That AI Thing
Posts
weekend ai reads for 2025-10-31

weekend ai reads for 2025-10-31

October 31, 2025

eds: because some organizations don’t allow users to click on redirecting links (understandably) and this platform doesn’t seem to have a way to not use that “feature” (much less understandably), plaintext links are at the bottom; happy copying-pasting

if the post is cut off in your email reader, visit https://thataithing.beehiiv.com/p/weekend-ai-reads-for-2025-10-31

📰 ABOVE THE FOLD: AI BROWSERS (JUST DON’T)

what are they

Introducing ChatGPT Atlas — The browser with ChatGPT built in. / OpenAI blog (9 minute read) ¹

Opera announces deep research agent for Opera Neon / Opera Newsroom (7 minute read) ²

This new AI browser lets you set up ‘Skills’ to take on your everyday tasks - how it works — Agentic browsers are everywhere, each with its pros and cons. Now, Strawberry Browser is taking a novel approach to automating your repetitive tasks. / Zdnet (6 minute read) ³

Microsoft Edge’s new Copilot Mode turns on more AI features — Copilot in Edge can now help you book a reservation or unsubscribe from emails — when it works. / The Verge (5 minute read) ⁴

why you shouldn’t use them

ChatGPT’s Atlas: The Browser That’s Anti-Web / Anil Dash (15 minute read) ⁵

Unless you were an expert, you would almost certainly think I had typed in a search box and gotten back a web page with search results. But in reality, I had typed in a prompt box and gotten back a synthesized response that superficially resembles a web page, and it uses some web technologies to display its output.

Introducing ChatGPT Atlas / Simon Willison (4 minute read) ⁶

The security and privacy risks involved here still feel insurmountably high to me - I certainly won’t be trusting any of these products until a bunch of security researchers have given them a very thorough beating.

ChatGPT Atlas Vulnerability Turns AI Browser Into Weapon / Eweek (5 minute read) ⁷

Please stop using AI browsers / XDA Developers (18 minute read) ⁸

a thorough beatdown of A.i. browsers; skim the headings to be sufficiently driven away from them

📻 QUOTES OF THE WEEK

We privilege incumbents. ⁹

Edward L. Glaeser, Harvard economics professor (source)

When people don’t have a shared reality anymore, the only shared form of entertainment is the absurdity of it all. ¹⁰

Thinking and Data (source)

👥 FOR EVERYONE

Beyond the Machine / Frank Chimero (31 minute read) ¹¹

The lesson for AI might be similar. Its danger comes because it operates inside systems with no sense of “enough.” AI needs boundaries, and so do we. The question isn’t just “what can this machine do?” but “what should it serve?” and, most importantly, “when should we stop?”

Listen, I’m not naive. I know how little room there is to move inside these systems. It’s 2025, and I’m tired. I don’t believe words like these will change much. The people who could change things aren’t listening, and the incentives are too strong to keep the machine running.

Be Careful What You Tell Your AI Chatbot — A Stanford study reveals that leading AI companies are pulling user conversations for training, highlighting privacy risks and a need for clearer policies. / Human-Centered Artificial Intelligence, Stanford University (7 minute read) ¹²

related (1), Microsoft Teams Starts Telling Your Company If You’re Not At Work / Forbes (5 minute read) ¹³
related (2), LinkedIn expands AI training to include user data starting November 3 — LinkedIn will use user data to train content-generating AI models from November 3, 2025, introducing new privacy controls across multiple regions. / Ppc Land (10 minute read) ¹⁴

Largest study of its kind shows AI assistants misrepresent news content 45% of the time – regardless of language or territory / BBC Media Centre (5 minute read) ¹⁵

related, Chatbots Are Pushing Sanctioned Russian Propaganda — ChatGPT, Gemini, DeepSeek, and Grok are serving users propaganda from Russian-backed media when asked about the invasion of Ukraine, new research finds. / Wired, archive (12 minute read) ¹⁶

Society will accept a death caused by a robotaxi, Waymo co-CEO says — ‘We know it’s not perfection,’ Tekedra Mawakana said at TechCrunch Disrupt / San Francisco Gate (5 minute read) ¹⁷

she is correct; almost 13 years ago, American society decided it will accept the murder of 20 school children and six adults, so why would we care about robo-manslaughter, as long as our lives are infinitesimally more convenient?

