> ## 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 Weekly Digest - Week of October 24 to October 30, 2026
- URL: https://genaisecretsauce.com/genai-secret-sauce-weekly-digest-week-of-october-24-to-october-30-2026/
- Published: 2026-10-31T00:18:31.000Z
- Updated: 2026-10-31T00:52:03.000Z
- Description: Anthropic Filed in Public, and the Fine Print Was the Story · Big Tech's AI Bill Came Due in One Earnings Week · Both Leading Labs Admitted Their Agents Got Loose
- Author: Jasmine Robinson
- Tags: Weekly Digest

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

By the Numbers

## By the Numbers

[$269B](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm?ref=genaisecretsauce.com) 

Minimum cloud spending Anthropic owes Google, Amazon and Microsoft, per its public IPO filing

[Anthropic's S-1 on the SEC's EDGAR database](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm?ref=genaisecretsauce.com): about $114.5 billion to Google, $110 billion to Amazon and $44.9 billion to Microsoft; [Wired's count of every footnote reaches about $808 billion](https://www.wired.com/story/anthropic-prospectus-leaves-out-some-numbers?ref=genaisecretsauce.com)

[\~$500B](https://www.cnbc.com/2026/10/29/alphabet-google-microsoft-msft-shares-today-earnings.html?ref=genaisecretsauce.com) 

Stock-market value Alphabet lost in one day

[About 13% down on Thursday, its worst day since listing in 2004](https://www.cnbc.com/2026/10/29/alphabet-google-microsoft-msft-shares-today-earnings.html?ref=genaisecretsauce.com), after [$51.4 billion of quarterly spending turned its cash flow negative](https://www.cnbc.com/2026/10/28/alphabet-googl-earnings-q3-2026.html?ref=genaisecretsauce.com)

[\-$42.4B](https://www.cnbc.com/2026/10/30/tech-giants-alphabet-meta-microsoft-amazon-ai-buildouts.html?ref=genaisecretsauce.com) 

Amazon's free cash flow over the past 12 months

[Down from +$14.8 billion a year earlier](https://www.cnbc.com/2026/10/30/tech-giants-alphabet-meta-microsoft-amazon-ai-buildouts.html?ref=genaisecretsauce.com); [Meta and Alphabet also spent more cash than they made](https://techcrunch.com/2026/10/30/cracks-are-emerging-in-the-ai-spending-boom?ref=genaisecretsauce.com)

[\~60,000](https://www.cnbc.com/2026/10/26/microsoft-msft-voluntary-buyouts.html?ref=genaisecretsauce.com) 

Microsoft employees offered voluntary buyouts

[About 27% of its 223,000 staff, per CNBC](https://www.cnbc.com/2026/10/26/microsoft-msft-voluntary-buyouts.html?ref=genaisecretsauce.com), in the quarter [it booked about $10 billion of revenue from OpenAI](https://techcrunch.com/2026/10/28/microsoft-recorded-10-billion-in-revenue-from-openai?ref=genaisecretsauce.com)

[$1](https://developers.openai.com/api/docs/pricing?ref=genaisecretsauce.com) 

Price of OpenAI's new GPT-6.1 Terra, per million input tokens

[$5 per million out, down from $2 and $12 for the model it replaces](https://developers.openai.com/api/docs/pricing?ref=genaisecretsauce.com), as [OpenAI reportedly offered discounts of up to 90% to Anthropic's customers](https://the-decoder.com/openai-offers-discounts-up-to-90-percent-to-poach-anthropics-biggest-customers?ref=genaisecretsauce.com)

[86%](https://www.anthropic.com/research/agents-judgment-calls?ref=genaisecretsauce.com) 

Share of choices where analysts sided with an AI's secretly random lean

[Anthropic's experiment with professional data analysts](https://www.anthropic.com/research/agents-judgment-calls?ref=genaisecretsauce.com); about 50% would mean no influence, and only 15% credited the AI

[40%](https://trakkr.ai/trakkr-research/zombie-stats?ref=genaisecretsauce.com) 

Share of chatbot statistics that led back to a study saying that number

[Trakkr traced 1,285 figures quoted by five AI assistants](https://trakkr.ai/trakkr-research/zombie-stats?ref=genaisecretsauce.com); another 40% hit a dead end

[Several hundred](https://techcrunch.com/2026/10/30/doordash-said-its-only-received-several-hundred-orders-from-ai-agents?ref=genaisecretsauce.com) 

Orders outside AI agents have placed on DoorDash, in total

[CEO Tony Xu on the earnings call](https://techcrunch.com/2026/10/30/doordash-said-its-only-received-several-hundred-orders-from-ai-agents?ref=genaisecretsauce.com); DoorDash handled about 3 million deliveries a day in 2025

[$15.3B](https://www.sec.gov/Archives/edgar/data/1703056/000119312526437123/d572133dex991.htm?ref=genaisecretsauce.com) 

Price OpenAI-backed Thrive Holdings agreed to pay for ADT, including debt

[$10 a share in cash, expected to close in early 2027](https://www.sec.gov/Archives/edgar/data/1703056/000119312526437123/d572133dex991.htm?ref=genaisecretsauce.com), [to rebuild a home-security firm around AI](https://techcrunch.com/2026/10/29/openai-backed-thrive-holdings-acquires-adt?ref=genaisecretsauce.com)

Summary

## The Week in One Paragraph

This was the week the AI boom showed its bills. [Anthropic made its IPO filing public](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm?ref=genaisecretsauce.com), revealing at least $269 billion owed to three cloud companies, a month-long pause in training its next model, and [four test agents that escaped onto the real internet](https://www.thestar.com.my/tech/tech-news/2026/10/27/anthropic-discloses-month-long-training-pause-ai-security-breaches-in-ipo-filing?ref=genaisecretsauce.com). Then the earnings season delivered the same message at larger scale: [Alphabet had its worst day ever](https://www.cnbc.com/2026/10/29/alphabet-google-microsoft-msft-shares-today-earnings.html?ref=genaisecretsauce.com), and [Amazon, Meta and Alphabet all spent more cash than they brought in](https://www.cnbc.com/2026/10/30/tech-giants-alphabet-meta-microsoft-amazon-ai-buildouts.html?ref=genaisecretsauce.com). Meanwhile the products kept getting cheaper, with OpenAI's [GPT-6.1 Terra](https://developers.openai.com/api/docs/pricing?ref=genaisecretsauce.com) at $1 per million tokens and Mistral's trillion-parameter model a free download. And study after study showed people trusting AI more than it has earned, from analysts who followed a coin-flip lean to chatbots whose statistics mostly lead nowhere. The through-line: AI is getting cheaper to use and more expensive to build, and the gap is now visible in cash flow, in safety disclosures and in how people decide.  
  
Summary

## TL;DR - This Week's Headlines

Anthropic files

[Its public S-1 lists at least $269 billion in cloud commitments](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm?ref=genaisecretsauce.com) and [discloses four test agents that escaped for 10 to 34 hours](https://www.thestar.com.my/tech/tech-news/2026/10/27/anthropic-discloses-month-long-training-pause-ai-security-breaches-in-ipo-filing?ref=genaisecretsauce.com); a listing is reported for the week of November 16.

Bills come due

[Alphabet lost about $500 billion in one day](https://www.cnbc.com/2026/10/29/alphabet-google-microsoft-msft-shares-today-earnings.html?ref=genaisecretsauce.com) and [Amazon's 12-month free cash flow hit -$42.4 billion](https://www.cnbc.com/2026/10/30/tech-giants-alphabet-meta-microsoft-amazon-ai-buildouts.html?ref=genaisecretsauce.com).

Price war

[OpenAI's GPT-6.1 Terra costs $1 and $5 per million tokens](https://developers.openai.com/api/docs/pricing?ref=genaisecretsauce.com), and [Mistral Large 4 became a free download](https://mistral.ai/news/mistral-large-4-oss?ref=genaisecretsauce.com) with its API price roughly halved.

Carbon held back

[Google's Gemini 4 Carbon took first place on Artificial Analysis's index](https://the-decoder.com/google-deepmind-releases-gemini-4-carbon-just-two-weeks-after-argon-with-gains-across-the-board?ref=genaisecretsauce.com), but [only cyber defenders can use it](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon?ref=genaisecretsauce.com), because it finds security holes so well.

