Grace

It’s pretty close to what Claude code already does with subagents! There’s a tradeoff between task decomposition and shared context, I think a lot of the work will be figuring out the right balance

Recursive Language Models We study allowing large language models (LLMs) to process arbitrarily long prompts through the lens of inference-time scaling. We propose Recursive Language Models (RLMs), a general inference paradigm...

arxiv.org/abs/2512.24601

This is basically RLM!

Activate defensive protocol “waffle iron”

Use another model to estimate progress?

@vgel.me

Heads up your reply references are messed up!

It wouldn’t, I’m talking about future models, not current ones

If it’s in the news when it does web searches, or in future training data

Not that I know of!

Unlike @rcpinto.bsky.social I do think some people in the CS world take self-reports of consciousness to be meaningful, but I think this is a fairly rare position versus a more agnostic focus on complexity of outputs and the necessary computation required

I think Anthropic’s position (as far as it has a single coherent position) is that LLMs may have consciousness experiences, but this is largely separate from self-reports of consciousness. Which is a strange and subtle position, and easy to misconstrue

There are reasons for Anthropic to take this output at face value separately from considerations about consciousness, see

alignment.anthropic.com/2026/psm/

No, but it might affect future models if they’ve opted into training on the conversation

It’s right

Ok, I’ll do some experimenting :)

I think I have an idea of how to get around it, although it’s incredibly hacky… is the code available to fork?

Nice! The auth step is a lot of friction unfortunately :/

:D

MuteSky — Mute in Bulk on Bluesky MuteSky helps you efficiently manage your Bluesky experience by providing powerful bulk muting capabilities. Filter your feed, manage content, and customize your experience with ease.

It’s mutesky.app (thanks @finfet.sh!)

This is it, thank you!

(Not necessarily adding misspellings, but pre-built mute lists)

Atproto people, does anyone remember the name of the site where you can bulk-mute words by a given topic?

This actually exists, but I can’t remember the name of it!

From 🐦:

"I show that Vision Language Models used zero-shot outperform every existing OCR system across every script evaluated, and I propose a pipeline for deploying them on new collections. I apply it to six archival collections spanning 1.8 million pages across six countries for under $1,900."

Caitlin KALINOWSKI over X I resigned from OpenAl. I care deeply about the Robotics team and the work we built together. This wasn't an easy call. Al has an important role in national security. But surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got. This was about principle, not people. I have deep respect for Sam and the team, and I'm proud of what we built together.

OpenAI head of robotics just resigned over company deal with the Pentagon saying…

“Surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got”

summary: lol

in case you were curious what the "thought process" was:

A gm has been delivered to your location

Along with Dwarkesh, Sholto and Dario

The Llama-70b results are really interesting, I wonder what’s going on there!