[ Chapter 04 — Horizon ]
After intelligence
The orb you're staring at doesn't exist. It's pure mathematics, deformed by noise and light every sixtieth of a second. Drag it. The future feels much the same — malleable, if you push.
[ Projections, not promises ]
The curve keeps bending upward
[ The road ahead ]
Milestones we can see from here
Agentic workflows
Models stop answering questions and start completing missions — planning, browsing, coding and verifying across hours of autonomous work with humans as editors, not operators.
Scientific co-pilots
Hypothesis generation at industrial scale. Materials, drugs and theorems discovered in simulation first; laboratories become places where machines propose and humans dispose.
Personal models
Your lifetime of context — files, health, taste, correspondence — distilled into an on-device intelligence that knows you better than any cloud service ever could. Privacy stops being a trade-off.
The open question
Somewhere past this line sits either the best tool our species ever built or its most interesting successor. Alignment research is the art of making sure those are the same thing.
The future is not something we enter. It is something we build — one gradient step at a time.