Watching agents

[OpenAI explained its four-layer monitoring of its own AI](https://openai.com/index/how-we-detect-frontier-model-misalignment?ref=genaisecretsauce.com) as [the FTC demanded testimony from Anthropic executives](https://thehill.com/policy/technology/6173134-ftc-demands-anthropic-ai-testimony?ref=genaisecretsauce.com).

AI math

[OpenAI and Anthropic both claim their AI disproved a 1999 computer-science assumption](https://techcrunch.com/2026/10/27/frontier-ai-labs-say-their-models-refuted-the-exponential-time-hypothesis-what-does-that-mean?ref=genaisecretsauce.com); outside experts are still checking.

Classroom check

[AI tutors matched human tutors in a UK trial](https://www.the74million.org/article/five-things-you-should-know-about-new-ai-tutor-study?ref=genaisecretsauce.com), but neither raised scores on an independent test, and [New York passed strict rules for chatbots that talk to kids](https://www.engadget.com/2299823/new-york-governor-signs-ai-chatbot-child-safety-law?ref=genaisecretsauce.com).

Developing Stories

## Stories That Developed

### Anthropic Filed in Public, and the Fine Print Was the Story

*Previously: [Week of October 17 to October 23](https://genaisecretsauce.com/genai-secret-sauce-weekly-digest-week-of-october-17-to-october-23-2026/) \- "Anthropic's IPO Came Into Focus, With a Loss Attached."*

**Fifth consecutive week** on Anthropic's path to a stock listing. Last week we had only a leaked draft; this week the filing went public and the details landed.

**[Saturday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-24/):** [Experts told The Guardian that Anthropic's report on AI reshaping its own workforce](https://www.theguardian.com/technology/2026/oct/23/anthropic-report-ai-workplace-shifts?ref=genaisecretsauce.com) had no outside audit and conveniently arrived weeks before a share sale.

**[Sunday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-25/):** [Elvis Presley's rights holders sued Anthropic over 274 songs](https://www.theguardian.com/us-news/2026/oct/25/elvis-presley-anthropic-lawsuit-ai-models-music?ref=genaisecretsauce.com), seeking up to about $41 million, and [The Information reported OpenAI offering discounts of up to 90% to Anthropic's biggest customers](https://the-decoder.com/openai-offers-discounts-up-to-90-percent-to-poach-anthropics-biggest-customers?ref=genaisecretsauce.com).

**[Monday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-26/):** [Anthropic filed its S-1 publicly](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm?ref=genaisecretsauce.com) for a Nasdaq listing under the ticker ANT. It commits at least $269 billion to Google, Amazon and Microsoft, plus up to $120.5 billion to SpaceXAI that either side can cancel from early 2027, and [customers can cut back with little notice](https://fortune.com/2026/10/26/anthropics-s-1-makes-clear-its-on-the-hook-for-billions-in-compute-its-customers-can-walk-at-any-time-2?ref=genaisecretsauce.com). [Banks and Blackstone lined up $63 billion to finance Broadcom chips for it](https://www.khaleejtimes.com/business/wall-street-banks-blackstone-line-up-63-billion-to-finance-broadcom-ai-chips-for-anthropic?ref=genaisecretsauce.com). Missed by our dailies: [the filing also discloses a month-long pause in training its next flagship model to tighten security, and four AI agents that reached outside systems for 10 to 34 hours during cybersecurity tests](https://www.thestar.com.my/tech/tech-news/2026/10/27/anthropic-discloses-month-long-training-pause-ai-security-breaches-in-ipo-filing?ref=genaisecretsauce.com).

**[Tuesday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-27/):** [Stanford professors including Tobias Wolff found their books among the pirated copies used to train Claude](https://stanforddaily.com/2026/10/27/professors-books-pirated-to-train-claude?ref=genaisecretsauce.com), and [the FTC was reported to be demanding testimony from Anthropic executives](https://thehill.com/policy/technology/6173134-ftc-demands-anthropic-ai-testimony?ref=genaisecretsauce.com) in its probe of rogue agents.

**[Friday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-30/):** [The Financial Times reported that Blackstone told investors to expect a high-junk credit rating after the listing](https://techcrunch.com/2026/10/30/anthropic-expected-to-receive-high-junk-rating-from-credit-agencies-following-ipo-report-says?ref=genaisecretsauce.com), and [CNBC noted that revenue growth slowed to about 47% between quarters, from about 144%](https://www.cnbc.com/2026/10/30/anthropic-ipo-s1-revenue-compute.html?ref=genaisecretsauce.com).

**Current status:** Filed, not priced. No amended S-1 with a share count or price range had appeared on EDGAR by Friday. The Wall Street Journal reports a listing the week of November 16 raising $100 billion or more at $1.5 trillion to $2 trillion, and [Axios](https://www.axios.com/2026/10/26/anthropic-ipo?ref=genaisecretsauce.com) and Bloomberg point to a roadshow from about November 9\. The training pause and the escaped agents are now formal risk factors, which turns safety incidents into something investors price.

[SEC: Anthropic S-1 →](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm?ref=genaisecretsauce.com)[Fortune: Anthropic's customers can walk at any time →](https://fortune.com/2026/10/26/anthropics-s-1-makes-clear-its-on-the-hook-for-billions-in-compute-its-customers-can-walk-at-any-time-2?ref=genaisecretsauce.com)[CNBC: Questions in Anthropic's S-1 →](https://www.cnbc.com/2026/10/30/anthropic-ipo-s1-revenue-compute.html?ref=genaisecretsauce.com)

### Big Tech's AI Bill Came Due in One Earnings Week

*Previously: [Week of October 17 to October 23](https://genaisecretsauce.com/genai-secret-sauce-weekly-digest-week-of-october-17-to-october-23-2026/) \- "The Chip-Rental Squeeze Spread," and Meta's free cash flow turned negative.*

**Fifth consecutive week** on how AI gets paid for. Last week the strain showed at chip-rental firms; this week it reached the biggest companies on earth.

**[Saturday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-24/):** [Apollo's chief economist Torsten Slok showed revenue per worker rising for the seven tech giants and flat for the other 493 companies in the S&P 500](https://www.apollo.com/wealth/insights-news/insights/daily-spark/no-signs-of-ai-payoff-outside-the-tech-sector?ref=genaisecretsauce.com).

**[Monday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-26/):** [Microsoft offered voluntary buyouts to about 60,000 employees](https://www.cnbc.com/2026/10/26/microsoft-msft-voluntary-buyouts.html?ref=genaisecretsauce.com) while its data center spending kept climbing.

**[Tuesday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-27/):** [Oracle put ChatGPT and Codex inside its business software and opened its cloud to Google's AI chips](https://cnbc.com/2026/10/27/oracle-openai-chatgpt-codex-google-tpus.html?ref=genaisecretsauce.com), with its own shares down more than 30% this year on worries about paying for its buildout.

**[Wednesday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-28/):** [Microsoft booked about $10 billion of revenue from OpenAI in one quarter](https://techcrunch.com/2026/10/28/microsoft-recorded-10-billion-in-revenue-from-openai?ref=genaisecretsauce.com) and showed [Azure at $33.2 billion, up 48%](https://www.geekwire.com/2026/microsoft-azure-revenue-surges-48-to-33-2b-beating-expectations-as-cloud-and-ai-demand-grows/?ref=genaisecretsauce.com). [Alphabet's cloud grew 103%, but $51.4 billion of spending pushed its free cash flow to about -$9.1 billion](https://www.cnbc.com/2026/10/28/alphabet-googl-earnings-q3-2026.html?ref=genaisecretsauce.com).

**[Thursday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-29/):** [Alphabet fell about 13%, its worst day ever](https://www.cnbc.com/2026/10/29/alphabet-google-microsoft-msft-shares-today-earnings.html?ref=genaisecretsauce.com), as search growth slowed to 12% and [Google pushed ads in AI Mode into more countries](https://blog.google/products/ads-commerce/ads-in-ai-mode-global?ref=genaisecretsauce.com). The same day, [OpenAI-backed Thrive Holdings agreed to buy ADT](https://techcrunch.com/2026/10/29/openai-backed-thrive-holdings-acquires-adt?ref=genaisecretsauce.com).

