A publication about AI and actual work
The cheaper AI gets, the more expensive the detour can become.
Age of AI Empires looks past token prices to examine what AI-assisted work actually costs: time, attention, corrections, scope creep, and the occasional useful rabbit hole.
The invoice records the engine.
The task absorbs the journey.
The task absorbs the journey.
Prompt
Useful result
errors, retries, correction optional discovery
Featured argument / 01
What does it really cost to get something done with an LLM?
Token prices tell us what computation costs. They do not tell us what the work costs.
01
The LLM Task-Cost Paradox
Why better AI can cost more per task. A plain-English examination of remediation leakage, expansion leakage, the completion boundary, and the difference between task cost and task value.
The library
Ideas you can use before the jargon arrives.
Frameworks
Short models for long conversations.
Five-minute versions of the ideas behind the essays, with definitions, equations, and diagrams.
↗Notebook
Where the argument is still being sharpened.
Working notes, test protocols, open questions, and the uncomfortable bits left visible.
↗About
A small publication for a large transition.
Built for people who want to understand AI without pretending every prompt is a revolution.
↗The working thesis
AI changes the cost of thinking. It does not remove the cost of deciding what deserves to be done.— Age of AI Empires