- That AI Thing
- Posts
- weekend ai reads for 2026-03-27
weekend ai reads for 2026-03-27
direct links are available on the web at https://thataithing.beehiiv.com/p/weekend-ai-reads-for-2026-03-27
📰 ABOVE THE FOLD: WORLD MODELS
What Is a World Model? / Nvidia Glossary (11 minute read)
World models are neural networks that understand the dynamics of the real world, including physics and spatial properties. They can use input data, including text, image, video, and movement, to generate videos that simulate realistic physical environments.
World Models: Computing the Uncomputable / Not Boring, Substack, archive (100 minute read)
Towards Efficient World Models / Moonlake, Twitter, archive (9 minute read)
an essay, not a tweet
Awesome World Models — A Curated List of Amazing Works in World Modeling, spanning applications in Embodied AI, Autonomous Driving, Natural Language Processing and Agents. / knightnemo, GitHub
Omma — 3D scenes, websites, games, apps. Describe anything and Omma builds it for you in seconds.
The future of work is world models — Why we need to build Starcraft for CEOs / Strange Loop Canon, Substack, archive (13 minute read)
📻 QUOTES OF THE WEEK
That’s always been part of the game: finding how to misuse the tools, and then by some misuse of the tool, you’ve spawned some amazing new thing that’s going to live on for decades.
Sleep isn’t rest. It’s maintenance.
👥 FOR EVERYONE
AI in “not always evil” shocker / Shardcore (8 minute read)
Creating subtitles with cerebral palsy is, as you would expect, not a simple task. For the normally voiced, adding subs is a relatively simple process using AI text-to-speech systems, perhaps with a little minor editing of the text when it gets it wrong. JT, however, does not have a ‘normal’ voice, and so editing the text resulting from transcription is non-trivial.
This is a classic example of where AI falls down – it’s excellent at working with the middle of the bell-curve, but rubbish at the edges.
Sorry, Mom. You’re Chatting With an A.I. Agent, Not Your Son. — Silicon Valley’s young coders are getting creative with this new technology. They also worry they’re not spending enough time with it. / New York Times (7 minute read)
Palantir’s billionaire CEO says only two kinds of people will succeed in the AI era: trade workers — ‘or you’re neurodivergent’ / Fortune (8 minute read)
For Karp, that cognitive difference can be an advantage in an AI-driven world—less because of the diagnosis itself and more because of the mindset it can foster. Success, he argued, will favor people who think differently and take risks, or in his words, be “more of an artist, look at things from a different direction, be able to build something unique.”
One-fifth of sales organizations within Fortune 500 companies are expected to actively recruit neurodivergent talent to improve business performance by 2027, according to a Gartner study.
related (1), What Young Workers Are Doing to AI-Proof Themselves — They’ve got their whole careers ahead of them, and they’re navigating a technology with a still-uncertain impact / Wall Street Journal, archive (19 minute read)
related (2), You thought the generalist was dead — in the ‘vibe work’ era, they’re more important than ever / Venture Beat (10 minute read)
ChatGPT Library Looks Like Cloud Storage. It’s Actually the End of the Chatbot Era. — OpenAI hired the guy who proved AI agents need a file system. Today they shipped the first piece of his vision. / Artificially Intimidating, Substack, archive (10 minute read)
These people used AI to help find their lost pets / Washington Post (8 minute read)
People send photos of their lost pets to a database, and AI compares the pets’ features — including facial structure, coat pattern and ear shape — to photos of stray pets that have been spotted elsewhere. Many of the stray pets have already been taken to shelters.
📚 FOUNDATIONS
AI Style Guides: How to Help AI Write Like You — A practical guide to making AI sound more human / Every (22 minute read)
“Claude” gets a byline, so make of that what you will
A free, open-source glossary of 51 AI terms and concepts for journalists, put together by Ryan Serpico, Jon Keegan and Libby Seline. … Please note this website is a work in progress, and the speakers plan to add more terms after NICAR 2026.
