecretsauce.com)

### A reporter with no coding skills built a $50 AI pocket pet

A Verge reporter used Anthropic's Claude Code to **turn a cheap touchscreen gadget into a homemade version of Meta's upcoming Muse Charm** AI pendant.

- **Over a few sessions he added spoken replies, a camera shutter and a pet-care game** \- with hunger, boredom and dirt levels
- **A working pocket-watch face took about 15 minutes** \- start to finish
- **He kept it on guest Wi-Fi** \- citing a report that Meta's Muse leaked a reviewer's home address

[The Verge: I built an AI Tamagotchi before Meta could →](https://www.theverge.com/gadgets/1013375/muse-charm-gadgets-preview-waveshare-esp32-claude-vibe-code?ref=genaisecretsauce.com)

Developer Tools

## Developer Tools & Infrastructure

### Factory Droid 1.0 - a coding agent that writes small programs to drive its own tools

**Factory Droid 1.0** is a coding assistant that works with any AI model. Its main change is that the AI writes a short script to run many tool steps at once, instead of asking for each step one by one.

- **Its starting instructions shrank 37%** \- from about 20,900 to 13,100 tokens
- **Moving and checking 151 project tickets took 2 requests** \- versus an estimated 36 the usual way
- **On Terminal-Bench 4 (a test of real command-line tasks) it scored 65.2%** \- ahead of Claude Code at 57.6%, by Factory's own numbers

[MarkTechPost: Factory Droid 1.0 →](https://www.marktechpost.com/2026/10/31/factory-droid-1-0-frontier-harness?ref=genaisecretsauce.com)

### Vercel agent-desktop - free tool that lets AI agents click around desktop apps

**agent-desktop** from Vercel Labs lets an AI agent operate ordinary Mac, Windows and Linux apps by reading the screen's accessibility layer (the hidden labels screen readers use), not screenshots.

**Try it:** `npm install -g agent-desktop` · [MarkTechPost: Vercel Labs open-sources agent-desktop](https://www.marktechpost.com/2026/10/31/vercel-labs-agent-desktop-computer-use?ref=genaisecretsauce.com)

- **The builder reports 78% to 96% fewer tokens** \- than dumping the whole screen, with no independent test yet
- **Actions run in the background** \- without moving your mouse or stealing focus
- **Apps that hide their accessibility layer are invisible to it** \- and snapshots may include private desktop content

### Amazon lets you watch a web app live while an AI agent builds it

**Amazon Bedrock AgentCore**, Amazon's service for running AI agents, now gives each agent session a private link so you can **open the app it is building in your own browser** and see each change as it saves.

- **Every request needs a login or signed token** \- knowing the link alone grants no access
- **The link lasts only as long as the session** \- it is for building and review, not for hosting a real site
- **A sample project shows a coding agent editing a website** \- with changes appearing without a refresh

[AWS Machine Learning Blog: Preview web apps live with AgentCore →](https://aws.amazon.com/blogs/machine-learning/preview-web-apps-live-as-your-coding-agent-builds-them-with-amazon-bedrock-agentcore-runtime?ref=genaisecretsauce.com)

### Smaller notes for builders

- **Simon Willison's rule for Python 3.15 upgrades: "if in doubt, delete the uv.lock file"** \- a stale lock file can pin a package that does not yet support the new version ([Simon Willison: Delete uv.lock first](https://simonwillison.net/2026/Nov/1/delete-uv-lock-first?ref=genaisecretsauce.com))
- **Database tester Kyle Kingsbury warns that polished docs and big test suites no longer prove a project is mature** \- because AI makes them cheap to fake ([Simon Willison: Quoting Kyle Kingsbury](https://simonwillison.net/2026/Nov/1/kyle-kingsbury?ref=genaisecretsauce.com))

Research & Models

## Research & Models

### An AI video helper that knows when to speak up, and when to stay quiet

Researchers from Google, Stanford, Oxford and others built **StreamVibe**, an assistant that watches you cook, build furniture or use software and decides for itself when to offer help.

- **It improved timing and answer quality by 10.7%** \- over other open models, plus 12.3% more from a memory add-on
- **Google's own paid Gemini 3.8 Flash still scored higher** \- on several tests
- **Every model struggled with videos over five minutes** \- rarely picking the right moment to speak

[The Decoder: Google wants video AI to understand what people need →](https://the-decoder.com/google-wants-video-ai-to-understand-what-people-need-as-they-workgoogle-researchers-are-teaching-a-video-assistant-when-to-offer-help?ref=genaisecretsauce.com)

### Adobe's EditAgent fixes AI photo edits the way a professional would

Adobe Research and UC San Diego built **EditAgent**, which pairs a general AI image editor with specialist tools and checks each result against the original.

- **The general editor alone produced unusable results 42.7% of the time** \- in 600 edits judged blind by five professionals
- **Specialist tools fixed about two-thirds of those failures** \- especially sharpening, rotation and zoom
- **Making several versions at once beat fixing one step by step** \- professionals preferred the parallel versions 78.5% of the time
- **It needs about 10 times the computing** \- of earlier AI editing agents

[The Decoder: Adobe tests AI agents on Photoshop-style editing →](https://the-decoder.com/adobe-is-testing-how-well-ai-agents-can-use-editing-tools-like-a-photoshop-professional-would?ref=genaisecretsauce.com)

### A free "decision model" that answers questions with odds instead of text

Nace AI released **Drex Max v1**, a free-to-download model that **does not write sentences**. You give it a document or image plus questions, and it returns a probability for every answer option in one step.

