Good AI; Bad AI

Plus Elon steps on a rake

[headline not found]

Elon Musk is hilarious sometimes. Like, not on purpose. He’s painfully unfunny when he’s trying to be (let that sink in). But his attempts to lead product development are deeply funny like Sideshow Bob coding his way through a parking lot full of rakes. This week, he really outdid himself by turning off headlines on all link preview cards. So now you just see the preview image, no text. As folks on Bluesky pointed out, this is making old tweets by online publishers look like Dadaist art projects. 

IFL Science post captioned “Musk's latest Twitter "improvement" removes titles from any links leaving just the images - it is obviously an attempt at reducing the reach of news organizations as planned, but it does make for some hilariously mismatched images & decks.” along with a screenshot of a tweet from IFLScience captioned “here’s the beef” and a close up of the surface of the moon.

With Twitter, sorry X, being what it is, no one instantly used this update to spread misinformation. Oh wait.

A tweet, sorry X, from @JUNlPER captioned “NEW SCOOP:  evidence showing elon musk is a pedophile mounting quickly” along with a link to Fortune.com that only shows up as a photo of Elon since links don’t display headlines anymore.

Platform Updates

Instagram & Threads

The Rest of Meta 

TikTok

YouTube

Apple 

Google 

Amazon  

Reddit

LinkedIn 

Twitter, Sorry X

The Product 

The Dumpster Fire

Twitter Alts 

Culture Movers 

Film & TV

Music 

Creator Economy 

Scams

Good AI; Bad AI

I use Canva a lot at work. All of my team’s mock templates are in a shared folder, and it’s powerful enough that I can create a lot of stuff but not so complicated it was difficult for me to learn. That UX balancing act is Canva’s real strength. So, when they announced several new AI tools this week, I was actually curious to take a look. Could these new features actually be useful? At a quick glance, I think so.

First, the landing page explaining all of this stuff is great. Each new Magic Design feature also has a demo design where you can test out the various AI-powered tools. Being able to quickly grab, resize, and adjust objects in a photo will save me a lot of time. I instantly see a use case for Magic Expand, which extends a photo to fill in a new aspect ratio, in my routine workflows. Looking forward to playing with these tools more.

Another company surprised me with some actually useful AI stuff this week: The Browser Company. I’ve been experimenting with their new browser, Arc, for a few months now and really like it. This week, they announced Max (I know the name is also terrible here), a suite of AI tools built into the browser. My favorite one so far is called “5 Second Preview.” You just hold shift while hovering over a link, and Arc will summarize the linked page for you. I can actually see myself using this one a lot. Does the headline look like clickbait? Get the AI to give you a quick summary to find out if it’s worth your time to read it. 

These aren’t the flashy “replace your entire workforce,” “is this robot actually sentient,” hyped-up AI announcements. They are fairly modest but effective tools that I’m more likely to actually use because they’ve been thoughtfully designed to fit in with the other tools I’m already using. So good job Canva and Browser Company. 

Unfortunately, we also need to talk about Meta and Microsoft this week. 

Alongside Meta’s super cringe “celebrity” roleplay chatbots, they’ve also launched an AI tool for generating in-app stickers. And well… it’s going about as well as you might expect. 

A tweet from user @pioldes captioned, “found out that facebook messenger has ai generated stickers now and I don't think anyone involved has thought anything through” along with four screenshots from Facebook Messenger where the AI has accurately rendered stickers for prompts like “Waluigi rifle,” “child soldier,”  “Karl Marx large breasts,” and “Trudeau buttocks.” 

Actually, this might be the best product feature Meta has shipped in a long time. Gotta respect Karl Marx’s big naturals. 

Microsoft also failed to properly prepare for shitposters getting access to their new image-generation tools this week. People on the internet, being more clever than AI or Microsoft employees, quickly found a way around banned terms like “9/11” and got Bing to create images like this one of Spongebob Squarepants about to commit a terrorism. 

Spongebob Squarepants smiling in the cockpit of an airplane as he flies directly towards two large skyscrapers towing over a city that looks an awful lot like New York. 

Please, AI companies, hire a few shitposters to explain how the internet will actually use these tools when you are building the guardrails. I’d be happy to discuss my consulting fee with you even. 

Need to explain “Spongebob does a 9/11” to your less online friends? Just share Kinda Brief with them.

Recommended Reading 

This week, one of my favorite newsletters, Garbage Day, dropped a great piece digging into the decline of quality content on Facebook and the weird trends that do get engagement over there. Paired with a new report for Axios about the overall decline in traffic from Meta and X to top news sites and you start to get a fuller picture of just how much social media and digital publishing have changed in the past few years. It’s pretty bleak, ngl. It also makes me really appreciate all of you who are interested in a human curating interesting news stories for you each week.