Google's new Gemini AI agent, Spark, just casually knew the name of a journalist's dog and his colleague's wife's first name - without either of them actually telling it. No prompts. No forms filled out. Just vibes and data, apparently.
The Verge's David Pierce and Jay Peters both went hands-on with Spark this week, and their shared takeaway was essentially: "This thing is unnervingly good." Which, sure, impressive! A little creepy! But also... now what?
The productivity treadmill nobody asked for
Here's the thing that's starting to keep people up at night - and it's not the AI knowing your pet's name. It's the growing suspicion that all of this jaw-dropping technological horsepower is being aimed directly at the least interesting target possible: making us marginally more productive at jobs we probably shouldn't have to do the way we currently do them anyway.
Every new AI feature announcement follows a suspiciously similar script. "It'll summarize your emails! Schedule your meetings! Draft your replies!" Great. Cool. But the emails are still there. The meetings are still happening. The replies still need to exist. We've built a rocket ship to go slightly faster on a hamster wheel.
Scary smart, wrong direction
The unsettling part isn't that Spark knows your dog's name - it's what that capability represents. The infrastructure, the training data, the inference costs, the engineering hours - all of it pointing toward... helping you get through your task list faster so you can fill it up again.
As The Verge notes in their coverage, what's truly revealing about AI getting better is that it exposes the emptiness at the center of the "productivity" promise. We're not being offered a different life. We're being offered the same life, slightly optimized.
There's a version of this technology that asks harder questions - about why we're overloaded in the first place, about what work actually matters, about what humans genuinely want more time for. That version is considerably less fun to demo at a keynote.
So what do we do with this?
Probably nothing, if history is any guide. We'll download Spark, be briefly amazed that it knows Frida is a golden retriever, let it summarize seventeen emails we shouldn't have received, and call it a win.
But maybe - just maybe - the fact that AI is now good enough to be genuinely spooky is also the moment to ask whether "spooky productivity assistant" was really the dream worth building toward. The technology clearly has the horsepower. Someone just needs to point it somewhere worth going.