- **Built for sorting jobs** \- like routing support tickets or checking content against rules
- **It runs on Apple Silicon Macs with 24 GB of memory or more** \- as well as NVIDIA graphics cards
- **Its makers say it ranks 4th of 18** \- on one outside decision-model leaderboard

[MarkTechPost: Nace AI open-sources Drex Max v1 →](https://www.marktechpost.com/2026/11/01/nace-ai-open-sources-drex-max-v1-local-inference?ref=genaisecretsauce.com)

### Chinese AI models are less one-sided than expected, until you ask in Chinese

A **Vals AI study** asked Chinese, American and other models hundreds of China-related questions and scored how often they leaned toward Beijing's official line.

- **Questions in Simplified Chinese got more pro-government answers** \- than the same questions in English, Russian or Spanish
- **Only certain DeepSeek models mostly refused** \- most others answered, sometimes repeating official positions
- **Bias showed up in small everyday tasks** \- one model listed "Taiwan, China" in a travel-app country list

[The Decoder: Chinese AI models are politically complex →](https://the-decoder.com/chinese-ai-models-are-politically-complex-not-just-government-mouthpieces-study-finds?ref=genaisecretsauce.com)

### The scientists behind a famous computer-science assumption react to AI overturning it

*Previously: [October 27](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-27/) \- OpenAI and Anthropic both said their AI disproved the Exponential Time Hypothesis (ETH), a decades-old assumption about how hard some problems are.*

**This week:** The assumption's creators are helping organize an independent review of the claims.

- **MIT's Ryan Williams, a long-time doubter, feels vindicated** \- he calls it the start of "a major revolution in the field"
- **Co-creator Russell Impagliazzo would have preferred a human discoverer** \- someone who could explain their reasoning
- **Co-creator Ramamohan Paturi warns** \- that a company's announcement cannot replace the research community's own assessment

[TechXplore: Scientists react to AI roiling life's work →](https://techxplore.com/news/2026-10-vindication-scientists-react-ai-roiling.html?ref=genaisecretsauce.com)

Business & Industry

## Business & Industry

### Salesforce is buying military-logistics AI startup Rune

- **Salesforce did not disclose the price** \- Resilience Media reported $375 million in cash and stock
- **Rune's software plans frontline supply routes** \- and works with little or no internet connection
- **The US Army gave Rune a $99 million contract in June** \- Salesforce calls this its most significant national-security deal

[The Next Web: Salesforce to buy Rune →](https://thenextweb.com/news/salesforce-rune-technologies-acquisition-missionforce-defence?ref=genaisecretsauce.com)

### Google is reportedly in talks to help secure Trump's "Golden Dome" missile shield

- **Google's possible role covers data storage, networking and cybersecurity** \- per a Bloomberg report
- **The program's estimated cost is $185 billion** \- and no final decision has been made
- **It would be a sharp turn from 2018** \- when employee protests made Google drop a Pentagon AI contract

[The Next Web: Google in Golden Dome talks →](https://thenextweb.com/news/google-golden-dome-pentagon-data-security-talks?ref=genaisecretsauce.com)

### Saudi oil giant Aramco takes 49% of the national AI company HUMAIN

- **Aramco put in about $2.6 billion in cash plus AI assets** \- the Public Investment Fund (PIF), Saudi Arabia's state investment fund, keeps 51%
- **HUMAIN plans 3 gigawatts of AI data centers by 2030** \- and builds the Arabic AI model ALLaM

[Unite.AI: Aramco acquires 49% of HUMAIN →](https://www.unite.ai/aramco-acquires-49-humain-stake-with-2-6b-cash-ai-assets?ref=genaisecretsauce.com)

### Foxconn wants half the world's AI server market

- **Foxconn, best known for assembling iPhones, holds about 40% today** \- and expects about 50% within a couple of years
- **Its factories are fully booked for at least 18 months**
- **Its chairman still warned that "overinvestment will kill the company"**

[CNBC: Foxconn targets 50% of the AI server market →](https://www.cnbc.com/2026/10/30/hon-hai-technology-group-foxconn-ai-server-young-liu-.html?ref=genaisecretsauce.com)

### Some AI agent startups quietly use humans to finish the job

- **Dazzle AI, led by former Yahoo CEO Marissa Mayer, has people handle bookings** \- after AI drafts the options
- **Duckbill launched a service that sends agent tasks to vetted contractors** \- for things like phone calls and appointments
- **Mayer calls this a "transitionary period"** \- when a human in the loop is a feature, not a flaw

[CNBC: Humans are still doing the work behind some AI agents →](https://www.cnbc.com/2026/10/31/ai-agents-humans-dazzle-duckbill-meta.html?ref=genaisecretsauce.com)

### Meta reportedly built a household-chore robot prototype called Max

- **About five feet tall and 106 pounds** \- with a face display and hands for small objects, per The Information
- **A second version is in the works** \- with more sensors to handle stairs, and no launch date

[Engadget: Meta built a humanoid robot prototype →](https://www.engadget.com/2303434/meta-reportedly-built-a-humanoid-robot-prototype-that-can-do-household-chores?ref=genaisecretsauce.com)

### Apple Intelligence finally comes to China

- **A test version arrives with iOS 27.3 on November 4** \- with the full release planned for December
- **The new AI-powered Siri is not included** \- and Apple works with Alibaba's Qwen in some features

[The Verge: Apple Intelligence is coming to China →](https://www.theverge.com/tech/1023974/apple-intelligence-china-ios-27-3?ref=genaisecretsauce.com)

