The irreducible epistemic atoms underlying the curriculum. 4,828 atoms across 8 types and 2 molecules
Validate that leading indicators actually predict lagging outcomes by periodically checking whether movements in the leading metric were followed by corresponding movements in the outcome, replacing proxies that fail this test.
When reality feedback and social feedback diverge, treat the divergence itself as high-value signal requiring investigation rather than choosing one channel over the other.
Identify the structural components of emotional feedback loops (sensation, interpretation, behavior, consequence) and intervene at the weakest link rather than attempting to suppress the emotion directly.
Replace ruminative cognitive responses with behaviorally incompatible absorbing activities rather than attempting to think your way out of recursive negative thought patterns.
Eliminate safety behaviors and avoidance responses in anxiety loops to allow catastrophic interpretations to be disconfirmed through exposure, weakening the loop with each cycle.
Reframe the meaning of physiological arousal symptoms before the emotional loop gains momentum rather than attempting to regulate after the response has peaked.
Design immediate, concrete rewards for new habits to strengthen the dopamine prediction error signal before the long-term benefits arrive.
To break a destructive habit, replace the routine while keeping the same cue and reward structure intact.
Supplement natural rewards with external reinforcement during early habit formation, then remove scaffolding once the loop becomes self-sustaining.
When encountering information that perfectly confirms existing beliefs, deliberately search for what you would expect to see if that belief were false.
To break a vicious feedback loop, identify and disrupt the specific reinforcement mechanism rather than attacking the behavior directly through willpower.
Strengthen an existing positive feedback loop by reducing friction at its weakest link rather than building new loops from scratch.
Stack beneficial feedback loops by strengthening shared nodes where multiple loops converge to achieve compounding-on-compounding effects.
Verify the long-term trajectory of a reinforcing loop before investing in acceleration, as amplification is indifferent to whether outcomes are beneficial or harmful.
Before adding corrective action in a delayed-feedback system, inventory actions already taken that have not yet produced results to avoid pipeline overfilling.
Match evaluation frequency to the system's natural delay timescale - daily habits on days, fitness on weeks, career on quarters, investments on years.
Make smaller adjustments and wait longer between them when operating in delayed-feedback systems to avoid amplifying oscillations through overcorrection.
Separate feedback loops by timescale, tuning faster inner loops before slower outer loops, and prevent slow loops from interfering with fast loop corrections.
The emotional intensity of your resistance to a piece of feedback serves as a proxy for its proximity to a genuine blind spot in your self-model.
Feedback loops carrying the most valuable corrective information are precisely the loops you have unconsciously disabled through defensive routines, making their reconnection a prerequisite for double-loop learning.
Pre-specify action thresholds that trigger behavioral responses before collecting data, preventing post-hoc rationalization from neutralizing feedback signals when metrics cross into concerning ranges.
Treat major changes in goals, team composition, or system architecture as mandatory recalibration events that require re-examining every existing feedback loop rather than assuming old metrics still apply.
Design learning systems with separate single-loop feedback (detecting and correcting errors within existing frameworks) and double-loop feedback (questioning the framework itself when single-loop corrections fail repeatedly).
Structure antifragile feedback loops by exposing systems to hormetic doses of stress (enough to trigger adaptive response, not enough to overwhelm capacity) and implementing over-correction rather than simple restoration to baseline.