Aug 27, 2026 • 1 min read
Open-Source Models Change the Economics
When a human chats with a model for a few minutes, inference cost is invisible. When an agent works for three hours and spawns twenty more, intelligence becomes a line item.
Part 5 of 7 — AI, Software Factories, and the Future of Work
Open-Source Models Change the Economics
Open-source AI isn't only interesting because of sovereignty or ideology.
It changes the economics of agentic software.
When humans use an AI chatbot for a few minutes, inference cost is barely noticeable.
Agents are different.
An agent might work for three hours.
Spawn 20 other agents.
Read thousands of files.
Run dozens of tests.
Repeat the same task multiple times.
Use one model for planning, another for implementation, and another for verification.
Suddenly intelligence becomes a line item.
And once you're operating thousands of agent-hours every day, model economics starts mattering enormously.
This is why I think the future agent stack will be aggressively multi-model.
Use the expensive frontier model where judgment matters.
Use smaller models where the problem is deterministic.
Use open models for high-volume tasks.
Continuously route workloads based on capability, latency, and price.
Companies won't ask:
"Which model do you use?"
The better question will be:
"How efficiently do you convert tokens into useful work?"