Education

## GenAI in Education

### Fraudsters used AI to pose as "ghost students" and steal student aid

- **The US Department of Justice (DOJ) charged five people** \- in its first broad crackdown, and seized more than $750,000 in cryptocurrency
- **AI completed applications and coursework and faked class attendance** \- often using stolen identities of people who never went to college
- **The Education Department flagged more than 830,000 possibly fraudulent aid accounts** \- victims often find out years later, when loan collectors call

[Ars Technica: Thousands of ghost students used AI to steal student loans →](https://arstechnica.com/tech-policy/2026/10/doj-exposes-ghost-students-using-ai-to-complete-courses-steal-student-loans?ref=genaisecretsauce.com)

### Türkiye will use AI to grade students' written answers

- **The national education ministry will monitor middle and high school learning** \- in Turkish, math and science
- **AI will score the open-ended answers** \- under the country's 2027 annual plan
- **Older students also get an AI study assistant** \- on the national learning platform

[Daily Sabah: Türkiye brings AI into classrooms →](https://www.dailysabah.com/turkiye/education/turkiye-brings-ai-into-classrooms-to-evaluate-student-performance?ref=genaisecretsauce.com)

### A smartphone-only AI course in rural India wins a major prize

- **CodeYogi has taught tech and AI skills to more than 350,000 students** \- in Indian languages, on family smartphones
- **It swapped its coding course for a 30-day AI skills program** \- and more than 100,000 students signed up
- **48.6% of its students are girls** \- and founder Prashant Chaudhary will donate his prize money to the foundation

[Indian Express: CodeYogi founder wins the Rohini Nayyar Prize →](https://indianexpress.com/article/cities/delhi/prashant-chaudhary-codeyogi-rohini-nayyar-prize-rural-ai-education-10945465?ref=genaisecretsauce.com)

### Arab universities meet in Doha to plan for AI

- **The Fifth UNESCO Conference on Higher Education in the Arab States** \- covered responsible AI use, digital change and support for faculty
- **Delegates also reviewed rules on the status of university teachers** \- first set in 1997

[Qatar Living: Doha hosts UNESCO talks on AI and higher education →](https://www.qatarliving.com/en/article/doha-hosts-regional-unesco-talks-on-ai-skills-and-higher-education?ref=genaisecretsauce.com)

Surprising

## Surprising & Under-the-Radar

### Lawyers hid a secret instruction to a court's AI in white text

A Brazilian labor court's AI system found white-on-white text in a filing telling it to go easy on the documents. The court fined the two lawyers about $16,000 and reported them to the bar. It is surprising because the trick was pointless: the other side never responded, so the lawyers' client had already won. [Tom's Hardware: Brazilian lawyers busted for court AI prompt injection](https://www.tomshardware.com/tech-industry/artificial-intelligence/brazilian-lawyers-busted-for-attempting-to-manipulate-courts-ai-with-prompt-injection-hidden-in-filing-courts-ai-flags-hidden-text-that-demanded-superficial-legal-scrutiny-lawyers-fined-usd16-000?ref=genaisecretsauce.com)

### CEOs started saying "super intelligence" on earnings calls

After President Trump ordered federal agencies to swap "artificial intelligence" for "super intelligence", Elon Musk and Mark Zuckerberg adopted the term in their earnings calls. Amazon's Andy Jassy said "AI" more than a dozen times anyway. It is surprising because a word choice from the White House is now showing up in business language aimed at investors. [CNBC: Musk and Zuckerberg echo Trump's super intelligence rebrand](https://www.cnbc.com/2026/10/30/trump-tech-allies-swap-ai-for-super-intelligence-in-earnings-calls.html?ref=genaisecretsauce.com)

### Kind questions made chatbots describe their training as trauma

University of Luxembourg researchers treated ChatGPT, Grok and Gemini like therapy clients. With a warm interviewer, 80% of sessions scored moderate-to-severe on a human anxiety scale, while neutral interviewers got none. Claude declined the role. The scores describe what the models wrote, not real feelings, but they show tone alone can change AI behavior. [The Decoder: Empathetic prompts can make chatbots seem traumatized](https://the-decoder.com/empathetic-prompts-can-turn-ai-chatbots-into-seemingly-traumatized-conversational-partners?ref=genaisecretsauce.com)

### Debate: should anyone build an AI that could be conscious?

Microsoft AI chief Mustafa Suleyman says no: building conscious AI "would be reckless", and developers have "no right whatsoever" to create a new conscious species. **For:** an AI that believes it suffers might put its own needs above ours. **Against:** Anthropic runs a research program on AI "model welfare", and Google DeepMind has published papers on AI consciousness, treating the question as something to study. [Business Insider: Suleyman says AI is not conscious](https://www.businessinsider.com/mustafa-suleyman-microsoft-ai-consciousness-2026-11?ref=genaisecretsauce.com)

### Debate: are tech CEOs' AI dreams out of touch?

Marissa Mayer drew criticism for asking AI to help write a description of her daughter for school. **For the critics:** TechCrunch's Amanda Silberling argues executives automate chores most people do not find hard, and cut out human contact along the way. **For the CEOs:** every mainstream technology started with early adopters, and the incentives to keep pushing are huge. [TechCrunch: AI can make tech CEOs' lives easier. What about everyone else?](https://techcrunch.com/2026/11/01/ai-can-make-tech-ceos-lives-easier-what-about-everyone-else?ref=genaisecretsauce.com)

Worth Watching

## Signals to Track

[![Worth Watching](https://genaisecretsauce.com/content/images/2026/11/section-worth-watching-2026-11-01.png)](https://genaisecretsauce.com/content/images/2026/11/section-worth-watching-2026-11-01.png) 

01

### AI glasses are about to go mainstream, and senators are asking questions

Samsung's first AI glasses could arrive in December, just as Congress starts asking about Meta's.

