Ian Danforth
Ian Danforth
@iandanforth.bsky.social
AI, Robots, Neuroscience, and Liberal Politics.
This is not *at all* what I intended to generate with Veo3, but it's actually pretty cute and satisfying to watch, so here you go!
September 1, 2025 at 12:58 AM
We use Baby Connect which is great, but it doesn't have this view which shows which allergens are due, which is why I asked Claude to build it for us. :) Maybe Huckleberry does!
July 10, 2025 at 2:55 PM
@jeffclune.com coincidental post from John Carmack today, crediting batch size for overcoming catastrophic forgetting in frontier language models.
May 23, 2025 at 7:07 AM
I want this game to exist.
May 18, 2025 at 12:29 AM
I don't care what your designer or brand manager says, don't do this
May 15, 2025 at 3:41 PM
Ok, share links don't include uploaded images apparently so here's the input image I asked it to work with.
April 11, 2025 at 11:21 PM
It's pretty good at flattery too
April 11, 2025 at 5:10 PM
FFS statista is not a reputable source. Latest report doesn't do the full breakdown, but, shockingly, a well established relationship doesn't reverse. (For me this is like hearing someone say that length has no impact on the volume of a box)
April 7, 2025 at 11:59 PM
I did, your numbers are bunk. Here are some from the CDC.
April 7, 2025 at 11:43 PM
I shared my opinion on AI's use of copyrighted works. (I'm a strong supporter of this use). Share your opinion as well: www.federalregister.gov/documents/20...
March 16, 2025 at 4:10 PM
It's not a panacea, but have you actually looked at a baking soda box? It has an FDA evaluated Drug Facts label. So for heartburn it really is medicine.
February 26, 2025 at 2:06 PM
Alibaba released an open video model called Wan 2.1 and it's pretty good! I was able to try it out on nim.video for free (new accounts get enough credits for 3 generations)
February 26, 2025 at 1:46 PM
Well done @anthropic.com 😂
February 25, 2025 at 12:54 AM
I'm quite pleased to report that I can load and interact with Llama 3.1 70b via Ollama on my new Mac Mini! It's not much faster than reading speed, but that's good enough in many scenarios.
November 17, 2024 at 12:54 AM
"So what's next?" you might ask. Well one of my leading contenders is summed up in this post from the other site by Kevin Black of Physical Intelligence.
November 13, 2024 at 6:52 PM
Has scaling plateaued? Perhaps! But recall that science is more often a series of sigmoids than a long term exponential. If you've been around you'll recognize this pattern even from ImageNet progress!
November 13, 2024 at 6:47 PM