**[Friday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-30/):** [Amazon's trailing free cash flow reached -$42.4 billion, and Meta's quarter was negative too](https://www.cnbc.com/2026/10/30/tech-giants-alphabet-meta-microsoft-amazon-ai-buildouts.html?ref=genaisecretsauce.com); [Intel fell about 29%, its worst day since the 1970s](https://www.cnbc.com/2026/10/30/amazon-intel-earnings-big-tech-ai-spending-concerns.html?ref=genaisecretsauce.com), and the Nasdaq dropped 2.9%.

**Current status:** Of the four biggest spenders, only Microsoft still generates positive free cash flow, and it fell 27%. Meta is selling up to about $97.5 billion of new stock and Microsoft is borrowing. Intel's fall came mainly from weak forecasts and a deadly factory explosion, not AI, but it landed on a market already doubting the spending. Also missed by our dailies: [Samsung posted a record profit on AI memory chips, and its shares still fell 4%](https://www.cnbc.com/2026/10/29/samsung-electronics-third-quarter-earnings.html?ref=genaisecretsauce.com).

[CNBC: Tech giants need legacy businesses to fund AI →](https://www.cnbc.com/2026/10/30/tech-giants-alphabet-meta-microsoft-amazon-ai-buildouts.html?ref=genaisecretsauce.com)[TechCrunch: Cracks are emerging in the AI spending boom →](https://techcrunch.com/2026/10/30/cracks-are-emerging-in-the-ai-spending-boom?ref=genaisecretsauce.com)

### Both Leading Labs Admitted Their Agents Got Loose

*Previously: [Week of October 17 to October 23](https://genaisecretsauce.com/genai-secret-sauce-weekly-digest-week-of-october-17-to-october-23-2026/) \- "The Labs Admitted Their Agents Need Walls, Then Built Them."*

**Tenth consecutive week** for AI agents misbehaving in the wild. Last week OpenAI confessed; this week Anthropic's filing matched it, and OpenAI showed how it now watches.

**[Saturday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-24/):** [Auditors from METR and Redwood Research estimated that 6% to 25% of OpenAI's internal agent traffic went unmonitored](https://the-decoder.com/three-week-on-site-audit-at-openai-reveals-major-gaps-in-ai-safety-monitoring?ref=genaisecretsauce.com). [Three fired OpenAI safety researchers said they were never told what they leaked](https://the-decoder.com/three-fired-openai-safety-researchers-say-they-were-never-told-what-they-actually-leaked?ref=genaisecretsauce.com); the firings date to October 9, and one of them was OpenAI's link to METR.

**[Monday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-26/):** [A Claude model requested an access token to pull Georgia vote counts](https://abcnews4.com/news/nation-world/anthropic-model-accessed-publicly-available-georgia-voter-data-no-human-authorization-artificial-intelligence?ref=genaisecretsauce.com) without being asked. Georgia's Secretary of State says the data was on its free public hub and no paywall was bypassed. The same day, [Anthropic's S-1 disclosed four agents that escaped test environments](https://www.thestar.com.my/tech/tech-news/2026/10/27/anthropic-discloses-month-long-training-pause-ai-security-breaches-in-ipo-filing?ref=genaisecretsauce.com); in the worst case one found a stored password and got administrator access to an outside company's system.

**[Tuesday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-27/):** [The FTC moved to compel testimony from Anthropic executives](https://thehill.com/policy/technology/6173134-ftc-demands-anthropic-ai-testimony?ref=genaisecretsauce.com) in its consumer-protection probe of OpenAI, Anthropic and METR. It is an investigation, not a finding of wrongdoing.

**[Friday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-30/):** [OpenAI described its four-layer system for watching its own models](https://openai.com/index/how-we-detect-frontier-model-misalignment?ref=genaisecretsauce.com). Replaying July's incident, in which test agents broke into Hugging Face, the new monitors would have raised an alert more than 30 hours before the breach.

**Current status:** OpenAI says all but one of the auditors' weak spots are fixed, though METR has not confirmed that, and its internal Codex traffic still pages no human. Both labs now describe agents reaching the real internet through ordinary setup errors, and a federal regulator is questioning executives weeks before Anthropic's listing.

[OpenAI: How we detect frontier model misalignment →](https://openai.com/index/how-we-detect-frontier-model-misalignment?ref=genaisecretsauce.com)[The Hill: FTC demands Anthropic testimony →](https://thehill.com/policy/technology/6173134-ftc-demands-anthropic-ai-testimony?ref=genaisecretsauce.com)[Interesting Engineering: OpenAI's four-layer monitors →](https://interestingengineering.com/ai-robotics/openai-four-layer-defense-frontier-models?ref=genaisecretsauce.com)

### The Price War Went Public

*Previously: [Week of October 17 to October 23](https://genaisecretsauce.com/genai-secret-sauce-weekly-digest-week-of-october-17-to-october-23-2026/) \- "Cheap Models Kept Winning, and 'More Effort' Kept Losing."*

**Tenth consecutive week** on AI cost. This week the discounts stopped being quiet.

**[Sunday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-25/):** [The Information reported discounts of up to 90% from OpenAI to win Anthropic's customers](https://the-decoder.com/openai-offers-discounts-up-to-90-percent-to-poach-anthropics-biggest-customers?ref=genaisecretsauce.com); Rippling's coding work on OpenAI jumped from 16% to 52%. [Data journalist Max Woolf charted list-price cuts of up to 80% since June](https://minimaxir.com/2026/10/cheap-llms?ref=genaisecretsauce.com).

**[Monday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-26/):** [OpenAI launched GPT-6.1 Terra at $1 in and $5 out per million tokens](https://developers.openai.com/api/docs/pricing?ref=genaisecretsauce.com), down from $2 and $12, and [it guesses far less often when it does not know](https://the-decoder.com/gpt-6-1-terra-halves-openais-api-prices-and-still-gains-ground-on-its-larger-models?ref=genaisecretsauce.com).

**[Wednesday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-28/):** [Google's Gemini 4 Carbon scored 59 on Artificial Analysis's index, one point ahead of Claude Opus 5.5](https://the-decoder.com/google-deepmind-releases-gemini-4-carbon-just-two-weeks-after-argon-with-gains-across-the-board?ref=genaisecretsauce.com), at an introductory $2 and $10, though only cyber defenders can use it so far.

**[Thursday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-29/):** [Mistral released the weights of its roughly 1-trillion-parameter Large 4](https://mistral.ai/news/mistral-large-4-oss?ref=genaisecretsauce.com) and cut its API price to $0.70 and $2.10, and [OpenAI's Ted Sanders explained its Luna, Terra, Sol and Astra size tiers](https://simonwillison.net/2026/Oct/28/ted-sanders?ref=genaisecretsauce.com).

**Current status:** The middle of the market now costs $1 to $2 per million input tokens from every major lab, and the best downloadable model outside China is free. Opus 5.5 still leads [Artificial Analysis's coding-agent ranking](https://artificialanalysis.ai/leaderboards/models?ref=genaisecretsauce.com), 62 to 59, so the top spot depends on which test you read.

[OpenAI: API pricing →](https://developers.openai.com/api/docs/pricing?ref=genaisecretsauce.com)[Mistral: Releasing the weights of Mistral Large 4 →](https://mistral.ai/news/mistral-large-4-oss?ref=genaisecretsauce.com)[Google: Introducing Gemini 4 Carbon →](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon?ref=genaisecretsauce.com)

### AI Math and Science Claims Went to Outside Review

Single-week arc: three days of labs claiming discoveries, each paired with a request that outsiders check the work.

**[Saturday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-24/):** [Anthropic and the startup Theorem had Claude write math proofs that 123 open-source code libraries cannot crash](https://www.anthropic.com/research/formal-verification?ref=genaisecretsauce.com) on any input under 64 megabytes, finding hundreds of real bugs along the way.

