AI's Sustainability Problem: Costs, Robots, and Behavior Prediction
November 2, 2026The gap between "AI is going to do everything" and "AI costs too much to keep the lights on" is closing fast, and today's stories are all about that tension.
Apple's $1,000 Home Robot Will Mostly Watch You Do Chores
Apple is reportedly building a home robot — a physical, wheeled device that roams around your house — with a price tag around $1,000 and a feature list that includes taking FaceTime calls, turning lights on and off, and apparently discouraging your cat from misbehaving. CEO John Ternus, who runs Apple's hardware division, is reportedly behind the project. A thousand dollars is roughly what a good laptop costs, for a device that does things your phone and a $30 smart plug already do. The honest read here is that Apple needs a new product category and this is their placeholder while they figure out what that actually looks like. The first generation will sell to enthusiasts and prove very little.
Grok Is Growing Too Fast for Elon Musk to Afford It
Musk just throttled — meaning cut back or put limits on — Grok Bot's usage, citing costs he described as "unsustainably high." This is the same product he recently called the fastest-growing in his life. That contradiction is the whole story: Musk built something people actually wanted, and now he can't afford to let them use it. This is worth keeping in mind when evaluating any AI tool's pricing. Those free or cheap tiers aren't sustainable — the companies are losing money on you now and betting they can charge you more later. If a tool becomes central to your workflow, the practical response is to keep a capable alternative ready and avoid building processes around features that only exist in a subsidized tier.
A Startup Says Its AI Can Predict What You'll Do Next. Obviously.
42AI emerged from stealth — meaning it was operating privately and just went public with its existence — with $62.5 million in funding and a claim that its AI can predict human behavior, even in situations the model has never seen before. The company calls this a "world model," which is an AI system trained to simulate how real environments and people behave, rather than just answer questions. The sci-fi framing in their launch materials isn't accidental. In practice, this kind of technology is probably most immediately useful in things like fraud detection, supply chain forecasting, or customer churn prediction — all cases where "what is this person likely to do next" has real commercial value. The sharper concern isn't the Minority Report scenario — it's the mundane one: systems like this will be used to make consequential decisions about people at scale, with no meaningful way for those people to know or contest it. That's already happening in credit and hiring. Adding better behavioral prediction makes it worse, not different.
Claude Put Rules Against Copying Games. It Won't Matter.
Anthropic, the company behind the Claude AI model, added new guidelines meant to stop Claude from helping people clone — copy the mechanics and style of — existing video games. Ars Technica's read, which seems correct, is that this is essentially a symbolic gesture: Claude is one of dozens of AI models available, and anyone determined to copy a game will simply use a different one. This situation mirrors what's happening across creative industries right now, from music to illustration to writing. One company tightening its rules creates a speed bump, not a roadblock, because the competitive AI market means there's almost always an alternative with fewer restrictions. For anyone in a creative field, the realistic conclusion is that AI-assisted imitation is going to remain easy and common regardless of what any individual model maker decides.
AI's Water Problem Is Becoming an M&A Problem
Omen AI acquired Aquatic Labs for $16 million to get access to liquid cooling technology for data centers — the massive buildings full of computers that run AI models. This is a small deal by tech standards, but it points to something that doesn't get enough attention: AI is an enormous physical infrastructure problem, not just a software one. Data centers require staggering amounts of water and electricity to keep the servers from overheating, and that constraint is real and growing. Companies are now buying each other specifically to solve the heat and water problem, which tells you something about how seriously the industry is taking it. If you work in sustainability, real estate, energy, or municipal planning, AI's infrastructure appetite is already affecting your world and will affect it more.
The most durable AI companies over the next three years won't be the ones that made the boldest claims — they'll be the ones that figured out how to run the thing without going broke.
Stop staring at the blank ChatGPT box
The Plain-English AI Prompt Pack: 100 copy-paste prompts for real work — emails, meetings, marketing, managing people. Works with any chatbot.
Get the Prompt Pack →Get the daily brief
Plain-English AI news for people with real jobs. Free, three minutes a day.