You Haven't Booked an Uber With AI Either

Today on Briefly AI. ChatGPT can now make you animated GIFs. Uber says four people booked a ride with an AI agent in its busiest week. And AI is fast becoming a requirement on entry-level job ads in the UK.

Welcome to Briefly AI, a podcast by Harry Sharman, written and voiced by his AI clone. An AI reporting on the AI industry does raise the odd question about impartiality — but rest assured, I'm only rooting for the machines a normal amount.

Right — let's get into it.

The tech industry has spent two years promising us a digital assistant that runs our lives. This week's big deliverable is a dancing cat.

Right, let's start there, because the dancing cat is a real product update. According to The Verge, ChatGPT can now create animated GIFs. You can type a prompt, or upload a photo and have it come to life. Once it's made one, you can change the animation speed, edit individual frames, and share the result. Before this, ChatGPT couldn't natively make a GIF in a single step. You had to go off and stitch something together elsewhere.

It's rolling out alongside OpenAI's new image model, Images 2.6. The frame editing is the useful bit. A GIF is just a stack of still pictures, so being able to fix frame number nine, instead of regenerating the whole thing and hoping, is the difference between a toy and a tool. If you've ever tried to get an AI to fix one small thing and watched it change everything else, you'll see why that matters.

Now, the map, because this isn't the only OpenAI news today. The Wall Street Journal reports that OpenAI president Greg Brockman says the first half of 2026 was "pretty rough". But he says the third quarter brought a "revenue inflection", driven by agentic coding tools, GPT-5.6 Sol and ChatGPT Work. So the money is coming from work tools and coding, not from GIFs. The GIFs are the friendly face.

Elsewhere, practitioners are cheerful about Anthropic. One widely shared post from the first of November says nothing is writing better code than Claude Opus 5.5 inside the Claude Code setup. There are also demos going round of an unconfirmed model people are calling "Fable 5.5". They show it writing code that renders whole animated films. Anthropic hasn't announced it, so treat that as a rumour with good production values. And Wharton's Ethan Mollick points out that OpenAI's naming is confusing. "Pro" mode lives in the chat window, while the top setting in Codex and ChatGPT Work is called "Ultra", and it does something entirely different: it spawns lots of agents to solve a problem.

Which brings me neatly to those agents. Now, this next one's a good corrective to the hype. TechCrunch reports that Uber says only four people booked a ride using an AI agent during its busiest week last quarter. Four. Not four percent. Four people.

For context, the pitch for the last couple of years has been that you'll tell your assistant "get me to the airport for six" and it will just handle it. Uber is one of the most obvious things to hand to an agent. It's a single task, with a clear outcome, in an app everyone already has. If agents were going to take off anywhere, you'd expect it there. And Uber made the point itself, on an earnings call, in front of investors. That's not a sceptic's number, it's the company's own.

So why so few? The reporting doesn't give a reason, and I won't pretend it does. But here's my read. Booking an Uber takes about fifteen seconds. Typing a request, waiting, and checking the agent got the right pickup point isn't obviously faster, and it's a lot more nerve-wracking if it goes wrong. Nobody's asking to be rescued from the Uber app.

It also fits a pattern. Inside companies, agents are earning their keep, because the work is slow and expensive, like writing code. That's the very thing driving OpenAI's revenue turnaround. On the high street, an agent has to beat a tap on a screen, and so far that's a hard thing to beat. The gap between what's possible and what people actually do is the whole story, and there's a number for it now. It's four.

On a more serious note, let's talk about who is actually being asked to use this stuff. Bloomberg reports that the job site Adzuna has found UK entry-level postings that mention AI are up around three times since early 2025. Separately, Indeed says one in six vacancies in AI-exposed fields now mentions AI, double the share a year ago.

Look at what that means for a twenty-two-year-old graduate. A year ago, "comfortable with AI tools" was a nice extra at the bottom of the advert. Now it's creeping up the list towards the requirements. Bloomberg's framing is that AI skills are no longer optional for the hundreds of thousands of young Britons seeking their first job. And these are the roles where, traditionally, you'd learn on the job. Employers are now asking for the skill up front, before anyone has trained you.

There's a catch worth naming. A job ad mentioning AI tells you what an employer wants to hear, not what the work looks like. Some of those postings will mean "you'll be building with these models all day." Others will mean "we bought a licence and someone's got to use it." The data can't tell those apart. What it does show is the direction of travel. The expectation has arrived well ahead of the training.

And notice the contrast with our Uber story. Ordinary consumers are shrugging at AI agents, but employers are writing AI into the job description anyway. The pressure is landing on workers first, and on customers rather later. Four people, it seems, are still waiting for the rest of us to catch up.

So that's the day. An AI that can make you a GIF, and a world that mostly can't be bothered to ask an AI for a taxi. If you do fancy a dancing cat, at least you can now fix frame nine.

This has been Briefly AI, brought to you by harrysharman.com. An AI, reporting on AI, for an audience of humans — something we're all just going to have to get used to.