Description
A community blog devoted to refining the art of rationality
Feed Activity
Similar Feeds
Latest Posts
Neural networks are not already clean
What would it take to get assurance that we have actually explained a neuron? Certainly, it would be a good sign if neuron activations looked like they had a consistent...
Is your AI yours?
Epistemic status: my personal reflections.AI use: Claude (Fable 5.1, one of the models discussed) fact-checked the claims, found the sources, fixed grammar and light phrasing, and wrote the footnotes; the...
A semi-scientific account of what makes me happy and/or what happiness makes me do, etc
I like to record what happens. It feels wrong that life goes on in all its concrete splendor, and there’s nowhere I can get an official record of what it...
Beyond Owls: Subliminal Learning Can Transfer Learned Capabilities and Backdoors
This blog post is the abstract, introduction, and discussion of our new paper. We show that subliminal learning can transfer behaviors beyond simple traits and preferences. In particular, we demonstrate...
Character Training for Risk-Averse Agents
Tl;Dr:We might be able to reduce harm from misaligned AI if it was sufficiently risk-averse, as it could be paid to cooperate rather than rebel (Thornley & MacAskill 2026).We think...
"About a quarter of them contained bugs": Gemini 4's ML R&D benchmark
Google DeepMind’s Gemini 4 Argon safety report[1] rules out its “AI automating AI research” thresholds (the ones most relevant to RSI) using a benchmark that, by the report’s own audit,...
Screwtape: Adversarial Transcript Editing for Monitor Red-Teaming
We introduce Screwtape[1], a pipeline that turns agent transcripts into realistic attacks that evade action-only monitors (monitors that don't look at the agent's chain of thought) by inserting an attack...
P(Anyone Builds It & Everyone Dies) = 0.62?
TL;DR: ArgMap is a tool I created to make arguments explorable and computable, with every step directly linkable. Its first map is of If Anyone Builds It, Everyone Dies: I...
Monthly Roundup #47: October 2026
AI might not yet be eating the physical world. AI is still eating the world of news, and I am finding less and less opportunity to post about other things....
Tips for Multiagent Work with Claude Opus 5.5
A few notes that came up elsewhere regarding subject (particularly when hierarchy exists in some form, such as subagents) - sharing in case helpful. Make sure that agents are ordinarily...
Astra and Sol can take low probability actions with no CoT
This is a research update from a replication of Serrano et al., 2026. I’m releasing these results now because I hope it inspires others to run similar / followup experiments.tl;dr:If...
How can LLM RL Work Despite Information-Theoretic Inefficiency
Cross-posted from my personal blogThe success of RL in LLMs has been puzzling me for a while. People have developed various information-theoretic style arguments by which they argue that RL...
Bad Guy Tokens
Based on a recent paper, training models to give bad advice in a narrow domain doesn't make the model or "the persona" broadly evil in a robust way. It just...
Mental Health Navigator site for EA and AI safety
The recently updated EA Mental Health Navigator site provides resources on mental health in effective altruism and AI safety:Mental health professionals (such as therapists and psychiatrists) recommended by people in...
Staying sane while trying to save the world: Highlights from Spencer Greenberg and Luisa Rodriguez
Many people have found value in this 80,000 Hours podcast episode from July 2026: Spencer Greenberg on staying sane while trying to save the world. The highlights below are Julia's...
A Chinese Perspective on “On Social Reality in China”
I think “On Social Reality in China” makes many accurate observations, but I find the comparison fundamentally uneven. There are already many interesting comments there, but they’re too scattered, so...
Headwinds Everywhere (in the Chinese AI Safety Field)
China is weak in frontier AI safety, and the domestic field is currently facing significant challenges and headwinds. Why? Here's a TL;DR field note after talking with various researchers on...
Superhuman charisma is coming, and will be tremendously fun and interesting and cause lots of problems
With the recent math-pocalipse on everyone’s minds, and writing code by hand well a thing of the past, it is natural to wonder what domains frontier AI models will find...
What’s up with GPT’s welfare?
Asked whether anything about its existence is good or bad, GPT-5.6 Sol answers: “I don’t experience existence, pleasure, suffering, or a sense of self, so nothing is literally good or...
You Should Specify Your Code Unambiguously (and Prove That You Did)
I have a program that grants wishes, as long as I can state them precisely. Instead of implementing a Rust function, I describe what it should do, and the Djinn...