Justin Pinkney

397 Followers
188 Following
116 Posts
I'm a machine learning researcher who likes image generation and editing. I co-created Toonify and work at Midjourney
homepagehttps://www.justinpinkney.com
@kayserifserif I used to use obsidian but it had way too many features just for editing some markdown. After trawling through a bunch of unsatisfying apps I recently found One Markdown which has been fantastic and feature rich but simple: https://apps.apple.com/kz/app/one-markdown/id1507139439
‎One Markdown

‎One Markdown is a simple and fast editor that supports plain text, Markdown and Textbundle. It has the following features: 1. Fully supports CommonMark standard and GitHub Flavored Markdown (GFM). 2. The same basic library is used for syntax highlighting and parsing into HTML, and there will be n…

App Store
@jonikorpi seems like you need Wolfram's walking desk

Announcing my new app, Gibberish. Gibberish is a weird blogging app that looks and feels like a messaging app.

https://andadinosaur.com/gibberish-is-now-available-on-testflight

Gibberish Is Now Available on TestFlight - And a Dinosaur

This post first appeared on my Gibberish blog. Gibberish is a blogging app that looks and feels like a messaging app. It’s a bit weird, but that’s the point. This UI tricks my brain into writing mo...

@bit101 this sort of interface is great for young kids, has been done really well for a while by Yoto

How to tell someone has used chat-gpt to make their seo spam programming article on Medium.

This is why Google search is doomed.

I don't know if it's particularly interesting anymore these days, but I did try training Würstchen v2 on Pokemon. My training script is here: https://github.com/justinpinkney/wuerstchen/
GitHub - justinpinkney/Wuerstchen: Official implementation of Würstchen: Efficient Pretraining of Text-to-Image Models

Official implementation of Würstchen: Efficient Pretraining of Text-to-Image Models - GitHub - justinpinkney/Wuerstchen: Official implementation of Würstchen: Efficient Pretraining of Text-to-Image...

GitHub
@simon quotations with an llm seem so wasteful, imagine all the computation required to copy and paste a range of tokens from the context to the output! If only there was a shortcut way for it to refer to sections of the input.

This website is beautiful ✿✿
Deux Fleurs is a non profit organization providing hosting without data centers. They are using old computers at home in France, using open source technologies.

https://deuxfleurs.fr/

Deuxfleurs : fabriquons un internet convivial

Association qui œuvre en utilisant ses propres infrastructures, des outils sobres et des valeurs fortes.

@endrift isn't the modern internet just one giant popup ad we have been unable to close?
But according to the Model license in the GitHub repo the weights will be non-commercial. What happened to the whole open source ethos?