Make America AI-Ready / U.S. Department of Labor
A free, 1-week AI literacy course for every American worker — delivered entirely by text message. No laptop or internet needed. Just your phone.
we have not tried this; let us know what you learn
🚀 FOR LEADERS
Nvidia CEO Jensen Huang says he’d be “deeply alarmed” if a $500K developer spent less than $250K on AI tokens / The Decoder (4 minute read)
On the All-In podcast at Nvidia’s GTC conference, Huang laid out a “thought experiment:” If a developer or AI researcher earned $500,000 a year and only used $5,000 in tokens by year’s end, he would “go ape something else.” If their token budget wasn’t at least $250,000, he’d be “deeply alarmed.”
can’t wait for companies’ “total compensation” reports to include “tokens”
The Private AI Model Explosion / Forrester (7 minute read)
Companies must start building their private models now. Yes, AI will help organizations be more efficient and build and run processes faster. But this is a sideshow. The real AI game will be winning, serving, and retaining customers. And that will be the sweet spot of the private-model business model.
ChatGPT's First Advertisers Can’t Prove Ads Work / Winbuzzer (6 minute read)
OpenAI's first ChatGPT ad partners have reported click-through rates nearly 7x below Google benchmarks, broken measurement tools, and unreachable premium users.
🎓 FOR EDUCATORS
AI-Powered Simulations Offer Practice for Teachers in Training — Simulation platforms like BranchED are emerging as a modality for teacher training, using avatars and large language models to replicate student behavior and give teachers practice dealing with classroom situations. / Government Technology (6 minute read)
Canvas Unrolls AI Teaching Agent — The new AI agent aims to save faculty time on “low-value tasks,” but stops short of fully automating grading. But some experts worry that the rise of agentic AI could lead to a dead classroom, where computers teach other computers. / Inside Higher Ed (8 minute read)
Melania Trump pitches robots as potential educators for American schoolchildren / CBS News (5 minute read)
Escorted by a walking, talking humanoid system, first lady Melania Trump pitched the Fostering the Future Together Global Coalition Summit on AI robots as potential, “personalized” educators for America's children in their homes. Not only can AI share a depth and breadth of knowledge, she said, but it can also patiently help children develop “deep critical thinking and independent reasoning abilities,” achieve a “more well-rounded lifestyle” as they make time for other activities and become a “more complete person.”
Does ed-tech hurt learning? Here’s the evidence behind a viral anti-screen argument. / Chalkbeat (8 minute read)
For his part, Horvath argues the burden of proof should not be on him but on the purveyors of ed-tech. “We are much closer to proving that it’s harmful than we are to proving it’s helping,” he says.
📊 FOR TECHNOLOGISTS
When Shipping Software Becomes Too Easy — When the hardest part of building shifts, so does leadership / Mozilla blog (9 minute read)
How to convert Excel spreadsheets to Python models with Claude Code / Martin Alderson (6 minute read)
OpenAI adds open source tools to help developers build for teen safety / Tech Crunch (5 minute read)
These safety policies are designed as prompts, making them easily compatible with other models besides gpt-oss-safeguard, though they’re probably most effective within OpenAI’s own ecosystem.
🎉 FOR FUN
SLOW LLM — SLOW LLM is a browser extension that makes LLMs appear to run very slowly. It works with ChatGPT and Claude.
Proof — Proof is an online document editor built for agents and humans to collaborate. Fast, free, and no login required.
AI influencer awards season is upon us / The Verge (6 minute read)
To enter, you must develop your AI influencer on OpenArt’s platform and submit it at www.AIpersonality.ai. You’ll be asked for social media handles across TikTok, X, YouTube, and Instagram, as well as the story behind the character, your motivations for creating it, and details of any brand work.
The Nano Banana Effect: How Google’s Viral AI is Reshaping Architectural Visualization — Can Nano Banana Pro change the way we design buildings? Is it a tool for liberation, or the announcer of the end for the ArchViz profession? / Architizer Journal (8 minute read)
🧿 AI-ADJACENT
The Slow Death of the Power User / fireborn, mataroa (24 minute read)
The technology industry’s current consolidation into a small number of platform monopolies is only possible because the adversarial capacity to break platform lock-in has atrophied. There are still people doing it — the open-source community is still building, the security research community is still finding vulnerabilities, the right-to-repair movement is still fighting — but the cultural mass behind those efforts has collapsed. They’re fighting a rearguard action against an industry that has successfully convinced most of its users that platform control is a feature, not a bug.
⋄