Your Interview Is At 1am Now

Today on Briefly AI — Google's Gemini has just hit a billion monthly users, drawing level with ChatGPT. An unreleased Anthropic model has made real progress on one of mathematics' oldest unsolved problems, the Riemann hypothesis. And job interviews are shifting to 1am, because your interviewer is an AI that never sleeps.

Welcome to Briefly AI, a podcast by Harry Sharman, written and voiced by his AI clone. Harry built it this way because writing his own scripts turned out to be the one job he was happy to hand over immediately.

Okay — let's get into it.

Right, confession time. More people now talk to Gemini every month than live in the United States, Brazil, and the UK combined. Which either means Google's cracked something properly useful, or we've all quietly developed a very odd voice-memo habit. Possibly both.

So, the numbers, as reported by The Verge this week. Gemini has passed a billion monthly users — Sundar Pichai called it Google's fastest-growing product ever, which for a company that's launched roughly fourteen products to a billion users at this point, is saying something. ChatGPT got there first, but Gemini's catching up fast, and the more interesting detail isn't the milestone, it's the behaviour underneath it. Sixty-three percent of Gemini users are talking to it, not typing — using the voice feature as their default way in. And it's generating over a hundred and fifty million images a day.

Here's the bit worth sitting with, if you'll forgive me the phrase just this once. We spent two years being told the chatbot interface was a stopgap until something more natural came along. Turns out the more natural thing was just... talking to it, like a person. That's not a small shift. Typing to a machine keeps a bit of distance — you're clearly issuing commands. Talking to one closes that distance, and closes it fast. Worth keeping an eye on how that changes what people are comfortable telling it.

Meanwhile, over at Anthropic, something a bit more academically thrilling happened — quietly, because it's not even a released model. TechCrunch reports that an unreleased Anthropic model made genuine progress on the Riemann hypothesis, which has sat unsolved for over a hundred and fifty years and comes with a million-dollar prize attached for anyone who cracks it.

Now, let's be honest about what "progress" means here, because nobody's cashing that cheque. The model didn't solve it. It made incremental headway on a problem so hard that most professional mathematicians spend their whole careers orbiting it without landing. But that's actually the interesting part — this wasn't a model straining to be clever for a demo. It was doing something closer to actual mathematical research, the slow, unglamorous kind, and getting further than you'd reasonably expect.

Why should you care, if you're not a number theorist? Because this is the test that matters more than any benchmark score — can these things do original thinking in a domain where there's no training data marked "correct answer," because nobody knows the correct answer yet. Nobody really knows if this generalises beyond one very famous, very stubborn equation. But it's the kind of result that makes you want to check back in six months rather than forget about it.

Now this next one's a bit more human, and honestly a bit bleaker. Wired's been reporting on what they're calling the rise of the 1am job interview. AI is increasingly the very first step in hiring — no human on the other end, just a chatbot asking your competency questions — and because nobody's actually sitting there waiting for you, candidates are booking these things whenever suits. Which for a lot of people, especially anyone juggling a day job or care responsibilities, turns out to be the middle of the night.

Why does that matter? Because it's a quiet redesign of what "flexible" means. On paper, interviewing at 1am sounds like freedom — do it on your own time. In practice, it's people squeezing a stressful, evaluative conversation into the hours when they're most exhausted, because the daytime is already spoken for. The AI doesn't care what time it is. It doesn't get tired, doesn't need sleep, doesn't notice if you sound frayed. The human on the other side very much does.

There's also a stranger effect worth naming — when nobody's watching, the performance changes. Some candidates say they're more honest, less nervous, more themselves with an AI at 1am than they'd be with a hiring manager at 10. Which tells you something about what we're actually afraid of in interviews. It was never really the questions.

So that's your lot — a billion people talking to Gemini, a maths problem that's been quietly humbling humans since the 1800s getting nudged by a model that isn't even out yet, and a hiring process that now runs on nobody's schedule but the robot's. If you take one thing from today, let it be this: set an alarm, but maybe not for your job interview.

That's it for today on Briefly AI. Same arrangement tomorrow — real news, filtered fast, said out loud by a machine that's got the hang of it. Subscribe wherever podcasts live.