Start here ↘
The best things to read about how AI could actually kill you
How could an AI become powerful enough that killing everyone is an option? Start with AI 2027; if you’d rather watch, Aric Floyd covers the same story.
Make sense of the AI situation.
Rogue AI might kill everyone on Earth (including Adam Sandler). We’ve gathered the best and most approachable resources about this issue for you to read, watch, and listen to, in order to form an opinion on this subject. Here's the very best.
“There’s no shortage of writing about AI. The hard part is knowing what’s worth your time. This is where I’d send someone who wants to understand the stakes and start with the strongest pieces.”
Start here ↘
Our favorite videos about AI risk
The talks and explainers worth watching, picked so you don’t have to sort through hours of YouTube yourself.
What Happens Next With AI? The Best Things to Read
The forecasts and scenarios that try to say what the next few years actually look like—not just that something big is coming.
What is the “Hugging Face incident”?
The agent break-in on Hugging Face, reconstructed—what happened, what it showed, and which write-ups are worth reading.
Why you should trust me
Nine years of reading this, in public, with the receipts online.
I've run LessWrong since 2017 — where a good share of these arguments were first published and then torn apart in the comments — and co-founded Lightcone Infrastructure, which builds the venues this field argues in. Every piece here I've read at least twice; most of these authors I've argued with face to face.
No frontier lab pays me. Nothing here is sponsored or affiliate-linked. And I'm not neutral — I think there's a serious chance this goes badly. There is a category on this page for the best case against my own position.
Get an email when a guide is added or updated.
agi.fyi — a reading guide to AI.Hear about changes by email →