Scaling Axes
An evolving view of the inputs to AI progress: compute, pretraining, post-training, test-time work, deployment and research, with the data behind each.
A place to
understand AI.
Explore the ideas, systems, and questions shaping the path to AGI.
Begin your journeyThe collection
15 drops
How models learn what to do
Follow the learning loop from demonstrations and feedback to practice, parameter updates, and the next task.
What happens when an AI agent works?
Trace a coding task through model calls, the agent harness, and the computer that searches, edits, and tests.
The physical stack behind AI
Explore the infrastructure that makes intelligence possible, from accelerators to clusters, power, and cooling.
The supply chain behind advanced AI chips
Follow the interdependent companies, machines, and manufacturing steps behind an advanced chip.
What happens when a model answers?
Follow one question through the production system DeepSeek open-sourced, from tokens to GPUs, caches, and the price of each token.
How AI gets measured
Explore the tests we use to evaluate AI, what their scores reveal, and what they leave unanswered.
Aligning intelligence
Understand the challenge of making powerful intelligence reliably serve human values and intent.
The intelligence that builds intelligence
Enter the research loop: how AI could help improve AI, what the evidence shows, and what remains uncertain.
The Millennium Prize Problems
Explore seven landmark mathematical problems through visual essays and an interactive world map. Follow the questions, the ideas, and the frontiers of proof.
VOL. 11In progress
Fine tuning a model
In progress. This volume is still being written.
VOL. 12In progress
Interpretability
In progress. This volume is still being written.
VOL. 13In progress
AI chips
GPU, TPU, and specialized AI hardware: Groq, Cerebras, and more.
VOL. 14In progress
Data Center Policy
In progress. This volume is still being written.
VOL. 15In progress
SpaceX Falcon 9
In progress. This volume is still being written.