/field-notes
Field notes
from the cycle.
Essays on agent-first development, the 720-hour loop, and the slow rebuild of the software-services economy. Mostly written between cycles, never ghostwritten.
- 06 engineeringJun 06, 2026
MCP is the right contract for agent tools
Tool calling done as ad-hoc per-app function definitions is a fragility tax. Model Context Protocol is the contract layer the agent-product economy needed. Here is why we default to it, and the one case where we don't.
- 05 manifestoJun 02, 2026
Why 720 hours is the right delivery window
Six months is a hedge, not an estimate. Two weeks is bravado. Thirty days is the only window that respects both founder runway and engineering reality.
- 04 manifestoMay 30, 2026
The refusal layer is the product
Most agent demos fail at refusals, not at execution. For agents that touch money, health, or sensitive data, what the agent will not do is a sharper product surface than what it will. A note on building refusal-first.
- 03 engineeringMay 28, 2026
The agent stack we use to ship in 30 days
Six agents, one human-shaped loop. Scope, build, QA, deploy, monitor, ops — how each one earns its keep and where we deliberately don't trust them.
- 02 engineeringMay 24, 2026
Why we build the eval harness before we build the agent
The standard order is: ship the agent, then evaluate it. The right order, for any agent product that will face real users, is the opposite. A note on eval-driven development and why it is the agent equivalent of test-first.
- 01 manifestoMay 20, 2026
What 'agent-first development' actually means
It's not 'we use Copilot.' It's not 'we use Cursor.' Agent-first means the agents own the delivery loop, and the humans own the judgement calls — not the other way around.