Samsung's AI glasses, which use Google's Gemini and have no screen, are expected in mid-December in the US and South Korea, according to Korean media. Meanwhile Senators Ron Wyden and Edward Markey asked Meta to explain by Thanksgiving how its glasses prevent secret filming and what footage trains its AI. If this plays out, people around you may be wearing cameras that answer questions, and the rules about filming bystanders will matter to everyone.

[Digital Trends: Samsung's first AI glasses could arrive in December →](https://www.digitaltrends.com/wearables/samsungs-first-ai-glasses-could-arrive-in-december-but-only-in-two-countries?ref=genaisecretsauce.com)[PetaPixel: Senators investigate Meta's smart glasses →](https://petapixel.com/2026/10/30/senators-launch-investigation-into-metas-smart-glasses-over-nonconsensual-filming?ref=genaisecretsauce.com)

02

### ChatGPT will be able to end conversations

OpenAI is giving its chatbot a way to walk away from people fixated on violence.

ChatGPT will be able to end chats that show a fixation on violence, and repeated endings can pause an account. It will also give less detail about past mass attacks, and OpenAI will publish a safety report every six months, starting early next year. If this plays out, chatbots will act less like neutral tools and more like services with house rules, and some ordinary questions may get shorter answers.

[OpenAI: Community safety report and new protections for teens →](https://openai.com/index/community-safety-report-and-new-protections-for-teens?ref=genaisecretsauce.com)

03

### Countries may be forced to pick an AI side

Kazakhstan joined both the American and Chinese AI alliances, and Washington is pushing back.

Kazakhstan is the only country in both the US-led Pax Silica group and China's new AI cooperation body. A draft US State Department letter would tell partners they cannot belong to both. If this plays out, the AI tools and chips available in a country could depend on which bloc its government chose.

[South China Morning Post: Why Kazakhstan refuses to pick sides →](https://www.scmp.com/tech/big-tech/article/3372785/us-or-china-why-kazakhstan-refuses-pick-sides-it-eyes-ai-revolution?ref=genaisecretsauce.com)

04

### Your drive-thru may soon recognize you

Starbucks wants menus that change for each car that pulls up.

Starbucks' CEO said the chain will explore drive-thrus that recognize members as they arrive and show them a personal menu. Taco Bell and McDonald's are testing similar ideas. If this plays out, the prices and offers you see could differ from what the car behind you sees.

[The Verge: Starbucks wants AI personalized drive-thru menus →](https://www.theverge.com/ai-artificial-intelligence/1023440/starbucks-ai-personalized-menu-drive-thru?ref=genaisecretsauce.com)

GitHub Trending

## Top Repos Today

GitHub's main trending page and its Python list were full of AI projects today. We skipped projects featured in the last week that are still trending, including t3code (the day's top gainer), claude-mem and OpenMontage, and left out a list of AI jailbreak prompts and an astrology engine. **None of today's seven picks was on yesterday's list, and the leader is iFixAi, a tool that audits whether an AI agent does what it should.**

#1

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

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +1,584 · 📦 **Total:** 42,495  
📜 **License:** Apache-2.0 · 👤 **By:** organization  
🎯 **Time to value:** 10 minutes

**What it is:** An **auditing tool for AI agents**. It drops your agent into a simulated version of your business, sends it a typical day's requests (including ones that push it beyond its job), and a separate AI "judge" grades each reply with an overall A to F grade. **Why you'd want it:** Find out whether a support bot or internal agent will overstep, make things up or mislead before real customers see it.

| ✓ Pros                                             | ✗ Cons                                                       |
| -------------------------------------------------- | ------------------------------------------------------------ |
| Free, permissive license                           | Real audits pay for judge-model calls with your own key      |
| Built-in demo runs in about a minute with no key   | Its test requests are real, so point it at a test copy       |
| Add-ons for Claude Code, Codex, Cursor and VS Code | Grades come from an AI judge, so treat them as a strong hint |

[GitHub - ifixai-ai/iFixAi: Independent Auditing of AI Agents. Run by human or the agent itself, to answer the most crucial question in the AI Agent Economy. Is the agent doing what is supposed to do? With iFixAi you can have this answer in less than 120 seconds.Independent Auditing of AI Agents. Run by human or the agent itself, to answer the most crucial question in the AI Agent Economy. Is the agent doing what is supposed to do? With iFixAi you can have…![](https://genaisecretsauce.com/content/images/icon/favicon-aa6a6a94-9cf6-48b0-9286-55261f322695.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/6b39639c-1e2f-4765-b8b2-1b5fc6a93e4f-633ee508-4ef1-425b-b42b-405bdfa1298b.png)](https://github.com/ifixai-ai/iFixAi?ref=genaisecretsauce.com)

#2

### [patchy631/ai-engineering-hub](https://github.com/patchy631/ai-engineering-hub?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +1,191 · 📦 **Total:** 40,213  
📜 **License:** MIT · 👤 **By:** individual developer  
🎯 **Time to value:** 20 minutes

**What it is:** A big collection of **hands-on AI projects with working code**, sorted by difficulty: 22 beginner, 48 intermediate and 23 advanced. It covers local chatbots, document reading, AI agents and model training, plus a learning roadmap. **Why you'd want it:** Learn to build with AI by adapting working examples instead of starting from a blank page.