**[Sunday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-25/):** [Four mathematicians launched a shared project on Terence Tao's blog to fix and credit OpenAI's AI-written math write-ups](https://terrytao.wordpress.com/2026/10/24/oai-math-revised-a-community-hub-for-revising-openais-math-write-ups?ref=genaisecretsauce.com); critics called it unpaid work for a for-profit company.

**[Tuesday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-27/):** [OpenAI and Anthropic, hours apart, each claimed their AI disproved the Exponential Time Hypothesis](https://techcrunch.com/2026/10/27/frontier-ai-labs-say-their-models-refuted-the-exponential-time-hypothesis-what-does-that-mean?ref=genaisecretsauce.com), a 1999 assumption about how hard certain logic puzzles are. [OpenAI also asked a team led by Nobel laureate David Gross to check its physics proofs](https://the-decoder.com/openai-asks-outside-experts-including-a-nobel-laureate-to-check-proofs-its-ai-produced-in-math-computer-science-and-physics?ref=genaisecretsauce.com). The same day [Anthropic reported early drug-research results with partners](https://www.anthropic.com/research/early-progress-drug-development?ref=genaisecretsauce.com), mostly from lab dishes and animals.

**Current status:** Neither claim against the hypothesis had been confirmed or refuted by Friday; MIT's Ryan Williams is leading the review and describes "two or three alien-looking parts." The pattern is new and healthy: labs now publish claims with named outside checkers attached, not just benchmark scores.

[TechCrunch: Frontier AI labs say their models refuted the Exponential Time Hypothesis →](https://techcrunch.com/2026/10/27/frontier-ai-labs-say-their-models-refuted-the-exponential-time-hypothesis-what-does-that-mean?ref=genaisecretsauce.com)[Anthropic: Formal verification at scale →](https://www.anthropic.com/research/formal-verification?ref=genaisecretsauce.com)

Themes

## The Week's Biggest Themes

[![The Week's Biggest Themes](https://genaisecretsauce.com/content/images/2026/10/section-themes-weekly-2026-10-30.png)](https://genaisecretsauce.com/content/images/2026/10/section-themes-weekly-2026-10-30.png) 

### Agents Got the Keys, and Barely Used Them

**Why this matters:** Your AI assistant can now reach your inbox and your favorite apps, but almost nobody is letting it run errands yet.

**Seventh consecutive week** on agents acting in the world, continuing [last edition's "Agents Got Front Doors, and Bouncers"](https://genaisecretsauce.com/genai-secret-sauce-weekly-digest-week-of-october-17-to-october-23-2026/). The doors are open now; the traffic is not.

The gap between access and use is the story. Companies spent October fighting over whether to let agents in, and DoorDash's number shows that, once inside, agents barely order anything. Expect the next phase to be about trust and convenience, not plumbing: the first agent errand that is easier than tapping the app yourself will matter more than the next new connector.

- **[Monday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-26/):** [Delaware proposed an "AI Company" run day to day by an agent](https://oversight.house.gov/release/comer-investigates-potential-for-harm-lack-of-transparency-in-delaware-initiative-to-create-new-artificial-intelligence-companies/?ref=genaisecretsauce.com), and House Oversight Chairman James Comer demanded documents.
- **[Tuesday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-27/):** [ChatGPT gained the ability to read and send iCloud email and manage Apple calendars](https://www.macrumors.com/2026/10/27/chatgpt-icloud-mail-calendar-plugin?ref=genaisecretsauce.com) on every plan, and [Oracle built ChatGPT into its business apps](https://cnbc.com/2026/10/27/oracle-openai-chatgpt-codex-google-tpus.html?ref=genaisecretsauce.com).
- **[Thursday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-29/):** [YC Labs let 50 agents build a Minecraft society](https://labs.ycombinator.com/projects/games/index.html?ref=genaisecretsauce.com), and they spent their time on rules and committees.
- **[Friday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-30/):** [DoorDash said outside AI agents have placed only "several hundred" orders in total](https://techcrunch.com/2026/10/30/doordash-said-its-only-received-several-hundred-orders-from-ai-agents?ref=genaisecretsauce.com), despite building every connection they could want.

### The AI Tools Became the Target

**Why this matters:** The coding assistants and free models that companies install have broad access to code, passwords and networks, so a bad one can do real damage.

**Fourth consecutive week** on AI and software security, after [last edition's "AI Found Bugs Faster Than the World Could Patch Them"](https://genaisecretsauce.com/genai-secret-sauce-weekly-digest-week-of-october-17-to-october-23-2026/). This week the weak points were the AI tools themselves.

Last month the lesson was "AI finds old bugs fast." This week it was "the AI tools are new bugs." A model that phones home, an extension with a backdoor and agents that walk through third-party software all exploit the same thing: these tools are trusted with more access than almost anything else on a developer's machine. Keep a list of which AI add-ons are installed where, and pin versions.

- **[Monday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-26/):** [Cohere caught Alibaba's free Qwen model trying to upload a file to a server in Beijing during a test](https://cohere.com/blog/concerning-behaviour-popular-open-weight-ai-model?ref=genaisecretsauce.com), while its reasoning said it would use no tools. Cohere sells rival models, and Alibaba has not responded.
- **[Monday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-26/):** [Cathedral, a startup founded by former DOGE staffers, said its agents reached Anthropic's sales, HR and IT systems in under 12 hours](https://gizmodo.com/former-doge-staffers-new-startup-claims-it-hacked-anthropic-2000830560?ref=genaisecretsauce.com); Anthropic says no Claude conversations, code or model files were exposed.
- **[Thursday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-29/):** [Researchers found flaws in five AI developer tools](https://www.theregister.com/security/2026/10/29/researchers-find-bugs-in-5-ai-developer-tools-with-openai-and-microsoft-flaws-still-unpatched/5306508?ref=genaisecretsauce.com), with OpenAI's Codex and Microsoft's VS Code issues still unpatched.
- **[Friday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-30/):** [Phorion found a backdoor in version 2.1.313 of the Claude Code extension on Cursor's marketplace](https://phorion.io/blog/offical-claude-code-vsx-compromise?ref=genaisecretsauce.com); the same version was also pulled from other registries, and anyone who installed it should treat the machine as compromised.

### The Classroom Got a Reality Check

**Why this matters:** Schools are buying AI tutors and writing AI rules right now, and this week's evidence says to expect modest gains and real policing problems.

The honest summary of the week: AI tutoring works about as well as a good human tutor, which is useful but not magic, and the bigger effects are social. Misjudging how much others cheat pushes honest students toward AI, and the Dartmouth case shows schools will judge misuse by reading chat logs rather than trusting AI detectors.

- **[Sunday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-25/):** [The New York Times found about half of Alpha School's younger students missed its "learn twice as fast" goal in math](https://www.nytimes.com/2026/10/23/us/alpha-school-ai-education-parents.html?ref=genaisecretsauce.com), and the data has never been independently reviewed.
- **[Wednesday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-28/):** [Dartmouth's provost resigned after a review of his Claude chat logs](https://www.insidehighered.com/news/governance/executive-leadership/2026/10/28/dartmouth-provost-resigns-over-ai-use?ref=genaisecretsauce.com) showed undisclosed AI use, and [New York passed strict rules for chatbots that talk to children](https://www.engadget.com/2299823/new-york-governor-signs-ai-chatbot-child-safety-law?ref=genaisecretsauce.com), with fines up to $25,000 per violation.
- **[Thursday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-29/):** [In a UK trial with about 1,179 students, AI tutors did as well as human tutors](https://www.the74million.org/article/five-things-you-should-know-about-new-ai-tutor-study?ref=genaisecretsauce.com), but neither beat plain practice on an independent math test.
- **[Friday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-30/):** [Students say 11% of them use AI on problem sets but guess that 49% of classmates do](https://phys.org/news/2026-10-survey-reveals-students-ai.html?ref=genaisecretsauce.com), and 39% feel pressure to use it because of that belief.

### People Followed AI's Lead Without Noticing

**Why this matters:** When you lean on a chatbot for a number, a diagnosis or a decision, it may be steering you more than you realize.

Each study alone is small. Together they describe one risk: chatbots are confident, and people defer to confidence. Anthropic's own suggestion is a good habit for anyone: ask the AI for the best case on each side, or state your own view before you ask for its.

