weekend ai reads for 2026-08-14

August 14, 2026

📰 ABOVE THE FOLD: PROVENANCE

Nobody Asks Who Made the Bolt — On warehouses, language models, and the wrong question about who made what. / Future Objects, Substack, archive (15 minute read)

  • separately:

… nowadays you have to pay more to have less screen time. And I suspect there'll be something similar with AI in school, and kids who bring less capability to the classroom will get less out of the learning when AI is injected. Perhaps the same old story, but at a velocity no one is prepared for.

  • users made realistic fake satellite images of wars, disasters, and refugee camps; the feature worked exactly as intended

Judge denies request by Elon Musk’s xAI to block MN nudification ban — The ruling clears the way for the first-in-the-nation law to go into effect on Saturday. / NBC News (8 minute read)

 

📻 QUOTES OF THE WEEK

What is the risk of what I’ve never seen before?

Dr Moritz Hanke, Johns Hopkins Center for Health Security (source)

 

If you measure bad behavior, and then train a system not to manifest what you’ve measured, you train it not just to do less of the bad thing but also to evade measurement of it.

Joshua Rothman (source)

 

👥 FOR EVERYONE

The Rise of the 1 am Job Interview — Since there’s no human on the other end, candidates are scheduling them whenever—even deep into the night. / Wired (5 minute read)

Who needs consultants in the age of AI? / The Financial Times, archive (7 minute read)

Doing industry research and presenting it clearly; isolating broad patterns among the strategies of multiple companies; and recommending options for action are all things the strategics have charged for. AI is brilliant at all of them. If it can go into a slide deck, it’s almost certainly turning into a commodity.

Vivian wears her Rokid glasses to navigate on her scooter, and to quietly photograph price tags so she can check if it’s cheaper online.

How kids feel about AI, in their own words — Does an AI-filled future seem like promise or peril to today’s teens and tweens? We went right to the source. / MIT Technology Review (19 minute read)

AI Pricing: A Reality Check on Effort, Usage, and Outcomes — As software companies layer new models alongside seat licensing, it’s important to understand the opportunities and risks. / Bain & Company (8 minute read)

 

📚 FOUNDATIONS

You can use the new AI to feed the old AI good data, and then use the old AI to answer questions for the new AI so the new AI doesn’t make things up all the time. And it works! By which I mean it breaks all the time until you iteratively improve your taxonomy and then it actually does something like what you expected.

Should You Self-Host Inference? — The economics of running your own models versus hosted APIs. Self-hosting becomes cost-effective only past roughly two million tokens a day. / The AI Engineer, Substack, archive (16 minute read)

A guide to AI workflow automation / Box blog (10 minute read)

Let’s explore how AI in workflow automation supports more informed decisions and helps reduce time spent on repetitive tasks, leading to real gains across your operations.

 

🚀 FOR LEADERS

6 critical success factors for AI transformation / Konecta blog (8 minute read)

6. Ensure data readiness, scalable infrastructure and seamless integration

“The only way for my data advantage to become my AI advantage is if I can customize the models with my data, and I cannot customize closed source models,” Natarajan said at the time. 

The Build vs Buy Head Fake — On the “Saaspocalypse,” the panic that ensued following Anthropic’s Claude Cowork. / Field Ventures Persepctives, Substack, archive (20 minute read)

Target names its first chief AI officer in a bid to remake the shopping experience — Former Lowe’s executive Chandhu Nair will lead the retailer’s AI push, spanning everything from employee tools and inventory management to how customers shop online. / Fast Company (6 minute read)

The CAIO position has exploded alongside the broader AI boom. Just two years ago, the CAIO role barely existed. Last year, 26% of organizations had one. Today that number stands at 76%.