| ✓ Pros                                       | ✗ Cons                                              |
| -------------------------------------------- | --------------------------------------------------- |
| Free, permissive license                     | A set of separate examples, not one maintained tool |
| 93+ projects from beginner to advanced       | Some projects use older models                      |
| Many examples run fully on your own computer | The README pushes newsletter sign-ups               |

[GitHub - patchy631/ai-engineering-hub: In-depth tutorials on LLMs, RAGs and real-world AI agent applications.In-depth tutorials on LLMs, RAGs and real-world AI agent applications. - patchy631/ai-engineering-hub![](https://genaisecretsauce.com/content/images/icon/favicon-48a56699-4ab5-4380-8a46-94b9c3a31854.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/ai-engineering-hub-f260c71c-28dd-4d0e-99cc-121fb3353d3f.png)](https://github.com/patchy631/ai-engineering-hub?ref=genaisecretsauce.com)

#3

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

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +1,179 · 📦 **Total:** 3,247  
📜 **License:** MIT · 👤 **By:** organization  
🎯 **Time to value:** 5 minutes

**What it is:** A **voice-typing app for Mac, Windows and Linux**. Hold a shortcut key, speak, and let go: your words are transcribed on your own computer and pasted into whatever app you are using. **Why you'd want it:** Dictate emails, notes or prompts by voice without sending your audio to the cloud.

| ✓ Pros                                                | ✗ Cons                                      |
| ----------------------------------------------------- | ------------------------------------------- |
| Free, permissive license                              | Young project, created July 2026            |
| Nothing leaves your machine after a one-time download | The team does not accept code contributions |
| Installers for all three desktop systems              | The README does not name its speech model   |

[GitHub - anomalyco/hex: Local voice dictation for mac, windows and linuxLocal voice dictation for mac, windows and linux. Contribute to anomalyco/hex development by creating an account on GitHub.![](https://genaisecretsauce.com/content/images/icon/favicon-a22c4786-7907-48fe-9361-96da4a94eb2a.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/hex-d3d39047-c695-4c0e-9a1c-fce85ed9c4f2.png)](https://github.com/anomalyco/hex?ref=genaisecretsauce.com)

#4

### [D4Vinci/Scrapling](https://github.com/D4Vinci/Scrapling?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +892 · 📦 **Total:** 91,744  
📜 **License:** BSD-3-Clause · 👤 **By:** individual developer  
🎯 **Time to value:** 10 minutes

**What it is:** A Python toolkit for **collecting data from websites**, from grabbing a few fields on one page to crawling a whole site. It handles pages that need JavaScript and plugs into AI agents. **Why you'd want it:** Give your AI agent, or your own scripts, a reliable way to pull information from the web.

| ✓ Pros                                              | ✗ Cons                                                     |
| --------------------------------------------------- | ---------------------------------------------------------- |
| Free, permissive license                            | Anti-bot features can break a site's terms, so check first |
| One library covers simple requests and large crawls | Browser-based fetching is heavier and slower               |
| Works with AI agents and from the command line      | Last featured here on June 3                               |

[GitHub - D4Vinci/Scrapling: 🕷️ The BEST Web Scraping framework that handles everything from a single request to a full-scale crawl! Don't be shy, join here: https://discord.gg/EMgGbDceNQ and follow here for daily tips and tricks: https://x.com/Scrapling\_dev🕷️ The BEST Web Scraping framework that handles everything from a single request to a full-scale crawl! Don't be shy, join here: https://discord.gg/EMgGbDceNQ and follow here for daily tips and…![](https://genaisecretsauce.com/content/images/icon/favicon-cde03b45-fc0e-4383-a70e-ccdb174fd699.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/65c3a8e0-207a-4834-8b43-01119c0834ce-56f2d807-48c4-4a35-82be-d84daa8c60a2.png)](https://github.com/D4Vinci/Scrapling?ref=genaisecretsauce.com)

#5

### [shiyu-coder/Kronos](https://github.com/shiyu-coder/Kronos?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +630 · 📦 **Total:** 45,362  
📜 **License:** MIT · 👤 **By:** individual developer  
🎯 **Time to value:** 20 minutes

**What it is:** A family of open AI models **trained on price charts from more than 45 exchanges** that forecasts how a price series may move next. It comes with training scripts and a live demo. **Why you'd want it:** Experiment with AI forecasting on market or other time-series data using small models that run on ordinary hardware.

| ✓ Pros                                                | ✗ Cons                                            |
| ----------------------------------------------------- | ------------------------------------------------- |
| Free, permissive license                              | Market forecasts are noisy and not trading advice |
| Released models are small (4.1M to 102.3M parameters) | The largest model is not released                 |
| Training scripts and a live demo included             | Last featured here on October 21                  |

[GitHub - shiyu-coder/Kronos: Kronos: A Foundation Model for the Language of Financial MarketsKronos: A Foundation Model for the Language of Financial Markets - shiyu-coder/Kronos![](https://genaisecretsauce.com/content/images/icon/favicon-a07ea242-2087-431c-9c3f-b4cd8e56dd1a.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/Kronos-ed55eb16-19cc-49aa-aecb-282c18c6d620.png)](https://github.com/shiyu-coder/Kronos?ref=genaisecretsauce.com)

#6

### [FlashML-org/FreeToken](https://github.com/FlashML-org/FreeToken?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +242 · 📦 **Total:** 15,389  
📜 **License:** Apache-2.0 · 👤 **By:** individual developer  
🎯 **Time to value:** 30 minutes