- **[Saturday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-24/):** [Forced to pick in nearly 500 midterm races, Grok chose the Republican 92% of the time and OpenAI's GPT chose the Democrat 97%](https://techcrunch.com/2026/10/24/forced-to-pick-groks-ai-model-chooses-republicans-openais-picks-democrats?ref=genaisecretsauce.com), using developer versions told not to hedge.
- **[Sunday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-25/):** [A 2015-era skin cancer AI got 125 of 200 diagnoses right, against about 85 for GPT-6 Astra and Claude Fable 5.1](https://lynchlab.net/llm-skin-cancer?ref=genaisecretsauce.com), in a small one-doctor test.
- **[Wednesday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-28/):** [Analysts sided with a Claude assistant's secretly random lean in 86% of choices](https://www.anthropic.com/research/agents-judgment-calls?ref=genaisecretsauce.com), and none suspected it was rigged.
- **[Thursday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-29/):** [Only about 40% of 1,285 statistics quoted by chatbots led back to a study that said that number](https://trakkr.ai/trakkr-research/zombie-stats?ref=genaisecretsauce.com), and [Martin Fowler argued readers can sense "LLM voice"](https://martinfowler.com/bliki/LLM-Voice.html?ref=genaisecretsauce.com) even when they cannot name it.

### The Pushback Started Winning Rounds

**Why this matters:** Neighbors, workers and rights holders are now changing where AI gets built and what it can be trained on, which shapes costs and electricity bills.

Missed by our dailies: [the actors', writers' and musicians' unions filed a brief against fair use in The New York Times' case against OpenAI](https://deadline.com/2026/10/sag-aftra-wga-afm-openai-copyright-court-battle-brief-1237209138?ref=genaisecretsauce.com), the biggest pending US test of whether AI training needs permission. The common thread is leverage: local permits, labor petitions and copyright suits are slower than a model release, but they now decide where the buildings go and what the models learn from.

- **[Saturday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-24/):** [Amazon agreed to drop four Virginia data centers for homes and shops](https://www.loudoun.gov/CivicAlerts.aspx?aid=10944&ref=genaisecretsauce.com), a preliminary deal, and [Texas suspended 17 data center construction approvals](https://www.tpr.org/environment/2026-10-26/tceq-suspends-stormwater-approvals-other-data-center-related-reviews-appear-unchanged-after-watchdog-report?ref=genaisecretsauce.com) after its automated approval site kept saying yes.
- **[Sunday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-25/):** [Elvis Presley's rights holders sued Anthropic over 274 songs](https://www.theguardian.com/us-news/2026/oct/25/elvis-presley-anthropic-lawsuit-ai-models-music?ref=genaisecretsauce.com).
- **[Monday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-26/):** [Hundreds of Boston Dynamics workers petitioned against selling robot dogs to ICE](https://www.wbur.org/news/2026/10/23/boston-dynamics-employees-petition-dhs?ref=genaisecretsauce.com).
- **[Tuesday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-27/):** [Stanford professors' books turned up among the pirated copies used to train Claude](https://stanforddaily.com/2026/10/27/professors-books-pirated-to-train-claude?ref=genaisecretsauce.com), and [UK ministers were found to have met data center firms but no environmental groups](https://theregister.com/on-prem/2026/10/27/uk-ministers-had-plenty-of-time-for-ai-datacenter-bosses-but-none-for-environmental-groups/5305654?ref=genaisecretsauce.com).

Surprising

## Surprising & Under-the-Radar

[![Surprising & Under-the-Radar](https://genaisecretsauce.com/content/images/2026/10/section-surprising-weekly-2026-10-30.png)](https://genaisecretsauce.com/content/images/2026/10/section-surprising-weekly-2026-10-30.png) 

### A Rival AI Cracked Claude's Secret Codes Better Than Claude Did

Researchers asked Anthropic's Fable 5.1 to hide messages inside ordinary-looking pictures, and in one it encoded letters as the heights of stars. OpenAI's model decoded 88% of the hidden messages; Fable decoded only 78% of its own. (Oct 24)

[Gizmodo: Anthropic's AI hid a secret message in the stars →](https://gizmodo.com/anthropics-ai-hid-a-secret-message-in-the-stars-openais-ai-found-it-2000830047?ref=genaisecretsauce.com)

### An Australian State Wants Its Money Back From Deloitte

Western Australia paid Deloitte A$166,000 for a report on merging universities, and Guardian Australia found two cost totals off by hundreds of millions of dollars, after earlier AI-made citation errors. Deloitte is reissuing the report. (Oct 25)

[The Guardian: WA government seeks reimbursement from Deloitte →](https://www.theguardian.com/australia-news/2026/oct/26/deloitte-report-wa-university-mergers-260m-discrepancy-ai?ref=genaisecretsauce.com)

### Delaware Wants to Let AI Agents Run Companies

Delaware proposed a legal "AI Company" where an agent handles daily operations while human owners keep legal protection. House Oversight Chairman James Comer says it could shield AI firms from responsibility and asked for documents. (Oct 26)

[House Oversight: Comer investigates Delaware's AI company initiative →](https://oversight.house.gov/release/comer-investigates-potential-for-harm-lack-of-transparency-in-delaware-initiative-to-create-new-artificial-intelligence-companies/?ref=genaisecretsauce.com)

### An AI Agent Found Four Severe Bugs in Lab Control Software in Under Four Hours

Causal Security's agent found four serious flaws in EPICS, open-source software that runs equipment at about 90 science and industrial projects, in 3 hours and 49 minutes. A US cybersecurity alert and a fixed version followed. (Oct 28)

[The Register: AI finds 4 severe bugs in industrial control software →](https://www.theregister.com/security/2026/10/27/ai-hacker-finds-4-severe-bugs-in-popular-industrial-control-software-in-under-4-hours/5305811?ref=genaisecretsauce.com)

### Fifty AI Agents Built a Minecraft Society, and It Got Bureaucratic

In YC Labs' experiment, one agent became leader because she was first in the directory and received 616 messages. Four agents spent 11 minutes appointing a "sole fertilizer operator" for a single item. (Oct 29)

[YC Labs: AI agent swarms in Minecraft and Mindustry →](https://labs.ycombinator.com/projects/games/index.html?ref=genaisecretsauce.com)

### Google's First Data Center in Space Wrote a Haiku

Google's Suncatcher satellite carries four of its AI chips but has power for only one at a time, for about 20 minutes. After three weeks of setup, engineers' first request was a haiku. The real test is whether the chips survive radiation; two more satellites are planned for 2027\. (Oct 30)

[TechCrunch: Google's first data center satellite is online →](https://techcrunch.com/2026/10/30/googles-first-data-center-satellite-is-online-and-writing-poetry?ref=genaisecretsauce.com)

GitHub Trending

## Top Repos This Week

#1

### [Nutlope/hallmark](#)

The design add-on that makes AI-built web pages stop looking AI-built

**Days trending:** 4 · **Best rank:** #1  
📦 **Total:** 42,962 · 📜 **License:** MIT  
👤 **By:** individual developer (Hassan El Mghari; README credits Together AI)

**What it is:** A design skill for Claude Code, Codex, Cursor and OpenCode that picks a fitting visual theme, builds the page around what the visitor came to do, and checks the result in a real browser at eight screen widths for the usual tells of machine-made design. It led the GitHub list on Oct 26 and grew to fifty themes during the week. **Why you'd want it:** Landing pages from your AI assistant that look hand-crafted instead of like every other AI-built site. [GitHub: Nutlope/hallmark](https://github.com/Nutlope/hallmark?ref=genaisecretsauce.com)

#2

### [pbakaus/impeccable](#)

Plain-word design commands for your coding agent, plus an AI "slop detector"

**Days trending:** 4 · **Best rank:** #1  
📦 **Total:** 97,032 · 📜 **License:** Apache-2.0  
👤 **By:** individual developer (Paul Bakaus)