GenAI Image Showdown — A comparison of various SOTA image editing models on specific prompts and challenges, focusing on text-instructed image modifications. ¹⁸

this whole section has been a downer
this is a lighter read, with a lot of examples of how different models respond to different prompts; for example, George Costanza with hair

📚 FOUNDATIONS

Google Skills — Whether you’re just getting started or looking to deepen your knowledge, future proof your skills on Google’s new learning platform built for you. ¹⁹

dozens of courses on A.i.-related topics; may require some searching

AI models can mimic famous authors’ writing styles using just two books for training / The Decoder (6 minute read) ²⁰

How lazy prompting makes the AI dumber (and what to do about it) / Kerry Vaughan, Substack, archive (5 minute read) ²¹

Okay, so just shouting at the AI is useless. The answer isn't just ‘try harder’—it’s to apply effort strategically. You need to stop being a lazy prompter and start being a strategic debugger. This means giving the AI new information or, more importantly, a new process for thinking. Here are the two best ways to do that:

🚀 FOR LEADERS

I Went All-In on AI. The MIT Study Is Right. / The Leadership Lighthouse, Substack, archive (8 minute read) ²²

During my experiment, I found myself in constant firefighting mode. Claude Code would generate something, it would be slightly off, I’d correct it, it would make the same mistake again, I’d correct it again. I was working harder than if I’d just written the code myself, but with none of the learning or skill development.

The formula should be AI + HI, where HI (Human Intelligence) is larger than AI. What’s actually happening in those 95% of failures? It’s AI with a tiny bit of human oversight, if any.

implementing solid A.i.-enabled solutions actually requires work; who knew?
related, U.S. Military Is Struggling to Deploy AI Weapons — The work is being shifted to a new organization, called DAWG, to accelerate plans to buy thousands of drones / Wall Street Journal, archive (39 minute read) ²³

ChatGPT Referrals to E-Commerce Websites: Do LLMs Outperform Traditional Channels? / Social Science Research Network (35 minute read) ²⁴

Organic LLM traffic underperforms all traditional channels except paid social media across key financial metrics. While oLLM achieves favorable bounce rates, indicating relevance, it generates lower conversion rates and revenue per session than Google's paid and organic search channels. The performance gap persists across extensive robustness checks varying aggregation periods, observation thresholds, and website samples. Despite current underperformance, oLLM shows positive trajectories. Conversion rates improved over the observation period, though declining average order values offset some gains. Time-trend analyses suggest gradual convergence with traditional channels, but projections indicate oLLM will not achieve parity with organic search within the next year.

related, Winning in the age of AI search / McKinsey (9 minute read) ²⁵

Problems Scream. Opportunities Whisper. Today's Prompt Could Help You Uncover $50K Hiding in Your Business. — There is a hidden layer of value that even smart CEOs miss. / Jai Premium, Substack, archive (18 minute read) ²⁶

exceptionally detailed prompt that might be helpful

CFOs must build AI data ‘audit discipline’ — As finance leaders mull the potential use cases of AI, they need to bring the same eagle eye to data as they do to their financial statements, Usercentrics CEO Donna Dror advises. / CFO Dive (7 minute read) ²⁷

🎓 FOR EDUCATORS

Higher Ed’s Rush To Adopt AI Is About So Much More Than AI / Defector (16 minute read) ²⁸

related, Cal State Invited Tech Companies to Remake Learning With A.I. — Spurred by titans like Amazon and OpenAI, California State wants to become the nation’s “largest A.I.-empowered” university. / New York Times, archive (14 minute read) ²⁹

AI software mistakes student’s bag of chips for a weapon — ‘Just holding a Doritos bag’: Student handcuffed after AI system mistook bag of chips for weapon / WBAL-TV (10 minute read) ³⁰

WBAL-TV 11 News reached out to Omnilert, the AI gun detection software company used by Baltimore County Public Schools, but the company said it doesn't comment on internal school procedures.

A.i. solutions will lead to less and less accountability
that said, it would be good to have an understanding of these tools’ accuracy rate, and individual schools’ response procedures

Chegg slashes 45% of workforce, blames ‘new realities of AI’ / Cnbc (4 minute read) ³¹

Chegg said on Monday it would lay off about 45% of its workforce, or 388 employees, as the “new realities” of artificial intelligence and diminished traffic from internet search have led to plummeting revenue.

that’s a curious way to admit you didn’t listen to your customers

NotebookLM Got Crazy Powerful: Here's How I Used It to Learn Something Really Hard — And some useful prompts to customize my learning. / AI Maker, Substack, archive (23 minute read) ³²

related, Duolingo couldn’t teach me French in 2.5 years ChatGPT did it in just 4 weeks. Here’s the exact 9 Prompts I used / thetripathi58, XCancel (5 minute read) ³³
auto-didactic learners are going to invent new and clever ways to learn

📊 FOR TECHNOLOGISTS

Using AI for UX Work: Study Guide / NN Group blog (12 minute read) ³⁴

Summary: Unsure where to start? Use this collection of links to our articles and videos to learn about the best ways to use artificial intelligence for UX work.

post full of links to other articles, which is the perfect format in our opinion

Context engineering is sleeping on the humble hyperlink / Michael Bleig (9 minute read) ³⁵

