Docs

Reliable by Design · 05.07

ASSESS & REFLECT: How the Loop Grades and Corrects Itself

A graded 0..1 progress signal for stall detection, and a bounded self-critique pivot ladder for when a loop stalls.

ASSESSREFLECT

Four stagnation signals, not one

The stall detector is a pure, dependency-free function fed only an iteration's observable facts: a structural signature of its output and a progress trajectory. Four signals can vote a loop stalled:

SignalFires whenRequires
structural_repeatThe exact same output signature repeats across a window of consecutive iterations.Always on.
oscillationA signature seen before reappears, but not on the immediately prior iteration (A → B → A).Assess mode.
plateauProgress stays essentially flat (within 0.02) across the window.Graded done-check.
regressionProgress drops by more than 0.05 versus the previous iteration.Graded done-check.

Plateau and regression only make sense with a graded (continuous) signal — a plain pass/fail done-check has no meaningful "distance to goal" to plateau or regress along, so those two stay off for it, avoiding false stalls on a deterministic check.

REFLECT: pivot, don't quit

When the loop is judged stalled, REFLECT feeds a self-critique of what's not working back into the next iteration's context and changes tack — up to a bounded number of pivots. This is the difference between a loop that silently keeps repeating a failing approach until it burns its whole budget, and one that notices it's stuck and tries something genuinely different before conceding.

Continue