Hi! Welcome to Techtember, when the gadget announcements flow like pumpkin spice and the AI agents do the back-to-school shopping. Yes, I handed my son’s school supply list to the big platforms—Claude, ChatGPT and Gemini—and told them to shop. The results are below.
Then: Tim Cook is out at Apple, John Ternus is in! Dyson wants to put cameras in your mouth, and Sonos is back. Plus, Apple Polishing Cloth jokes, because some traditions are sacred.
Hey! Thanks for being a free subscriber to New Things!
It’s still summer—technically—so our end-of-summer sale is still on.
That’s less than one $9 Apple Polishing Cloth a month.
AI labs have all sorts of fancy benchmarks for testing their agents, but they don’t have The School Supply Shopping Benchmark™. It’s grueling. It involves online shopping carts and finding the perfect handheld pencil sharpener and packs of pre-sharpened pencils requested by a fourth-grade teacher.
Last year, during my AI Year Takeover for my book, I AM NOT A ROBOT, I used Perplexity’s Comet browser and ChatGPT’s Atlas (RIP), the cutting edge of agentic browsers at the time, to do my son’s back-to-school shopping.
A year later, Claude, ChatGPT, Gemini and others now all have agents that can take action across the web, and specifically in their own native web browsers. (Claude added this last week.) So naturally, I’ve turned this into an annual benchmark. Forget Humanity’s Last Exam. This is the Buy the Right Glue Sticks Exam.
The results were surprisingly clear: These anonymous tasks on the web have gotten so much better in just the past 12 months. One agent absolutely crushed the task, another wasn’t far behind and the third still seems to be looking for the fine-tip black Sharpies.
Fun fact: Sharks have a continuous tooth replacement system—as one row of teeth gets worn down, others push forward. Some sharks go through up to 35,000 teeth in their lifetime. Selling sharks $500 toothbrushes would be a terrible business.
This newsletter was written and curated by Joanna Stern and Adele Lowitz.
|
