In partnership with
Smart starts here.
You don't have to read everything — just the right thing. 1440's daily newsletter distills the day's biggest stories from 100+ sources into one quick, 5-minute read. It's the fastest way to stay sharp, sound informed, and actually understand what's happening in the world. Join 4.5 million readers who start their day the smart way.
Beginners in AI
Good morning and thank you for joining us again!
Welcome to this daily edition of Beginners in AI, where we explore the latest trends, tools, and news in the world of AI and the tech that surrounds it. Like all editions, this is human curated and edited, and published with the intention of making AI news and technology more accessible to everyone.
THE FRONT PAGE
AI Trainers Ask to Film Kids at School, a Growing Pattern
TLDR: A University of Washington study wanted preschool teachers to wear cameras and film 4-year-olds all day to train AI, and it's part of a much bigger pattern where everywhere is becoming a recording studio for AI training data.
The Story:
Parents at a University of Washington preschool got a strange form in their kid's backpack. It said the lead teacher might wear a small camera that captures her view of the classroom, plus a fixed camera in the room. The footage would be used to train AI models. If parents didn't want their 4-year-old recorded, they had to manually opt out. Many of the families at the school speak languages other than English, and the form wasn't translated. After 404 Media's investigation and parent backlash, the study was scrapped. This follows last month's protest that killed a planned AI-powered high school in New York City.
Its Significance:
This isn't just about schools. Meta started recording its own employees' keystrokes, mouse clicks, and screen activity on Google, LinkedIn, Wikipedia, and hundreds of other sites, all to train AI agents to do office work. Meta Ray-Ban smart glasses now store your voice recordings by default with no opt-out, and human contractors in Kenya have reviewed footage that includes people using the bathroom and undressing. AI models need huge amounts of real-world data, and the cheapest way to get it is to record people doing normal things. Your face, your voice, your kid's playtime, your typing habits, your bathroom trip caught by someone else's glasses, it's all training data now. The question for each person is whether they know they're being recorded, whether they can say no, and what happens to the footage after.
QUICK TAKES
The story: Penn scientists made a hybrid particle called an exciton-polariton that combines light's speed with matter's ability to interact. They used it to do all-light switching at about 4 quadrillionths of a joule, less energy than briefly powering a tiny LED. Today's photonic AI chips have to keep converting light back to electrons for decision-making steps, which slows them down. This skips that step.
Your takeaway: Data centers burn massive amounts of electricity to run AI. If this scales, chips could process camera input directly as light, no conversion needed. It's early lab work, but it points to a future where AI hardware uses far less power than it does today.
The story: Chaac Pizza Northeast, which runs 111 Pizza Hut stores, says the chain's mandatory AI delivery tool called Dragontail tanked its business. Before the AI, more than 90% of deliveries arrived in under 30 minutes. After, sales growth in New York City swung from +10.19% to -9.78%, because DoorDash drivers could see when pizzas would come out of the oven and waited up to 15 minutes for multiple orders to stack up.
Your takeaway: AI gets sold as a no-brainer fix, but if it's pushed onto people who didn't ask for it, the side effects can wreck the thing it was supposed to help. Lukewarm pizza and a $100 million lawsuit is not the win Pizza Hut was looking for.
The story: A California jury took less than two hours to throw out Elon Musk's lawsuit against Sam Altman and OpenAI. The reason: Musk waited too long to file. His claims of breach of charitable trust and unjust enrichment had a three-year statute of limitations. The court never ruled on whether OpenAI actually did anything wrong.
Your takeaway: The case got dismissed on timing, not merit. Musk says he'll appeal. For now, OpenAI walks away from a fight that could have forced it to give back up to $150 billion.
TOOLS ON OUR RADAR
💾 Duplicati Freemium: A highly reliable automatic backup system that securely encrypts your personal files and safely stores them on your preferred cloud storage provider without requiring complicated setup menus.
📖 Foliate Free and Open Source: A beautifully designed electronic book reader for your computer that features customizable text layouts reading progress tracking and built in dictionary support.
🚑 Rescuezilla Free and Open Source: An incredibly user friendly graphical utility that lets you easily backup and restore your entire computer hard drive just like an emergency digital life raft.
🌐 Ghost Free and Open Source: A gorgeous publishing platform that empowers writers and creators to easily launch their own personal blogs and newsletters without relying on restrictive corporate publishing networks.
TRENDING
Amazon's Alexa Can Now Generate Custom AI Podcasts on Any Topic — Ask Alexa+ about any subject and it'll build a podcast-style episode with two AI-generated hosts chatting back and forth. Episodes pull from Reuters, AP, The Washington Post, and 200 local newspapers. Free for Prime members.
