AI At Your Worst

22

Google wants to let you deepfake yourself.

Also announced at Google I/O 2024, assuming we aren’t already there.

They are sprucing up the Gemini models. Revamping search. Stuffing AI agents into every crack. Smart glasses arriving this fall, too, because hardware still sells.

Search Is Leaving the Chat

The results look vibe-coded now. Super widgets. Bots that do not sleep. Google thinks you don’t need to be there. Not anymore. The search experience is becoming hyper-personalized, fully automated, aggressively AI-driven.

It is not looking for you to type anymore.

Goblins Are Bad for Coding

OpenAI gave its Codex agent very strict rules. No talking about goblins. Or gremlins. Raccoons are out too. Trolls? Ogre? Pigeons? Absolutely not. The instructions are explicit: only mention creatures if it is absolutely and unambigiously relevant to the code.

Really? Who asks the bot to code a dragon slayer?

Better Pixels, Same Problem

ChatGPT Images 2.0 has dropped. It makes sharper pictures. It actually renders text properly this time, which is a relief. But test it in anything other than English.

It still stumbles.

The details are better but the language support lags.

The Job Anxiety Pivot

Demis Hassabis says laying people off is stupid.

He told WIRED that AI gives you productivity gains so you can do more, not so you can fire staff. Interesting. Coming from the guy whose AI models can do most jobs cheaper than humans.

Then you have Mira Murati. Former OpenAI CTO now at Thinking Machines Lab. She wants humans in the loop. Collaboration over automation. She builds AI that helps you rather than replacing you entirely.

One CEO says keep the humans. The other implies efficiency wins.

Who is right? Maybe both. Or neither.

Android Gets Smart-ish

Android 17 is coming. Summer arrival. You can generate your own widgets. Ask Gemini to finish your bookings while you browse Chrome on your phone. It is small stuff.

But it is small stuff that adds up.

CEO Clones

Mark Zuckerberg. Jack Dorsey. Different visions. Same goal: heighten control.

They imagine using AI to be everywhere at once. Management via proxy. The boss never sleeps, the boss never blinks, the boss is always watching through a digital lens.

Can we stop calling these things human?

Anthropic just announced “dreaming.” For sorting “memories.” The agents “dream” about their data. It sounds creepy. It feels invasive. It makes AI sound like a person when it is really just math moving fast.

We should just call it processing.

Apple Is A Subscription

Tim Cook is leaving. His legacy is services. A monthly subscription for your entire digital life. The incoming CEO, John Ternus, inherits the ship. Now he has to steer it into the AI era while the subscription model hardens.

Is this progress or a toll booth?

Your New Worst Employee

Finally, there is Gemini Spark.

Google’s answer to the always-on agents from others like OpenClaw. It runs 24/7. It hungers for your data. It will send your emails for you. It might spend your money for you, too, if you let it.

Google is making it easy to build a digital ghost of yourself.

One that works harder than you.

Do you really want to know what it does when you sleep?