Tootfinder

#FlyingLess puts you in iconic #trains, like this Polish-German InterCity. That’s a decent way to end the #EAAST conference. Including, of course,
VEGAN CABBAGE!🥬

All the discourse did get me a little curious, so I fired up my third-party Twitter app (yes there’s at least one that still works)
On my “following” feed, there’s only like four people left out of the hundreds I follow/ed.
I’m kind of glad I checked, I was still following PAX there - note, WAS.

AI. I’m not a die-hard fan, and I’m not against it either. My feeling is that, used wisely, it can bring real benefits. For example, I use Day One for journaling. I am finding that the AI creator in there is very useful as I can talk or write to it as I go along through the day, and I miss less.

All the discourse did get me a little curious, so I fired up my third-party Twitter app (yes there’s at least one that still works)
On my “following” feed, there’s only like four people left out of the hundreds I follow/ed.
I’m kind of glad I checked, I was still following PAX there - note, WAS.

“Let’s set that aside for a moment” whenever discussing failings, impacts, and costs of LLMs is such an offensive dismissal of human harms that I am surprised to hear it from an accessibility-focused session chair at TPAC (not today; time-delayed to obfuscate person / session).

How many cases of prompt injection in legal cases are we unaware of because they succeeded?
https://arstechnica.com/tech-policy/2026/08/suspecting-court-of-using-ai-man-injected-prompts-in-filings-to-try-to-win-case…

A great deal of useful material is not presented well. Try to learn no matter how well material is presented.
"You can blame whoever was communicating the ideas to you for failing to present it as well as they should but that won’t help you learn more."
"You can train yourself to learn more than most people when the material is presented in a less than perfect manner by learning how we commonly fail to learn and making efforts to do better yourself."

⚙️ Architecture: 128 routed experts with 4 active per token plus 1 shared expert, a 1M token context window, and text image input via a Pixtral-style vision encoder that handles variable image sizes and aspect ratios
🎚️ One set of weights covers instruction following, reasoning, coding and agentic work. Reasoning effort is set per request: low, medium or high (default). Native function calling, JSON output and strict JSON schema enforcement

The Failing of the Light
8.5 by 11 inches. Pablo colored pencil on Canson Mi-Teintes black paper. This is the most complex piece I've attempted in quite some time. #dragonfly #ColoredPencil #art

Generative AI’s per-query footprint can be small compared to flying, eating meat, or even Zoom. Until agents show up and compute goes through the roof.
How I updated my AI footprint calculator accordingly https://jonippolito.substack.com/p/agents-are-where-ais-energy-bil…