AI As Time Compression
AI as time compression is Li Jigang / 李继刚’s claim in E45 孟岩对话李继刚：人何以自处 that AI changes the time dimension of knowledge work. In his three-world frame, the atomic world is constrained by location, the internet or bit world removes much of the spatial distance between points, and the AI or vector world compresses the time needed to search, read, extract, simulate, and receive feedback.
The episode’s example is reading. In the internet era, a person could instantly download a hundred books but still had to read them with personal time and brainpower. In the AI era, a model can summarize, structure, compare, and discuss those materials as if a large share of human textual history had been crystallized into a callable medium.
The concept does not say time literally disappears. It says many cognitive loops become faster enough that the remaining human bottleneck moves toward purpose, question choice, AI Use Pacing, Human Agency Under AI, and Wet-State Human Agency.
Key Claims
- AI compresses the elapsed time between question, context gathering, synthesis, and feedback.
- Faster cognitive loops can make the world feel like it has a higher flow rate.
- Time compression changes reading, investing, company research, prompting, education, and personal automation.
- Speed does not settle which questions deserve attention; it makes human agency and pacing more important.
- Compression can turn ordinary use of AI-Assisted Reading and Context Engineering into a major personal leverage difference.
Connections
- Li Jigang / 李继刚 — source speaker for the frame.
- AI-Assisted Reading and Reading As Frame Training — reading practices changed by time compression.
- AI Use Pacing — discipline needed when the AI work loop accelerates.
- Human Agency Under AI and Wet-State Human Agency — human layers that remain after cognitive speed increases.
- Context Engineering and Persistent Agent Memory — mechanisms for keeping compressed knowledge loops useful across sessions.
- AI Inference Cost Structure — practical cost layer when compressed cognitive work consumes tokens.