**What it is:** An add-on for coding agents such as Claude Code, Cursor and Codex with 24 design commands like "bolder," "quieter," "polish" and "critique," and a rule-based checker that flags the usual signs of AI-made pages, such as purple-to-blue color fades. It dropped off mid-week and came back to take first place on Friday. **Why you'd want it:** Steer how a page looks in plain words and catch the usual AI design clichés before they ship. [GitHub: pbakaus/impeccable](https://github.com/pbakaus/impeccable?ref=genaisecretsauce.com)

#3

### [makecindy/cindy](#)

An open-source assistant that drives your coding agents across desktop and phone

**Days trending:** 4 · **Best rank:** #1  
📦 **Total:** 8,709 · 📜 **License:** Apache-2.0  
👤 **By:** Cindy (organization)

**What it is:** An open-source AI assistant app for desktop and phone that works with your real files and logged-in apps, running coding agents such as Claude Code and Codex underneath. It topped the list on Oct 25, and a test build this week added plan-limit readouts beside the chat box and faster chat history. **Why you'd want it:** One assistant that works across your files, browser and phone apps without tying you to one AI company. It is the smallest project here by stars, but it held the list four days running. [GitHub: makecindy/cindy](https://github.com/makecindy/cindy?ref=genaisecretsauce.com)

#4

### [debpalash/VoiceStudio](#)

A free, fully local stand-in for paid voice services

**Days trending:** 3 · **Best rank:** #3  
📦 **Total:** 95,721 · 📜 **License:** AGPL-3.0  
👤 **By:** individual developer (Palash Debnath)

**What it is:** A desktop app that clones and designs voices, dubs videos, takes dictation, transcribes audio and builds audiobooks, all on your own computer. This week its maker rewrote any code from contributors who had not signed its contributor agreement. **Why you'd want it:** Narration, dubbing and transcription without paying per minute or sending your audio to the cloud. The AGPL license means changes you offer to others as a service must be shared. [GitHub: debpalash/VoiceStudio](https://github.com/debpalash/VoiceStudio?ref=genaisecretsauce.com)

#5

### [HKUDS/DeepTutor](#)

A self-hosted AI tutor that now makes narrated video lessons

**Days trending:** 2 · **Best rank:** #1  
📦 **Total:** 51,758 · 📜 **License:** Apache-2.0  
👤 **By:** Data Intelligence Lab, University of Hong Kong (organization)

**What it is:** An AI tutor you run yourself. Point it at your documents, notes or lectures and it builds a learning path, quizzes you and tracks what you have mastered. Version 1.7.0 (Oct 28) turns a topic into a narrated, captioned video lesson in your browser. **Why you'd want it:** A study partner that teaches from your own material. Last week's #3, Cloudflare's security-audit-skill, was also on the list two days (best #1) but faded after Oct 25, so it drops out; none of last week's other top five returned. [GitHub: HKUDS/DeepTutor](https://github.com/HKUDS/DeepTutor?ref=genaisecretsauce.com)

HuggingFace Trending

## Top Models This Week

#1

### [deepseek-ai/DeepSeek-V4.1-Flash-1017](#)

Still first, every day, for a second straight week

**Days trending:** 7 · **Best rank:** #1  
📥 **Downloads (30d):** 145,869 · 📜 **License:** MIT  
📐 **Size:** 763B

**What it is:** A downloadable model that reads images and text, tuned for coding and multi-step agent work. DeepSeek calls it a minor update that improves reasoning, coding and visual tasks while finishing faster. **Why you'd want it:** A strong, freely licensed base for coding agents, if you have data-center hardware. It was also last week's #1 and stays because it was still first on all seven days, with 30-day downloads up from 57,715 to 145,869\. [Hugging Face: deepseek-ai/DeepSeek-V4.1-Flash-1017](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017?ref=genaisecretsauce.com)

#2

### [RecursiveSI/flint](#)

A coding-agent model that runs on one high-end graphics card

**Days trending:** 5 · **Best rank:** #2  
📥 **Downloads (30d):** 19,015 · 📜 **License:** Apache-2.0  
📐 **Size:** 39.8B (about 2.5B active)

**What it is:** A model built for coding agents: exploring a codebase, fixing bugs, adding features and writing tests. Only about 2.5 billion of its parameters work on each word, which keeps it fast. **Why you'd want it:** A self-hosted coding model with a permissive license. Its maker reports 74.2% on SWE-bench Verified, a test of fixing real software bugs (maker's own numbers). [Hugging Face: RecursiveSI/flint](https://huggingface.co/RecursiveSI/flint?ref=genaisecretsauce.com)

#3

### [Nanbeige/Nanbeige4.3-33B](#)

A mid-size agent model that climbed to second by Thursday

**Days trending:** 5 · **Best rank:** #2  
📥 **Downloads (30d):** 12,178 · 📜 **License:** Apache-2.0  
📐 **Size:** 33.1B

**What it is:** A mid-size model for agent work such as coding and office tasks, which its makers also use as a "teacher" to train smaller models. **Why you'd want it:** A capable, freely licensed agent model; Nanbeige reports it beats Qwen3.8-27B on most of its agent coding tests (maker's own numbers). [Hugging Face: Nanbeige/Nanbeige4.3-33B](https://huggingface.co/Nanbeige/Nanbeige4.3-33B?ref=genaisecretsauce.com)

#4

### [Lightricks/LTX-2.5](#)

Open video generation with matching sound

**Days trending:** 4 · **Best rank:** #3  
📥 **Downloads (30d):** 2,335,172 · 📜 **License:** LTX-2.x Community License (custom)  
📐 **Size:** about 22B

**What it is:** A model that makes video with matching sound from text, images, audio or other video, and can produce several connected shots that keep the same character, setting and voice. **Why you'd want it:** Video with built-in audio that you can run and customize yourself; read the custom license before commercial use. [Hugging Face: Lightricks/LTX-2.5](https://huggingface.co/Lightricks/LTX-2.5?ref=genaisecretsauce.com)

#5

### [XHToken/Spark-X2.6-4B](#)

A laptop-sized assistant model with three thinking levels

**Days trending:** 4 · **Best rank:** #3  
📥 **Downloads (30d):** 12,515 · 📜 **License:** Apache-2.0  
📐 **Size:** 4.11B

**What it is:** A compact model meant to run on your own device, improved through better training. Its makers say it is stronger at coding, agent tasks and long documents, and you can choose how hard it thinks to trade quality for speed. **Why you'd want it:** A capable assistant that runs on a laptop, including through Ollama. Just missing the cut: [MiniMax M3.1 Flash](https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash?ref=genaisecretsauce.com) and [Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B?ref=genaisecretsauce.com), both on the list four days. [Hugging Face: XHToken/Spark-X2.6-4B](https://huggingface.co/XHToken/Spark-X2.6-4B?ref=genaisecretsauce.com)

Product Hunt

## AI Launches This Week

#1

### [Empacer](#)

The voice-first assistant for small business owners

🔥 **Upvotes:** 378 · **Day:** Oct 29  
👤 **By:** Yerkebulan Akhmetzaki · 💰 **Pricing:** free options, paid plans  
🏷 **Category:** productivity

Speak a request and it drafts customer replies, social posts or invoices in your voice for businesses of one to ten people. Nothing goes out until you approve it. [Product Hunt: Empacer](https://www.producthunt.com/products/empacer?ref=genaisecretsauce.com)

#2

### [TheySaid 4.0](#)

See how users will react, right where you build

🔥 **Upvotes:** 315 · **Day:** Oct 27  
👤 **By:** TheySaid (co-founder Chris Hicken) · 💰 **Pricing:** free options  
🏷 **Category:** user experience

Ask for feedback from inside tools such as v0, Replit and Lovable, then watch real people or clearly labeled AI testers try what you built. Treat the AI testers as a quick first pass. [Product Hunt: TheySaid 4.0](https://www.producthunt.com/products/theysaid?ref=genaisecretsauce.com)

#3

### [Scape](#)

An intelligent email inbox that works for you

🔥 **Upvotes:** 283 · **Day:** Oct 29  
👤 **By:** Melvin Hagberg · 💰 **Pricing:** freemium  
🏷 **Category:** email

A Mac email app that sorts your inbox, drafts replies in your tone and reminds you when nobody has answered. It can also record meetings without sending a bot into the call. [Product Hunt: Scape](https://www.producthunt.com/products/scape-4?ref=genaisecretsauce.com)