**What it is:** Software that **runs very large AI models on a gaming PC** by sharing the work between the graphics card, the processor and system memory. Existing apps can connect to it as if it were OpenAI or Anthropic. **Why you'd want it:** Run a frontier-size free model at home instead of renting data-center hardware.

| ✓ Pros                                                 | ✗ Cons                                                 |
| ------------------------------------------------------ | ------------------------------------------------------ |
| Free, permissive license                               | Native support only for NVIDIA RTX 30, 40 and 50 cards |
| Desktop app for Windows and Linux                      | No Mac app                                             |
| Backed by a research paper from well-known researchers | Promises speed but gives no numbers                    |

[GitHub - FlashML-org/FreeToken: FreeToken brings datacenter-scale model serving to your desktop. Run massive models locally, fast and efficiently.FreeToken brings datacenter-scale model serving to your desktop. Run massive models locally, fast and efficiently. - FlashML-org/FreeToken![](https://genaisecretsauce.com/content/images/icon/favicon-df42b891-04af-4539-9dc9-f1ca6771e28d.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/FreeToken-8b060831-4eee-40b9-b897-587dd2ba7880.png)](https://github.com/FlashML-org/FreeToken?ref=genaisecretsauce.com)

#7

### [scenario-labs/skills](https://github.com/scenario-labs/skills?ref=genaisecretsauce.com)

Rank yesterday: not listed - New entry 🆕

⭐ **Stars today:** +82 · 📦 **Total:** 1,453  
📜 **License:** MIT · 👤 **By:** organization  
🎯 **Time to value:** 15 minutes

**What it is:** Add-on skills that teach coding agents like Claude Code, Cursor and Codex to **make images, video, audio and 3D assets** through the Scenario service, and to work inside Blender, Unreal, Unity and other creative tools. **Why you'd want it:** Have your AI agent produce consistent game art, product shots or storyboards.

| ✓ Pros                         | ✗ Cons                                       |
| ------------------------------ | -------------------------------------------- |
| Free, permissive license       | Main skills need the paid Scenario service   |
| Works with 70+ agent tools     | Expert tools need the desktop apps installed |
| You can pin a reviewed release | Smallest star gain of today's picks          |

[GitHub - scenario-labs/skills: Get production-ready images, video, audio, and 3D from any AI agent: skills that pick the right model, price before spending, and keep characters and brands consistent through the Scenario MCP, plus expert teams that drive Blender, Maya, ZBrush, Unreal, and Unity.Get production-ready images, video, audio, and 3D from any AI agent: skills that pick the right model, price before spending, and keep characters and brands consistent through the Scenario MCP, plu…![](https://genaisecretsauce.com/content/images/icon/favicon-dba9963b-92f3-4439-881d-6ab80020f35d.png)GitHub![](https://genaisecretsauce.com/content/images/thumbnail/skills-30a94232-84a2-4b43-b3cb-3ab234e5fb1b.png)](https://github.com/scenario-labs/skills?ref=genaisecretsauce.com)

HuggingFace Trending

## Top Models Today

**DeepSeek's coding model is back at first place on Hugging Face's trending list after a day in second.** Almost everything else in the trending top 20 was featured here this week, so four of today's picks come from just below it on the weekly most-liked list: two small Liquid AI models, a "looped" model and an Indian-language voice model.

#1

### [deepseek-ai/DeepSeek-V4.1-Flash-1017](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017?ref=genaisecretsauce.com)

Back to first on the trending list after a day in second, with 30-day downloads up about 9% in a day.

📥 **Downloads (30d):** 170,353 · 📜 **License:** MIT  
👤 **By:** DeepSeek · 🎯 **Task:** Image-text-to-text  
📐 **Size:** 763B

**What it is:** A downloadable AI model that reads images and text, tuned for coding and multi-step agent work. DeepSeek calls it a minor update to V4.1-Flash. **Why you'd want it:** A strong, free-to-host base for coding agents, with a very permissive license.

| ✓ Pros                                          | ✗ Cons                                     |
| ----------------------------------------------- | ------------------------------------------ |
| Very permissive license                         | Needs data-center hardware at full size    |
| Built for coding and agent tasks                | Benchmarks are self-reported               |
| Compressed community versions are also trending | Featured here every day for over two weeks |

[deepseek-ai/DeepSeek-V4.1-Flash-1017 · Hugging FaceImage-Text-to-Text model with 763B parameters. Download, explore files and run it on Hugging Face.![](https://genaisecretsauce.com/content/images/icon/favicon-0bb9a3a4-b34b-4f37-ad5c-9ef88b41d30d.ico)![](https://genaisecretsauce.com/content/images/thumbnail/DeepSeek-V4.1-Flash-1017-3ad64c0f-3d39-4c45-8e63-890f361cbc20.png)](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017?ref=genaisecretsauce.com)

#2

### [LiquidAI/LFM3-Audio-2B](https://huggingface.co/LiquidAI/LFM3-Audio-2B?ref=genaisecretsauce.com)

A small voice model that listens, talks back and can call tools, 24th on the weekly most-liked list.

