weekend ai reads for 2025-08-22

šŸ“° ABOVE THE FOLD: DEVICES

 

šŸ“» QUOTES OF THE WEEK

You can say ā€œall are welcome,ā€ but if wolves and sheep are both welcome then you’re only going to get wolves. The the smart sheep will go somewhere else and the naive sheep will be eaten and processed. If you welcome Islamophobes and Muslims then you’ll get Islamophobes. If you welcome Klan members and people of color then you’ll get Klan members. If you welcome nativists and immigrants you’ll get nativists.

Refusing to choose is a choice. It’s a choice in favor of the people who prey on others and who refuse to acknowledge the humanity of those they hate.

Adam Bates (source, via Jason Kottke)

 

You’re not supposed to be scared of the president!

Andrew Callaghan (source)

 

šŸ‘„ FOR EVERYONE

LLMs are slot-machines / Cory Doctorow, Pluralistic (11 minute read)

In the case of LLM coding assistants, this manifests as an unconscious overestimation of how often the LLM saves you time. That’s because a coding program that produces a bug that you have to ā€œfutz withā€ for a while before it starts working is normal, and thus unmemorable, while a coding tool that turns a plain-language prompt into a working computer program is amazing, so it stands out in your memory.

If these moderate views of A.I. are right, then in the next few years A.I. tools will make steady but gradual advances. Many people will use A.I. on a regular but limited basis, whether to look up information or to speed up certain annoying tasks, such as summarizing a report or writing the rough draft of an event agenda. Certain fields, like programming and academia, will change dramatically. A minority of professions, such as voice acting and social-media copywriting, might essentially disappear.

How churches use data and AI as engines of surveillance — Spiritual care and surveillance are converging across the country, reshaping the theology of trust. / MIT Technology Review (29 minute read)

Google says a typical AI text prompt only uses 5 drops of water — experts say that’s misleading — Google shared a study of Gemini’s environmental impact, but it omits some key data. / The Verge (8 minute read)

What could have been. / Emilio Coppola (5 minute read)

Tech executives are robbing every investor blind. Promising the biggest breakthrough in technology while our current technology rots to the core. The operative systems we run, the browsers we use, our critical infrastructure gets consistently neglected to chase a promised wonderland of automation that never arrives.

  • we agree; but the hope of leaving a lot of technical debt behind when you move to a new system is part of the appeal of that new system

 

šŸ“š FOUNDATIONS

This is actually a really solid context engineering template. / mattpocockuk, XCancel (1 minute read)

That’s the thing with making anything. If you persevere, if you slog through the trough of disillusionment, ā€œa certain fascination gets through,ā€ she wrote. ā€œIt can be like those times you hear someone playing the piano beautifully or a sax wailing through jazz improvisations, and the sound ignites a longing in you, a desire to take up the difficulties and learn how to play that music.ā€

 

šŸš€ FOR LEADERS

95% of generative AI implementations in enterprise ā€˜have no measurable impact on P&L’, says MIT — flawed integration cited as why AI projects underperform ā€” The study shows that AI tools must adjust to the organization’s processes for it to work effectively. / Tom’s Hardware (6 minute read)

  • the report; ā€œ52 structured interviews across enterprise stakeholders, systematic analysis of 300+ public AI initiatives and announcements, and surveys with 153 leaders.ā€

C.E.O.s Want Their Companies to Adopt A.I. But Do They Get It Themselves? — Some are being nudged to learn how to use the nascent technology. Coming to the C-suite retreat: mandatory website-building exercises using A.I. tools. / New York Times, archive (11 minute read)

  • our experience: too often, ā€œnoā€

Responsible Enterprise AI in the Agentic Era [PDF] / Infosys (12 minute read)

 

šŸŽ“ FOR EDUCATORS

  • this seems like a good thing, on the whole

New findings from Michigan State University researchers suggest that some popular approaches for learning may not be as practical as parents think. Preschoolers who played more literacy games scored lower on reading assessments, while those engaged in print-focused activities saw the most significant gains.

Google just turned Gemini into a full-blown AI school system. Teachers can now assign AI experts to students. Students can auto-generate quizzes and visual explainers. And it’s all free in Google Workspace for Education. Here’s what just dropped / heyrimsha, XCancel (5 minute read)

An Ed Tech Insider Pleads for More Equitable Tools / The 74 Million (18 minute read)

I guess that’s where I’ve ended up: I believe in the promise of ed tech. I don’t think that the promise of ed tech and the free-market business model are compatible.

U-M Guidance for Students / University of Michigan (7 minute read)

Understand that your usage of GenAI-based tools can give you the means to better not just yourself, but also society as a whole, and there is an ethical responsibility towards doing so.

  • an ā€œethical responsibilityā€ to use A.i.???

  • archived here because they will undoubtedly modify this

 

šŸ“Š FOR TECHNOLOGISTS

Getting Good Results from Claude Code / Chris Dzombak (8 minute read)

The era of AI hacking has arrived — Hackers and cybersecurity companies have entered an AI arms race. / NBC News (7 minute read)

 

šŸŽ‰ FOR FUN

What are your best/most interesting use cases for AI right now? / Megan McArdle, XCancel (10 minute read)

Even so, the researchers were befuddled by the AI’s design. ā€œIf my students had tried to give me this thing, I would have said, ā€˜No, no, that’s ridiculous,ā€™ā€ Adhikari said. But the design was clearly effective.

Meet Dedrick Kane: The Villain Rapper Built by AI / Computerguru, Medium (6 minute read)

Google launches new AI-powered flight deals tool — It’s designed for flexible travelers whose number one goal is saving money on their next trip. / Google blog (3 minute read)

Ollie and other meal-planning AI apps aim to simplify running a household — Using technology to manage the mental load can be tempting. But it’s not a cure-all, experts say. / Washington Post (8 minute read)

Do LLMs Have Good Music Taste? / Tyler Cosgrove (6 minute read)

  • Anthropic (Sonnet 4) puts together a decent list but forgets how to count to 20

 

šŸŽ„ VIDEOS [BONUS SECTION]

Making cash off ā€˜AI slop’: The surreal business of AI video — A mad rush of creators is using AI video tools to flood the internet — and turn a profit — with videos that can seem remarkably real. / Washington Post (16 minute read)

ā€œHiggsfield is taking Product Placement to the next level. Just drop a product image onto any scene and get a fully animated, cinematic shot.ā€ / AngryTomtweets, XCancel (seven 9 second videos)

ā€œThis week’s AI community space was packed with talent Huge participation and intense creativity made it incredibly competitive!ā€ / MayorKingAI, XCancel (many short videos)

 

🧿 AI-ADJACENT

Curate your own newspaper with RSS / Citation Needed (16 minute read)

Stop padding your resume — Include these three things instead / Personal Math, Substack, archive (8 minute read)

 

ā‹„