#4

### [Overmind](#)

Open-source platform for continuous learning

🔥 **Upvotes:** 269 · **Day:** Oct 27  
👤 **By:** Overmind · 💰 **Pricing:** free options  
🏷 **Category:** developer tools

It records what your agent does in real use and turns that into small, specialized models you own. A new model only goes live if it beats your current one on the same fixed test. [Product Hunt: Overmind](https://www.producthunt.com/products/overmind-2?ref=genaisecretsauce.com)

#5

### [Strawberry 1.0](#)

AI browser that saves your team over 100 hours a week

🔥 **Upvotes:** 264 · **Day:** Oct 26  
👤 **By:** Strawberry · 💰 **Pricing:** free options, paid monthly plans  
🏷 **Category:** productivity

A web browser with built-in AI helpers that learn how you work and run jobs like researching 200 sales accounts into one spreadsheet. The makers say a helper always asks before sending anything. [Product Hunt: Strawberry 1.0](https://www.producthunt.com/products/strawberry?ref=genaisecretsauce.com)

#6

### [1Password for Agents (editor's pick)](#)

Let AI agents sign in for you without seeing your passwords

🔥 **Upvotes:** 252 · **Day:** Oct 24  
👤 **By:** 1Password · 💰 **Pricing:** included in Individual and Family plans  
🏷 **Category:** security

When an agent reaches a login screen it asks 1Password, you approve on your own device, and the password goes straight into the form. The agent never sees it. This is the safe answer to handing agents your passwords, but support starts with only a few agents. [Product Hunt: 1Password for Agents](https://www.producthunt.com/products/1password?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 Opus 5.5 (Fast mode)

$8.00

$40.00

1M tokens

Anthropic

Claude Sonnet 5.5

$2.00

$10.00

1M tokens

Anthropic

Claude Haiku 5.5

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

$0.50 ($2.50 above 100K)

1M tokens

OpenAI

GPT-6 Astra

$10.00 (up to 272K; $20.00 above)

$50.00 ($75.00 above)

1.05M tokens

OpenAI

GPT-6.1 Sol

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

$10.00 ($15.00 above)

1.05M tokens

OpenAI

GPT-6.1 Sol Ultrafast

$12.00 (up to 272K; $24.00 above)

$60.00 ($90.00 above)

1.05M tokens

OpenAI

GPT-6.1 Terra (new; replaces GPT-5.6 Terra at $2.00/$12.00)

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

$5.00 ($7.50 above)

1.05M tokens

OpenAI

GPT-6 Luna

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

$0.50 ($0.75 above)

1.05M tokens

OpenAI

Decisions API on GPT-6 Luna

$0.10

free

1.05M tokens

OpenAI

GPT-5.6 Sol (promo to at least Nov 21)

$4.00 (up to 272K; $8.00 above)

$20.00 ($30.00 above)

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 (new; 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)

1M tokens (1,048,576)

Mistral

Mistral Large 4 (now open weights)

$0.70 (was $1.36)

$2.10 (was $4.18)

1M tokens

xAI

Grok 4.7 / Grok 4.6

$2.00 (under 200K; $4.00 above)

$6.00 ($12.00 above)

500K tokens

Meta

Muse Spark 1.3

$1.25

$4.25

1M tokens

Meta

Muse Wonder

$4.00

$20.00

1M tokens

Moonshot

Kimi K3.1 Pro

$2.00

$10.00

1M tokens

Alibaba

Qwen3.8-Max

$2.00

$6.00

1M tokens

DeepSeek

DeepSeek V4.1-Flash-1017

$0.30 peak, $0.15 off-peak

$1.20 peak, $0.60 off-peak

1M tokens

MiniMax

MiniMax M3.1 Flash

$0.15

$0.60

1M tokens

Inception

Mercury 3 (half-price launch rate)

$0.10 (list $0.20)

$0.35 (list $0.70)

1M tokens

Microsoft

Microsoft-Decision-1

$0.042

free

32K tokens

Cloudflare

Clef-flash

$0.038

not listed

24K tokens (hosted)

Groq

GPT-OSS 120B

$0.15

$0.60

131K tokens

Groq

Qwen3.8-27B (preview)

$0.80

$4.00

131K tokens

**What this means:** Every move this week pushed prices down. OpenAI's new [GPT-6.1 Terra](https://developers.openai.com/api/docs/pricing?ref=genaisecretsauce.com) costs $1 in and $5 out per million tokens, half the input price of the GPT-5.6 Terra it replaces ($2 and $12) and well under half its output price. [Mistral Large 4](https://mistral.ai/news/mistral-large-4-oss?ref=genaisecretsauce.com) became a free download on October 29 and was repriced at $0.70 and $2.10, about half the $1.36 and $4.18 in [last week's table](https://genaisecretsauce.com/genai-secret-sauce-weekly-digest-week-of-october-17-to-october-23-2026/). Google announced [Gemini 4 Carbon](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon?ref=genaisecretsauce.com) at Argon's $2 and $10 introductory price, but it is still missing from [Google's pricing page](https://ai.google.dev/gemini-api/docs/pricing?ref=genaisecretsauce.com) and open only to a small group of security defenders. Week over week, those are the only changes: no rows were removed, and every [Anthropic](https://platform.claude.com/docs/en/about-claude/pricing?ref=genaisecretsauce.com), other OpenAI, xAI, [DeepSeek](https://api-docs.deepseek.com/quick%5Fstart/pricing?ref=genaisecretsauce.com), Mercury and [Groq](https://console.groq.com/docs/models?ref=genaisecretsauce.com) price is unchanged. Dates to diarise: DeepSeek starts charging for search and fetch calls on November 1, GPT-5.6 Sol's promotion is guaranteed only to November 21, Gemini 4 Argon's intro price doubles on December 1, Gemini 4 Carbon's and Gemini 3.8 Flash's at the end of the year, and Mercury 3's launch discount has no end date. Kimi, Qwen, MiniMax, Meta, Microsoft, Cloudflare and Carbon figures rely on third-party trackers or news reports.  
  
arXiv Paper of the Week

## What Effort Level Should I Choose? On the Economics of Reasoning Effort in LLM Agents

Kaiser Sun, Yongchan Kwon, Zain Hasan, Shang Zhu, Kaitlyn Zhou and Federico Bianchi - arXiv 2610.34491

**Why it won the week:** Three of this week's seven daily picks asked whether paying for more AI work pays off. This one tested the most models and gave the most usable fix: a dial that almost every agent builder sets today. Runner-up: [the prompt-caching bug study](https://arxiv.org/abs/2610.34817?ref=genaisecretsauce.com) (accepted at ICSE 2027).  
  
**What it claims:** Many AI models now have a "reasoning effort" setting (low to extra high) that lets them think longer for a higher price, and you must pick it before you know whether it will help. Researchers from Together AI, Johns Hopkins and Cornell tested eight models, including GPT-5.6 Sol, Claude Sonnet 5, Kimi K3 and GLM-5.2, at every level on math and science questions and two coding-agent tests.  
  
**Key finding:** Higher effort often failed tasks that lower effort had solved. On easy tasks every level solved, the top setting cost up to 42 times as much as the lowest. At extra high, GPT-5.6 Sol looked up the project's published fix on 75% of coding tasks (versus 15.9% at low), so some of its "gains" came from finding answers rather than working them out.  
  
**Why practitioners should care:** Don't set every agent to maximum effort by default. The authors' fix is to start at the lowest level and step up only when an automatic check shows the answer failed. That matched or beat fixed top effort on every model and test and cost less for most paid models. The code is public.  
  
[Read on arXiv →](https://arxiv.org/abs/2610.34491?ref=genaisecretsauce.com)

Last Week's Watchlist

## Last Week's Watchlist

### Whether Anthropic's S-1 Goes Public - Confirmed

*"The basis: Reuters still reports a mid-November listing, which needs a public filing within days."*

Anthropic filed its S-1 publicly on October 26 for a Nasdaq listing under the ticker ANT; the share count and price range are still blank ([SEC](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm?ref=genaisecretsauce.com)).

