Featured · Field Notes 01 · August 30, 2026
Wait. Remember. Recover.
Access to models is not the bottleneck. The actual problem is putting agents into production to complete jobs end-to-end.
Featured · August 30, 2026
Cayu — everything you need to build long-horizon agents
Runtime to build, Cloud to run, Compound to improve — a standard framework for consequential work like demurrage claims.
Field Notes 02 · August 30, 2026
Agent vs Copilot
Copilot, agent, AI-native, and agent-native are not interchangeable — and they need different ROI models.
Field Notes 03 · August 30, 2026
FDE or not?
Four hats for startup FDEs in 2026 — and why you need an agent-native software factory underneath every deployment.
Field Notes 04 · August 30, 2026
Token cost: $/1M tokens? No!
The lazy answer is dollars per million tokens. Different AI workflows need different KPIs — and control while running is not the same as reporting after the outcome.
Field Notes 04 · August 30, 2026
Gateway or Not?
Gateways tag model calls. Long-horizon agents need durable executions with budget, lineage, and cost attached to the outcome.
Field Note 5 · August 29, 2026
To Agent or Not?
If you can specify the procedure, build software. If the system needs judgment to determine what the procedure should be, use an agent.
