Gah, its been a long while since I used them. Thanks for the tip.

Had success before with www.abebooks.co.uk

Yes. But it is important to realise that models are getting larger and deeper to the point that its a legitimate to be concerned about their internal “thoughts” and not just the text they output. Vid explains why really well.

I find Robert Miles really helpful as a navigator of AI safety issues. This video in particular is good. He acknowledges that we can anthropomorphise LLMs too much, but it is still helpful to think of them as “thinking” rather than just generating language. www.youtube.com/watch?v=iuHd...

See also: jamesking.io/2026/06/10/h...

It _feels like_ Astra remembered choices it made at lower reasoning levels (assuming those ran first) or Astra just makes more consistent artistic choices when approaching the task from scratch.

FYI: Link to pelican transcript in your post is missing.

Observation: Pelicans for Astra are more consistent in the drawing choices from low->high. Arbitrary details like position of the sun, number of tail-feathers, shape of wing/arm is the same, but for other models no consistency between levels.

Congrats! Been following your video reports on Audacity’s progress. Great to see it come to fruition.

I like this post by Emilie Rosenlund-Soysal ergodicityeconomics.com/blog/time-sp...

Oops - I made a small mistake in the gif above - the notebook that I download is new.observablehq.com/@nxrix/binar..., but the one I then open in Observable Desktop is a remix: new.observablehq.com/d/cca8d2013f... which I downloaded earlier.

Poundstone is really good at exposition of mathematical concepts. I really like this description of ergodicity (pic, with alt text included). The book is chock full of efficient explanations like this.

Do read the book!