Anthropic Buys Stainless for Over $300 Million, Cutting Off OpenAI and Google — Stainless makes the software that turns API specs into developer tools. It powered SDKs for OpenAI, Google, Cloudflare, and Anthropic itself. Now Anthropic is shutting down hosted Stainless products for everyone else, forcing rivals to rebuild their developer tools from scratch.
Cursor's Composer 2.5 Matches Claude Opus 4.7 at One-Tenth the Price — Cursor's new coding model scored 79.8% on SWE-Bench Multilingual, close to frontier models from Anthropic and OpenAI, but at $0.50 per million input tokens and $2.50 per million output tokens. Cursor is also working with SpaceXAI on a bigger model trained with 10 times the compute.
Oppo Open-Sources an Android AI Agent That Runs on Your Phone — Oppo's X-OmniClaw uses the phone's camera, screen, and microphone to do tasks across apps. Point it at a bottle and ask the price, it opens Taobao and scrolls. Show it a math problem, it solves the next ones too. Most of the work happens on the phone, not in the cloud.
Vitalik Buterin Says AI Can Make Crypto Safer, Not Just Riskier — Ethereum's co-founder argues that AI-assisted formal verification, where math proves code does exactly what it's supposed to do, could shore up Ethereum, smart contracts, and post-quantum cryptography against AI-powered hackers. He warns it's not a complete fix, but says it beats the alternative.
The 'Unintended Consequences' of AI Deciding Your Health Insurance Claims — A former United Healthcare appeals worker told WUSF that AI denials are already widespread, Centers for Medicare and Medicaid Services are rolling out AI at scale, and there's no holding the dam back. Florida lawmakers have proposed bills to require humans, not algorithms, to make the final call on denied claims.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🎯 State any prediction. Get a calibrated probability, base rates, and exactly what would change your mind.
Build a single-file HTML app with vanilla HTML/CSS/JS. Forecasting Coach — calibrate any prediction against base rates. Persist to localStorage key 'forecasting_coach_v1'.
Aesthetic: dark navy (#0a0e14), faint cyan grid overlay, vignette. Spectral serif for body, Inter for UI, Space Mono for labels. Teal (#78c8d0) accent.
Form: prediction textarea, confidence slider (1-99 with live % readout), reasoning textarea, horizon dropdown, domain dropdown.
System instructions to the model: act as a forecasting coach trained on Tetlock's superforecasting research. Be honest — don't flatter the user's estimate. Return raw JSON: question_clean, your_probability, calibrated_probability, calibrated_range (e.g. "35-55%"), gap_note (honest comparison), reasoning (2-3 paragraphs with base rates and asymmetries), base_rates array (3 items: rate + claim), update_higher array, update_lower array, trap (the most likely cognitive bias affecting their reasoning).
Render: side-by-side user vs calibrated probability cards with big numbers (amber for user, teal for coach), gap note callout, reasoning paragraphs, base rates list with rate + claim, two-column update higher/lower (green/red), and an amber-left-border trap warning. Save to localStorage with calibrated probability in archive.What this does: Type a prediction, set your confidence with a slider, add your reasoning. Get back a calibrated probability with a range, an honest comparison to your estimate, the base rates to anchor on, specific evidence that would push it higher or lower, and the cognitive trap most likely affecting your thinking. Saves to localStorage.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Generate full custom podcast episodes with two AI hosts conversing back and forth, pulling from licensed news sources (Alexa Podcasts).
❌ Still Can't: Reliably figure out which fast-food deliveries to prioritize without making service worse (Pizza Hut Dragontail).
✅ AI Can Now: Operate as an on-device Android agent that uses your real camera, screen, and microphone to do tasks across apps without sending data to the cloud (Oppo X-OmniClaw).
❌ Still Can't: Get used in classroom recording studies, employee keystroke tracking, or smart glasses data collection without major public backlash about privacy and consent.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
A bleak post-apocalyptic film set after the collapse of synthetic biology, where a 13-year-old girl scavenges a swamp using small AI drones and bioengineered organisms to survive. Made by Lithuanian and French directors for almost no money, with practical effects that put most Hollywood productions to shame. One of the best films ever made about a world where AI and synthetic biology have run together and broken down. Released to almost zero fanfare in 2022.
Thank you for reading. We’re all beginners in something. With that in mind, your questions and feedback are always welcome and I read every single email!
-James
By the way, this is the link if you liked the content and want to share with a friend.
Some * designated product links may be affiliate or referral links. As an Amazon Associate, I earn from qualifying purchases. This helps support the newsletter at no extra cost to you and Amazon makes a tiny hair less.
