Quadratique Engineering Discipline for AI Agents
Course / M3
Module 3 of 12

Feedback Loops

The sensor net that makes delegation safe

Build the ladder of checks that lets agents correct their work and lets you accept code you did not read line by line: unit tests, acceptance tests, gates, telemetry, and human review.

8 sections

Start this module →

  1. 01

    How you know you're making progress

    ✎ Text

    Feedback loops are the mechanism of delegation itself.
  2. 02

    The loops form a ladder, by speed

    ✎ Text

    A full feedback ladder catches each kind of failure at the fastest rung that can catch it.
  3. 03

    Tests are critics, and they criticise different things

    ✎ Text

    Unit tests criticise code. Acceptance tests criticise the product.
  4. 04

    The anatomy of a good unit test

    ✎ Text

    A good test has one behaviour, a business-intent name, and assertions about behaviour rather than implementation.
  5. 05

    Test the core through the command handler — and keep it fast

    ✎ Text

    Exercise behaviour through the natural entry point, fake the ports, and never reach for internals.
  6. 06

    You own the failure model — agents don't grade their own homework

    ✎ Text

    The most valuable tests come from a human-owned model of what must never happen.
  7. 07

    The other loops, by what each catches

    ✎ Text

    Each rung catches a different class of failure, and not every rung has the same weight.
  8. 08

    The payoff, and the seed forward

    ✎ Text

    A sensor net lets agents correct themselves and lets you trust code you did not read line by line.