Subscribe
By Tom Davenport —

AI Tweets You Missed This Week: 7th June

The first of the weekly roundup, where I share the tweets you might have missed from the from frontline of the AI twitter community. Thanks for the positive response to the first issue where I tried to cover all of May.

If you're reading this on email, click through to the full post to see the tweet embeds and full videos/images in action.

Rumours (that didn't happen)

The Countdown to GPT-5 bot on Twitter got everyone very excited, then inevitably didn't play out. Here's the final act of the account's rug pull:

Creative

Incredible anime progress showing Animatediff's seamless results:

Open source consistent models available on Replicate:

Sora level video generation from China. We're currently at the fun end of the arms race wedge:

Honestly, I thought this next one was real. The comedy timing on Sanders is perfect. I still can't tell what's going on, but according to the replies: this is a conversation that never happened and apparently a (very well edited) deep fake:

The creative drama this week was an AI artist with a wonderful style and concept called 'Babes' which other artists loved... until they realised it was made with AI. In perfect irony, an artist stole the style, proving that artists are trained on other art, just like AI. The final twist: community notes attempting to claim the original AI artist was stealing from the person who stole her art 😵‍💫

Perplexity now lets you publish results as a page, effectively letting it build surface area for itself online with the offcuts from user searches. The effect may end up being a wikipedia-scale content play:

Websim

Websim is the AI edgelord playground. Scratch here with another upgrade to my previously reported MS-DOS emulation, this time extending with a Discord app:

More in Websim: a language builder. Tolkien would be proud.

And this one with a 3D game environment:

Speech

A new open source speech model, Ultravox:

Tools

A new open source AI search engine:

A fun latent space browser:

Coding

Interesting concept here to force AI to complete code that tests itself:

This one runs code in a sandbox to debug and test:

Another Devin copy is out in the wild:

Prompt hacking

Pliny keeps making Godmode GPTs that get taken down by OpenAI, typically using l33t speak to push messages through:

Confusing but involves Ilya so have it

Open AI's latest research is supposedly what Ilya was working on before their conscious uncoupling:

Opinions

Claire is one of the premier AI artists:

Mike proving that regular people on the front line like you and I are in a better position that Forrester to understand and assess the current landscape:

Them's fighting words on Langchain:

Leopold caused a stir with his Dwarkesh podcast appearance this week. A few takeaways:

Adam found GPT-4o giving meaningful feedback on his MRI scan images:

Thanks for reading and leave a comment if you have feedback or a hot take, I might include it next week. And sign up if you want this every Friday.

Subscribe to Tom Davenport

Don’t miss out on the latest issues. Sign up now to get access to the library of members-only issues.
jamie@example.com
Subscribe