 

🎓 FOR EDUCATORS

Universities would prefer not to. — AI and the degree monopoly. / Hollis Robbins, Anecdotal Value, Substack, archive (20 minute read)

Solvely — AI Study Tools built for students

How Behavioral Bias Broke Ed Tech — and How to Fix It / The Learning Agency (8 minute read)

Mathematics in the Age of AI [PDF] / Terence Tao, GitHub

  • presentation to the International Congress of Mathematicians 2026, and not anti-A.i., but a thoughtful perspective on how to to bring AI into mathematics (he was awarded the Fields Medal in 2006)

  • related, What sort of maths are LLMs good at? / Gowers’s Weblog (23 minute read)

Let us consider two things that we can be confident that LLMs are good at. One of them is knowing a lot of mathematics: if a problem can be solved by means of a relatively standard argument, it is highly likely that an LLM will be able to find and use that argument. The other is the ability that an LLM has simply by virtue of being a computer: it can work at huge speed (compared with humans at least) and can therefore afford to make a large number of unsuccessful attempts at a problem before it finds a solution.

 

📊 FOR TECHNOLOGISTS

Some thoughts about Anthropic’s new cryptanalysis results / Matthew Green, A Few Thoughts on Cryptographic Engineering (13 minute read)

I don’t know. If you’re under the impression that these models are “glorified autocomplete” or that progress is slowing down, I need to urge you: stop thinking that. The models are very intelligent and capable, they are getting better at a fast clip. … If there’s a ceiling out there, I don’t yet see evidence of it. The people who think models are dumb are mostly using Google’s free AI search results, and not interacting with the high-end stuff (which only costs $20, so it’s not out of reach.) And they’re mostly not working in new areas.

On the other hand: if you think that models are super-intelligent or that AGI is already here, you should also stop thinking that. Working with these tools is like swimming in a pond where the ground drops off sharply. One minute you’re wading comfortably and there’s support under your feet. Then suddenly you cross a specific line, and you’re back to swimming on your own. This analogy is my best way to explain what it feels like when the model goes from helpful to clueless. Right now it’s easy for a human being to find that line if you’re doing advanced research, so you know where the intelligence drops off. But the line is moving. You can feel it slowly drifting outwards under your feet.

The future of the website / Joost blog (7 minute read)

openworker / andrewyng, GitHub (7 minute read)

AI that gets your everyday tasks done. OpenWorker is an open-source AI coworker that lives on your desktop and delivers finished work, not just chat: a polished document, a Slack reply with the numbers, an updated calendar, a triaged inbox.

Cost of a Data Breach Report 2026 — The AI Tipping Point [PDF] / IBM (17 minute read)

 

🎉 FOR FUN

AI Economics for Dummies / Andrew Singleton, McSweeney’s (4 minute read)

Benjamin owns a farm. He employs 100 workers plowing his fields. His total payroll is $10 million/year. One day, he buys a mule, which provides the worker who uses it with a modest 10 percent productivity gain. Benjamin fires 99 of his workers and purchases 99 mules, expecting a 1,000 percent productivity gain. The driverless mules cause plow damage to his property in excess of $50 million.

AI Group Call — Live group voice calls with generative AI

  • we have not tried this but this seems poised for unintentional humour

Can A.I. Get You Where You Want to Go for Less? / New York Times (8 minute read)

“The consumer effectively gains the opportunity to monitor the market,” Mareike Mohlmann, the chair of A.I. and business at Bentley University in Waltham, Mass., wrote in an email. “It is a small but meaningful shift from A.I. acting ‘on’ consumers to A.I. acting ‘for’ consumers.”

Please Use AI / Shawn Smucker, The Courage to Live It, Substack, archive (5 minute read)

And be sure to use AI when your next child
gets married, so that you can write them
the perfect toast or poem or speech or song
because no one wants to hear your
words, the actual poorly written words
of a parent (you) who changed
hundreds of diapers for said child or fed
them in the middle of the
night from your actual body.

 

🧿 AI-ADJACENT

What’s the Purpose of Information Architecture? — Fifty years after its formal introduction, the discipline is still widely misunderstood. Here’s a corrective. / Jorge Arango (4 minute read)

 

⋄