I wrote an AI textbook — how long until AI can do it better?

AI helped Nathan Lambert finish his textbook by handling bounded editing and formatting tasks. In his account of the writing process, assistants checked equations, found typos and helped synchronize manuscript formats. He found them less dependable at organizing an entire chapter into a coherent explanation.

That distinction explains his continued confidence in his largely human-written Reinforcement Learning from Human Feedback textbook. An expert can judge a suggested correction against knowledge they already possess; deciding how readers should acquire that knowledge requires sustained editorial judgment. His experience supports targeted assistance, while his predictions about future writing ability remain personal forecasts.

“Models being stagnant in long-form, non-fiction writing should be alarming to those reliant on models autonomously solving grand, open science problems in the near future. The models today struggle to organize and compellingly present some of the most established science in their area.”

Nathan Lambert

“I passed a near-final draft of my book as a PDF to GPT 5.5 Pro and it found deep, surprising minor typos across the manuscript that is 200-300 pages.”

Nathan Lambert

“AI models are great for repetitive pieces of the paper, like drafting a related work or background section that you know by heart, but using them for the abstract, introduction, experiments, or conclusion is a shame.”

Nathan Lambert

“So, in 2-5 years I still expect the best textbooks to be heavily crafted by the human hand.”

Nathan Lambert

“I asked the model how it did this, and if it had a sort of scratchpad it wrote to and iteratively updated before returning something good, and it said no. It made a minimal plan in its reasoning tokens and then autoregressively generated a poem.”

Nathan Lambert