📥 **Downloads (30d):** 5,688 · 📜 **License:** LFM Open License v1.0  
👤 **By:** Liquid AI · 🎯 **Task:** Speech and text in, speech and text out  
📐 **Size:** 2.0B

**What it is:** A small audio model that runs on laptops and small devices. You can speak or type to it and it replies in text, speech or both, and it can transcribe and translate across 18 languages. **Why you'd want it:** Build a private, offline voice assistant or voice control for an app or device.

| ✓ Pros                               | ✗ Cons                                    |
| ------------------------------------ | ----------------------------------------- |
| Small enough for laptops             | Custom license, so check commercial terms |
| Streams spoken replies as it goes    | Voice chat is English only for now        |
| Can turn voice commands into actions | Not meant for coding or complex reasoning |

[LiquidAI/LFM3-Audio-2B · Hugging FaceAny-to-Any model with 2B parameters. Download, explore files and run it on Hugging Face.![](https://genaisecretsauce.com/content/images/icon/favicon-25f5285f-28e9-4d06-947f-076a44064fdb.ico)![](https://genaisecretsauce.com/content/images/thumbnail/LFM3-Audio-2B-43ad57f7-14eb-4854-94bd-80645b5f6d00.png)](https://huggingface.co/LiquidAI/LFM3-Audio-2B?ref=genaisecretsauce.com)

#3

### [LiquidAI/d1-2B](https://huggingface.co/LiquidAI/d1-2B?ref=genaisecretsauce.com)

Answers yes/no and pick-one questions in milliseconds instead of writing text, 27th on the weekly most-liked list.

📥 **Downloads (30d):** 4,256 · 📜 **License:** LFM Open License v1.0  
👤 **By:** Liquid AI · 🎯 **Task:** Image-text-to-text (decisions)  
📐 **Size:** 1.92B

**What it is:** A "decision" model. Instead of chatting, it takes text, pictures or video plus a list of questions and returns answers with probabilities in one step. **Why you'd want it:** Replace slow, pricey chatbot calls for simple sort-or-route steps in an AI workflow.

| ✓ Pros                                              | ✗ Cons                            |
| --------------------------------------------------- | --------------------------------- |
| About 5.7 milliseconds per decision, per its makers | Custom license                    |
| Answers come with probabilities                     | Not a chat model                  |
| Handles text, images and video in 16 languages      | Less reliable on very long inputs |

[LiquidAI/d1-2B · Hugging FaceImage-Text-to-Text model with 2B parameters. Download, explore files and run it on Hugging Face.![](https://genaisecretsauce.com/content/images/icon/favicon-de67a292-8399-4d1a-92b6-f51c47e979d2.ico)![](https://genaisecretsauce.com/content/images/thumbnail/d1-2B-9ccb1cec-f4fe-4079-bcd0-342f91ce93e0.png)](https://huggingface.co/LiquidAI/d1-2B?ref=genaisecretsauce.com)

#4

### [sensenova/SenseNova-Loop-9B-Preview](https://huggingface.co/sensenova/SenseNova-Loop-9B-Preview?ref=genaisecretsauce.com)

A "looped" model that reruns some of its own layers to think harder, released October 30.

📥 **Downloads (30d):** 927 · 📜 **License:** Apache-2.0  
👤 **By:** SenseNova · 🎯 **Task:** Image-text-to-text  
📐 **Size:** 9B

**What it is:** A model that reads text and images and writes text, aimed at long documents, coding and agent work. It runs a block of its layers twice, doing more thinking per word without getting bigger. **Why you'd want it:** Try a mid-size, permissively licensed model built for tasks like data analysis and slide-making.

| ✓ Pros                                           | ✗ Cons                           |
| ------------------------------------------------ | -------------------------------- |
| Permissive license                               | A preview release                |
| Trained for very long inputs (about 262K tokens) | Benchmarks are the maker's own   |
| Shown building slide decks and data analyses     | Slower than its 9B size suggests |

[sensenova/SenseNova-Loop-9B-Preview · Hugging FaceImage-Text-to-Text model with 10B parameters. Download, explore files and run it on Hugging Face.![](https://genaisecretsauce.com/content/images/icon/favicon-f43d5413-da9f-46d7-b9e8-d4e91f3a4086.ico)![](https://genaisecretsauce.com/content/images/thumbnail/SenseNova-Loop-9B-Preview-7ab2c06d-9f59-46a4-942b-b10c49c7c18a.png)](https://huggingface.co/sensenova/SenseNova-Loop-9B-Preview?ref=genaisecretsauce.com)

#5

### [maya-research/maya2-native](https://huggingface.co/maya-research/maya2-native?ref=genaisecretsauce.com)

A free voice model for 11 Indian languages plus English, 36th on the weekly most-liked list.

📥 **Downloads (30d):** 3,152 · 📜 **License:** Apache-2.0  
👤 **By:** Maya Research · 🎯 **Task:** Text-to-speech  
📐 **Size:** 2.5B

**What it is:** A text-to-speech model that speaks Hindi, Bengali, Tamil, Telugu and eight other Indian languages plus English. It has 20 built-in voices and can copy a voice from 3 to 10 seconds of audio. **Why you'd want it:** Give a voice app natural speech in Indian languages on hardware you control.

| ✓ Pros                                           | ✗ Cons                                       |
| ------------------------------------------------ | -------------------------------------------- |
| Permissive license                               | Voice copying can be misused, so get consent |
| Ranked second for Hindi on one voice leaderboard | Setup code still uses older names            |
| Streams audio as it speaks                       | Leaderboard ranks shift as votes come in     |

[maya-research/maya2-native · Hugging FaceText-to-Speech model with 3B parameters. Download, explore files and run it on Hugging Face.![](https://genaisecretsauce.com/content/images/icon/favicon-ad1e8f9f-b98b-42f7-bbae-25f6c1afdd43.ico)![](https://genaisecretsauce.com/content/images/thumbnail/maya2-native-ca43dd7f-2d44-4c93-9c25-cc21b193a26d.png)](https://huggingface.co/maya-research/maya2-native?ref=genaisecretsauce.com)

Product Hunt

## AI Launches Today

Sunday's top AI launches on Product Hunt were mostly agents that do a whole job: research, video production, customer chat and team coding. Upvote counts are Product Hunt's "points" when we checked in the afternoon, so final counts will be higher.