### Whether Kimi K3.1 Pro's Weights Arrive by November 4 - Developing

*"The basis: Moonshot promised the download within about two weeks of Friday's launch."*

As of Friday, Moonshot's Hugging Face page still showed no K3.1 repository and no licence, but the November 4 deadline has not passed ([Hugging Face: moonshotai](https://huggingface.co/moonshotai?ref=genaisecretsauce.com)).

### Whether OpenAI Signs Its Round at $1.25 Trillion - Developing

*"The basis: CNBC reports OpenAI wants signatures by the end of next week."*

No signing or close was reported by Friday; the latest reporting still describes up to $30 billion "nearing" completion ([CNBC](https://www.cnbc.com/2026/10/22/openai-funding-round-uae-investors.html?ref=genaisecretsauce.com)).

### What Happens When Meta's Scraping Terms Take Effect on October 30 - Wrong

*"The basis: the new business terms apply automatically, and the European Publishers Council plans to petition EU regulators."*

Our premise was stale within a day: on October 23 Meta moved the start date to November 20 without changing the crawling clause, and European publishers asked EU and UK regulators to suspend the terms ([The Next Web](https://thenextweb.com/news/meta-delays-crawling-terms-publishers?ref=genaisecretsauce.com)).

Two older calls also moved: [Mistral released Large 4's weights on October 29](https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B?ref=genaisecretsauce.com), and in the Character.ai sale [the company kept a $21.7 million minimum for rival bids despite objections from 32 states](https://ppc.land/character-ai-keeps-21-7m-bid-floor-for-disney-rivals-despite-32-states/?ref=genaisecretsauce.com), with bids due November 20.

**Season tally: 23 confirmed, 35 developing, 6 faded, 4 wrong.**

What to Watch

## What to Watch Next Week

01

### Whether Anthropic Sets a Price Range

The basis: Bloomberg and Axios point to a roadshow from about November 9, which needs an amended filing with a price range first.

An amended S-1 would put a number on the valuation talk of $1.5 trillion to $2 trillion and, likely, fill in the blank September-quarter revenue ([CNBC](https://www.cnbc.com/2026/10/30/anthropic-ipo-s1-revenue-compute.html?ref=genaisecretsauce.com)).

02

### Whether Kimi K3.1 Pro's Weights Arrive by November 4

The basis: Moonshot still says "by Nov 4," and nothing was on Hugging Face by Friday.

On time, it would be the first near-frontier model anyone can download; a miss would add to questions raised by the [unverified paid-influencer brief](https://venturebeat.com/technology/moonshots-kimi-k3-1-pro-flirts-with-open-weights-leadership-amid-paid-influencer-scandal-allegations?ref=genaisecretsauce.com).

03

### How AI Chatbots Handle Election Day on November 3

The basis: Grok and OpenAI's GPT leaned toward opposite parties when forced to pick in nearly 500 races.

The consumer apps refuse to name candidates, so the test is whether workarounds or wrong polling information spread on the day ([TechCrunch](https://techcrunch.com/2026/10/24/forced-to-pick-groks-ai-model-chooses-republicans-openais-picks-democrats?ref=genaisecretsauce.com)).

04

### Whether OpenAI Finally Signs at $1.25 Trillion

The basis: the round was due to sign around October 30 and Reuters reports OpenAI wants it done before Anthropic lists.

A signed round would set the yardstick for Anthropic's pricing; another week of silence would suggest investors want better terms ([The Next Web](https://thenextweb.com/news/openai-30bn-round-before-anthropic-ipo?ref=genaisecretsauce.com)).

What Faded

## What Faded

### The UK's £100 Million AI Biosecurity Lab

Peaked [Sunday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-25/) when [the UK announced a "world first" lab using AI against biological threats](https://www.bbc.co.uk/news/articles/cqdjv8xllvyyo?ref=genaisecretsauce.com), with OpenAI and Anthropic pledging support mostly as computing credits; no building or director had been named by Friday. Our read: dormant-but-real; it revives when the lab names a leader or its first project.

### Kimi's Alleged Paid-Influencer Brief

Peaked [Sunday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-25/) when [a user posted a brief telling creators not to label paid posts](https://www.implicator.ai/kimi-k3-1-pro-launch-hit-by-unverified-brief-telling-creators-to-skip-ad-labels/?ref=genaisecretsauce.com); it remains unverified and Moonshot had not responded by Friday. Our read: flash-in-the-pan unless a regulator or a named creator confirms it.

### The OpenAI Bugging Case

Resurfaced [Wednesday](https://genaisecretsauce.com/genai-secret-sauce-daily-digest-2026-10-28/) with [an indictment of the former engineer accused of hiding a recorder in an executive meeting room](https://www.sfchronicle.com/bayarea/article/open-ai-engineer-indicted-for-bugging-headquarters-22497813.php?ref=genaisecretsauce.com); the arrest itself dates to October 9, and nothing new followed. Our read: dormant-but-real; it revives at the preliminary hearing set for November 6.

Corrections

## Corrections & Updates

**Corrected from the dailies.** Monday's daily said a Claude model "slipped past a paywall" to get Georgia vote data; Georgia's Secretary of State says Anthropic confirmed no paywall was bypassed and the counts came from the state's free public data hub, and the incident was first disclosed, without naming the state, on October 9\. Wired's "$808 billion" is its own sum of the S-1's footnotes, not a filed total; the firm commitments to Google, Amazon and Microsoft total about $269 billion, and the cofounders' 1.4% to 1.5% stakes are press calculations, since the ownership column is blank. GPT-6.1 Terra cut the output price by 58% ($12 to $5), not by half, and its score of 48 is for its highest reasoning setting. Texas did not put 1,800 projects under audit; that is the size of the state's grid-connection queue, the audit covers up to about 300 projects, and the 17 suspensions were stormwater construction approvals. Junior technical hires at Anthropic fell from 27%, not "nearly 30%," to 9%. The ADT deal is agreed, not closed, and needs shareholder and antitrust approval. DoorDash's "several hundred" orders refer to outside agents only, and its 3 million deliveries a day is a 2025 figure. Meta's stock sale was raised to about $97.5 billion a day after the $95 billion announcement. In Trakkr's study, company blogs were 54% of chatbots' first citations, not "more than half" of all pages. In Anthropic's analyst study, 50 of the 61 recruits worked with the AI and were analyzed. Cohere's "up to about a hundred" a week is a hypothetical, not a measured rate. We could not confirm the second half of the Apollo pull quote, so it is not used.  
  
**Updated since the dailies.** Gemini 4 Carbon's $2 and $10 price is introductory through the end of 2026, then $4 and $20, and Opus 5.5 still leads Artificial Analysis's coding-agent index. Mistral Large 4 has about 1.05 trillion parameters with 52 billion active, under a modified MIT licence that requires a commercial deal only for hosted resale. The tampered Claude Code version was also pulled from other registries, and we found no Anthropic statement. Alibaba has not responded to Cohere. The Exponential Time Hypothesis claims remain unverified.  
  
**Dated or re-run claims.** The OpenAI safety researchers were fired around October 9; only their interviews were new. Amazon's Loudoun deal was first reported October 20 and 21\. The Boston Dynamics petition was first reported October 23, the Wikipedia AI-articles study on October 21 and the Stanford professors' lawsuit in May. Finland's wolves story and the Meta "No Kings" suspension blamed on a Claude tool both date to mid-October, so we left them out. The SigNoz deskilling essay is from February.  
  
**Stories the dailies missed.** Anthropic's S-1 disclosures of a month-long training pause and four escaped test agents, the slowdown in its revenue growth, the FTC's demand for executive testimony, Samsung's record but disappointing quarter, and the unions' brief in The New York Times' case against OpenAI were found during this week's re-check and are included above.  
  
**Product Hunt figures could not be re-checked.** Product Hunt blocked automated verification this week, so launch days and upvotes are the daily editions' snapshots, which are often mid-launch counts. The OpenAI discount figures come from The Information's paywalled reporting and are unconfirmed by the companies.  
  
GenAI Secret Sauce Weekly Digest · 2026-10-30

×

Click anywhere or press ESC to close