Short posts
New from me: Researchers tested AI on hundreds of real freelance jobs. The best system did 2.5% successfully.
New: We got data behind a ChatGPT wrongful death lawsuit. The results are horrifying.
More details here (🎁): https://wapo.st/4qxaGGe
New AI image technique by Qwen is the first I’ve seen that uses layers natively — like actual artists and designers do. Once layer-aware AI gets built into design software, look out.
Qwen-Image-Layered: Towards Inherent Editability via Layer Decomposition
I partnered with
@geoffreyfowler.bsky.social
to test a bunch of AI editing tools, and something very interesting happened.
We asked Gemini to generate a professional photo of an actor crying at the Oscars. It did — including a fake copyright notice from a real AP photographer.
Google said it was “refining” its safeguards. The AP didn’t answer questions about whether Gemini had rights to train on its photos.
The full test results 🎁 -> https://wapo.st/48IQawf
How interesting that OpenAI did not have to train a new video model to start generating Disney characters … :| https://openai.com/index/disney-sora-agreement/
Caught up on some blogs post-Thanksgiving and gotta recommend these two gems:
“You should write an agent” I endeavored to do this myself one week, and then somehow just 30 minutes later it was done) -> https://fly.io/blog/everyone-write-an-agent/
“Parsing PDFs with Antigravity” Another win for using AI to do journalism tasks in a reproducible way aka by writing code you can check and rerun ->
Parsing PDFs with Antigravity – Matt Waite’s Collection of Miscellany
Made a chart showing how OpenAI’s lead has evaporated, according to the Artificial Analysis intelligence index
Full story -> 🎁 https://wapo.st/3Xy8Xnz
Interesting study -> New research suggests AI chatbots can shift people’s political views more effectively than campaign ads on TV. https://wapo.st/49RSstP
So many nuggets in here, like “Mental health experts told his team, for example, that sleep deprivation was often linked to mania. Previously, models had been “naïve” about this, he said, and might congratulate someone who said they never needed to sleep.” https://www.nytimes.com/2025/11/23/technology/openai-chatgpt-users-risks.html?smid=url-share
On the one hand, ChatGPT is super popular. On the other hand, AI companies need to make SO MUCH revenue.
Bubble or no bubble? More data here -> 🎁 https://wapo.st/3KfxpXE