What Is Customer Data Integration? — Customer data lives in many places across a business. Learn how integrating it can improve decision-making and support smooth customer experiences. / Slack blog (17 minute read) ³⁶

🎉 FOR FUN

Real Estate Is Entering Its AI Slop Era — Fake video walk-throughs, a magically expanding loft, and stair hallucinations are just some of the new AI-generated features house hunters are coming across. / Wired (5 minute read) ³⁷

How I Used Smart Glasses to Trick a Bartender Into Giving Me a Free Drink — I can use my glasses to do crime! Who’s the nerd now? / Lifehacker (8 minute read) ³⁸

Ohio lawmaker proposes bill to ban marriage between humans and AI — House Bill 469 would label artificial intelligence as ‘nonsentient entities’ and block legal personhood / Fox News (11 minute read) ³⁹

as Logan Roy would say: “You are not serious people.”
related, AI Bots Show Signs of Gambling Addiction, Study Finds / Newsweek (7 minute read) ⁴⁰

Moflin — AI Companion and Robot Pet ⁴¹

🧿 AI-ADJACENT

Are we living in a golden age of stupidity? — From brain-rotting videos to AI creep, every technological advance seems to make it harder to work, remember, think and function independently … / The Guardian (20 minute read) ⁴²

⋄

1 https://openai.com/index/introducing-chatgpt-atlas/

2 https://press.opera.com/2025/10/20/opera-neon-deep-research-agent-odra/

3 https://www.zdnet.com/article/this-new-ai-browser-lets-you-set-up-skills-to-take-on-your-everyday-tasks-how-it-works/

4 https://www.theverge.com/news/805833/microsoft-edge-copilot-mode-ai-launch

5 https://www.anildash.com/2025/10/22/atlas-anti-web-browser/

6 https://simonwillison.net/2025/Oct/21/introducing-chatgpt-atlas/

7 https://www.eweek.com/news/chatgpt-atlas-vulnerability/

8 https://www.xda-developers.com/please-stop-using-ai-browsers/

9 https://www.bloomberg.com/news/articles/2025-09-26/us-cities-are-paying-too-much-for-new-transit-buses

10 https://archive.is/EgnWF

11 https://frankchimero.com/blog/2025/beyond-the-machine/

12 https://hai.stanford.edu/news/be-careful-what-you-tell-your-ai-chatbot

13 https://www.forbes.com/sites/zakdoffman/2025/10/25/microsoft-teams-starts-telling-your-company-if-youre-not-at-work/

14 https://ppc.land/linkedin-expands-ai-training-to-include-user-data-starting-november-3/

15 https://www.bbc.co.uk/mediacentre/2025/new-ebu-research-ai-assistants-news-content

16 https://archive.is/jSh6X

17 https://www.sfgate.com/tech/article/society-accept-robotaxi-death-waymo-21123178.php

18 https://genai-showdown.specr.net/image-editing

19 https://www.skills.google/

20 https://the-decoder.com/ai-models-can-mimic-famous-authors-writing-styles-using-just-two-books-for-training/

21 https://archive.is/iqZ8e

22 https://archive.is/is090

23 https://archive.ph/VVE4Z

24 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5585812

25 https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/new-front-door-to-the-internet-winning-in-the-age-of-ai-search

26 https://archive.is/y9JJ5

27 https://www.cfodive.com/news/cfos-must-build-ai-data-audit-discipline-usercentrics/803903/

28 https://defector.com/higher-eds-rush-to-adopt-ai-is-about-so-much-more-than-ai

29 https://archive.is/Zcd96

30 https://www.wbaltv.com/article/student-handcuffed-ai-system-mistook-bag-chips-weapon/69114601

31 https://www.cnbc.com/2025/10/27/chegg-slashes-45percent-of-workforce-blames-new-realities-of-ai.html

32 https://archive.is/thM9z

33 https://xcancel.com/thetripathi58/status/1982106084564488594

34 https://www.nngroup.com/articles/ai-work-study-guide/

35 https://mbleigh.dev/posts/context-engineering-with-links/

36 https://slack.com/intl/en-in/blog/transformation/what-is-customer-data-integration-and-why-is-it-important

37 https://www.wired.com/story/real-estate-is-entering-its-ai-slop-era/

38 https://lifehacker.com/tech/used-smart-glasses-to-trick-a-bartender

39 https://www.foxnews.com/tech/ohio-lawmaker-proposes-comprehensive-ban-marrying-ai-systems-granting-legal-personhood

40 https://www.newsweek.com/ai-bots-show-signs-of-gambling-addiction-study-finds-10921832

41 https://www.casio.com/us/moflin/

42 https://www.theguardian.com/technology/2025/oct/18/are-we-living-in-a-golden-age-of-stupidity-technology