Principles of Outcome Engineering

Outcome as the new unit of work. We as humans provide a statement of what should remain true at the end. “Build this Android app” is a goal, “Build an Android version of our iOS app and keep it at parity forever” is an outcome.

Machinery should be invisible. Skills, Loops, /goal, automations, verification, memory, plugins, budgets, connectors, and subagents all matter, but almost none of it should be our responsibility. We should bring an outcome, the system should manage / abstract away everything else.

Human attention is for decisions. People should step in only for judgment calls a human should make. The system should be smart enough to surface these and notify us when an answer is needed.

“Done” is a state. The system should be responsible for the work for as long as we depend on the outcome.

Ambition is unbound. The most challenging part of a big idea was never starting it, it was finishing it and maintaining it forever. Now that our systems can do both we can be more ambitious.

Experimentation as a first principle. Every open question becomes something we can just test. A wrong guess is no longer an opportunity cost because the system can be wrong N number of times on the way to being right (and in parallel).

Outcome engineering is where we fully give in to our imagination, embrace ambition, and forget that the underlying system and machinery exist.

This mental model is already being embraced by some of the best engineers I know and I think will define Software Engineering in 2027.