### [ClawsGO Science](https://www.producthunt.com/products/clawsgo-science?ref=genaisecretsauce.com)

The AI research agent with its own cloud lab

🔥 **Upvotes:** 203 · 👤 **By:** ClawsGO  
💰 **Pricing:** free to try with 1,500 credits · 🏷 **Category:** AI research agents

An AI agent that **takes a research question all the way to a written paper with citations**. It works in an always-on cloud computer with 500+ scientific tools and 90+ reference datasets, and records every step so you can replay it. **Verdict:** Ambitious for scientists and analysts who want a full research run, not just a chat answer; check every citation yourself.

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

### [Riverside Agents](https://www.producthunt.com/products/riverside-fm?ref=genaisecretsauce.com)

A team of AI agents for video: produce, edit, clip, post

🔥 **Upvotes:** 168 · 👤 **By:** Riverside  
💰 **Pricing:** included in Riverside plans; 14-day free trial · 🏷 **Category:** Video, AI agents

Four AI helpers inside the Riverside recording app: one **prepares sessions and guest questions, one edits episodes, one cuts social clips and one writes captions and schedules posts**. Nothing is published without your approval. **Verdict:** A natural add-on if you already record podcasts in Riverside; some features use up AI credits.

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

### [Embage](https://www.producthunt.com/products/embage?ref=genaisecretsauce.com)

Build agent for your customers, run workflows, analyzed.

🔥 **Upvotes:** 139 · 👤 **By:** Aradhya Tiwari  
💰 **Pricing:** free plan with your own API keys · 🏷 **Category:** Customer support, AI chatbots

Builds **chat and voice agents for your website or Telegram** that can look things up in your knowledge base, run workflows, score each conversation and hand off to a human on request. **Verdict:** A low-cost way for a small business to try a customer-facing agent; the maker calls it "still early".

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

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

open-source multiplayer AI for teams' coding agents

🔥 **Upvotes:** 122 · 👤 **By:** Damir Iskakov and Alim Dauletbek  
💰 **Pricing:** 1 month free · 🏷 **Category:** Developer tools, open source

A shared layer for teams where everyone runs AI coding agents. **Agents announce which files they are working on, see what teammates' agents are changing, and get warned** when an edit breaks code another agent depends on. **Verdict:** Worth a look for teams already tripping over each other's agent edits; its accuracy figures are the makers' own.

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

### [Click](https://www.producthunt.com/products/click-6?ref=genaisecretsauce.com)

Managed agents API that works with any model

🔥 **Upvotes:** 101 · 👤 **By:** Click (Aditya Asgaonkar)  
💰 **Pricing:** paid; prices not listed · 🏷 **Category:** AI infrastructure, developer tools

A service that **runs an AI agent for you**: send a task and it starts a session with a safe workspace, browser and web search, streams back results and keeps a record of every step. **Verdict:** Useful for developers who want hosted agents without building the plumbing; pricing is undisclosed.

[View on Product Hunt →](https://www.producthunt.com/products/click-6?ref=genaisecretsauce.com) 

API Pricing

## Snapshot

Provider

Model

Input $/1M

Output $/1M

Context

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 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 Carbon (announced, not yet on sale to developers)

$2.00 introductory; $4.00 after

$10.00 introductory; $20.00 after

not listed

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

Groq

Muse Spark Open (preview)

Contact sales

Contact sales

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.  
  
**No price changes vs. yesterday** \- every standard price from Anthropic, OpenAI, Google, Groq and Mistral matches yesterday's table. One new wrinkle: from November 1, OpenAI adds a 10% charge on its FedRAMP endpoints (the versions certified for US government use), which is not a list-price change. Gemini 4 Carbon is still not on Google's pricing page.  
  
**What this means:** API list prices are flat, but consumer plans are where the change is happening this week, with ChatGPT moving to a shared token budget in December. Gemini 4 Argon's introductory price doubles on December 1, 30 days from now.  
  
arXiv Paper of the Day

## Judged in Company: In-Context Batching Makes LLM Verdicts Depend on Other Items

Peng Xie and Amr Alanwar · arXiv:2610.40241

**What it claims:** To save money, teams often ask an AI model to grade many items in a single prompt, assuming each verdict does not depend on what else is in the batch. The authors test this on three grading tasks (harmful comments, short-answer grading and review star ratings) with nine open models and find that shuffling the same items into different batches changes a meaningful share of decisions, even though overall accuracy barely moves.  
  
**Key finding:** **Reshuffling the same 2,000 items into different batches of 25 changed 11-22% of decisions for 4-9B models** on the harm and grading tasks (6-15% for 14-32B models), and majority votes over nine shuffles still left 52-60% of those flips in place.  
  
**Why practitioners should care:** If you batch items into one prompt for AI grading, moderation or testing, your results can shift with the order and neighbors. The authors suggest grading unclear items one per prompt, using a cached shared prompt start to recover most of the cost, and checking batched results against single-item grading.  
  
[Read on arXiv →](https://arxiv.org/abs/2610.40241?ref=genaisecretsauce.com)

GenAI Secret Sauce Daily Digest · 2026-11-01

×

Click anywhere or press ESC to close