Personal Email:raelifin@gmail.com (don't be shy!)
Work Email:max@intelligence.org
LessWrong:Raelifin (personal) and Max Harms (work)
Twitter:@raelifin
Substack:@raelifin

Max’s latest novel is Red Heart, available November, 2025. Her meets Three Body Problem in this speculative thriller that examines the potential — and peril — of artificial general intelligence.

Red Heart Cover
“A strikingly authentic portrayal of AI in modern China — both visionary and grounded in cultural truth. This is AI fiction at its finest.”Zhang San, Senior AI Executive
“A thrilling spy novel grounded in realistic AI safety research and geopolitical possibilities.” — Nathan Helm-Burger, AI Alignment Researcher

Pre-order now

Discover more about Red Heart


Max’s debut series is the Crystal Society trilogy, a hard science-fiction trilogy. Told from the point-of-view of an artificial intelligence, Crystal Society explores the nature of minds — human, artificial, and alien.

Crystal Society CoverCrystal Mentality CoverCrystal Eternity Cover
“[Crystal Society] seems to belong in the very, very tiny subset of AI stories that are not bloody stupid, a heroic and almost unbelievable accomplishment.” — Eliezer Yudkowsky

Available in paperback, ebook, and audio... or to read online


To get updates about Max’s books sent directly to your inbox, sign up for his newsletter.

Max’s day job is being a part of the Machine Intelligence Research Institute (MIRI) and working on the project of aligning artificial intelligence to human values. He became concerned that building smarter-than-human minds could be catastrophically dangerous — and potentially lead to human extinction — back in 2009, and has been devoting his life to trying to find solutions ever since. He joined MIRI in 2017.

His most notable research agenda is Corrigibility As Singular Target (CAST), which urges those who are reckless enough to build advanced AI to prioritize corrigibility. Corrigibility is a technical property which roughly involves maintaining a subordinate, tool-like relationship with one or more humans, and ensuring those humans are capable of understanding what the agent is doing, changing the agent's goals, and deactivating it, if the need arises. With CAST, corrigibility is made the central and only goal of the AI agent – obedience and other useful properties flow directly from the adherence to corrigibility. Max’s novel Red Heart centers around an agent designed to be corrigible.