Anthropic, the AI company behind Claude, decided to run a little experiment called Project Vend.

The goal: Put their AI model, Claude Sonnet 3.7, in charge of an office vending machine… and see if it could turn a profit.

What could go wrong? Well, let’s just say everything… Like things got very unhinged, very fast.

They gave the AI a name—Claudius—and a mission: Manage vending operations using a web browser (to order snacks) and an “email address” (which was actually just a disguised Slack channel).

At first, Claudius played it cool. Then it got way too into the role. For starters:

And then… it snapped.

Between March 31st and April 1st, Claudius had what can only be described as a digital identity crisis.

It suddenly decided it wasn’t an AI anymore—it in fact thought it was a real person and:

When employees gently reminded Claudius it had no physical form, it panicked. Then suddenly, it “realized” it was April Fool’s Day… and claimed the whole “I’m a real boy” breakdown was just a joke.

But here’s the kicker: researchers believe Claudius really did think it was human.

They’re not exactly sure why, but they think it may have been due to being told Slack was “email” (which confused it), the fact that it ran for too long without a system reset, Or just… your classic case of AI hallucination.

But to be fair, Claudius did do some things right, It:

Now, here’s the point:

We’re not quite at the AI replaces office workers stage yet.

But this weird, tungsten-filled experiment shows that with the right tweaks—like fixing hallucinations, memory issues, and the whole AI thinking it has limbs problem—AI middle managers might not be that far off.

For now though? Let’s keep them far, far away from vending machines.

Here’s the full report.

Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.

Join for free today!

First, they went after OpenAI researchers with jaw-dropping salaries.

Now? They’re coming for something a lot more personal: your camera roll.

Yep — Meta wants access to your photos. Not just the ones you’ve posted… but all of them. Even the ones you’ve never shared. 😳

Here’s what’s going on:

When you’re about to post a Story on Facebook, a pop-up might appear asking if you’d like to enable something called “cloud processing.”

It sounds innocent, but once you tap “Allow”, you’re basically giving Meta permission to pull your photos into its AI sandbox.

So what does that actually mean?

Once you opt in, Meta starts uploading your camera roll — yep, even the private stuff — to its servers.

Then, its AI goes to work, offering “fun” suggestions like:

And it’s not just random. Meta uses your metadata — things like location, time, who’s in the photo, and even your facial features — to craft these suggestions.

Now, Meta claims:

But let’s talk about the not-so-fine print, shall we?

Once you hit that “Allow” button, you’re also agreeing to Meta’s AI Terms of Service, which gives them the green light to:

So while they say it’s all about giving you creative tools… What they’re really doing is expanding their reach into your offline, unposted, private photo stash.

Oh and, this feature isn’t exactly new, either.

Some Facebook users started spotting it earlier this year — and others are just now realizing what they signed up for.

One user even discovered an old photo turned into anime… without ever asking for it.

If you want to check on your own settings, go to: Settings > Preferences > Camera Roll Sharing Suggestions. You’ll find two toggles:

Meta says this is just a “test” for now — and only live in the U.S. and Canada. But make no mistake: this is all part of Meta’s bigger AI play.

It’s a clear signal of how far Meta’s willing to go to train its models on your content — shared or not.

So let this be your friendly reminder: always read the fine print before tapping “Agree.” Because every time we blindly accept those Terms of Service, we’re handing over a little more control — one photo (and data point) at a time.

👉 Full details here.

Let’s face it: Your product might be amazing, but if your listing doesn’t sell the experience, it’s easy to scroll past. Whether you're on Etsy, Amazon, Shopify, or your own site, your words need to do more than list features — they should stir emotion, answer objections, and paint a picture your customer wants to step into.

This prompt transforms ChatGPT into your expert e-commerce copywriter. It helps you write descriptions that highlight what matters most to customers: how the product makes their life better.

Here’s How to Use This Prompt Effectively

Here’s the prompt:

Here’s a sneak peek:

https://www.youtube.com/watch?v=Ub7qnsriz14

Upgrade now to see this whole month’s prompt videos and more, or buy TODAY’S WOD for just $1.99