Feedback Loops

Closed-loop feedback control as a unifying pattern across engineering, software, and management. The trend toward shorter cycles and the implications of AI-speed feedback.

Control theory foundations

• Control Theory (Wikipedia) — overview of the field: Maxwell → PID → cybernetics → modern control (bib)
• Open-Loop vs Closed-Loop Systems — the fundamental distinction: does the system re-check actual state?
• Stability in Feedback Systems — overshoot, oscillation, settling time, steady-state error
• PID Control — the canonical controller: proportional + integral + derivative
• Controllability and Observability — prerequisites: can you steer it? can you see it?
• Robustness in Control Systems — handling model imprecision and disturbance
• 2039567024205267136 — closed-loop feedback control: measure, compare, actuate
• 2039566010131558818 — Kubernetes reconciliation loop as a specific implementation
• 2039567945400254882 — Kubernetes’ loop is Watt’s governor (1788) applied to software
• 2039613059178873301 — dead band: the range where the controller deliberately does nothing

Desired state in software

• Controller Pattern — watch, diff, reconcile: the generic loop
• Kubernetes Controllers — K8s implementation: many simple controllers via API server (bib)
• Perpetual Disequilibrium — the system never settles, and that’s correct
• 2039568269989036219 — desired state systems = feedback control: sensor is observation, comparator is diff, actuator is API call
• 2039569197378965946 — Kubernetes standardized this: CRDs + controllers for any desired state
• 2039302841450537164 — desired state wraps mutable with declarative (Jenco)
• 2039308018211586082 — NixOS as a particular case of desired state
• 2039290956810448985 — NixOS: describe desired, compute path to it

Management as feedback control

• Goal-Setting Theory — specific difficult goals outperform vague goals; four mechanisms, three moderators (bib)
• Self-Efficacy and Goals — Bandura’s self-efficacy as the central mediating variable
• Discrepancy Production vs Reduction — feed-forward (setting harder goals) vs feedback (closing the gap)
• Learning vs Performance Goals — on complex tasks, learning goals outperform performance goals
• Locke & Latham (2002) — source summary: 35-year retrospective on goal-setting theory (bib)
• 2039578081070137359 — Locke & Latham: goal-setting as feedback control (but see discrepancy production for the nuance)
• 2039578424977850584 — Locke & Latham answered the management question in 1990
• 2039578726334394790 — they explicitly cite cybernetic engineering as their model
• 2039579041183998043 — but in practice the management loop is open: quarterly checks, manual action
• 2039579188970365285 — AI-native companies can close this loop to seconds

Cycle frequency in management

• OKR — quarterly goal-setting framework: objective + measurable key results (bib)
• 2039590834027508118 — management has been shortening the feedback cycle for decades
• 2039585611682918581 — OKR: 90-day feedback loop
• 2039585794495803405 — OKR from control theory: a thermostat that checks once a season
• Scrum — three nested feedback loops: daily, sprint review, retrospective (bib)
• 2039587182000959844 — Scrum: three nested inspect-and-adapt loops
• 2039588170195710144 — Scrum implements empirical process control
• Kanban Method — continuous flow, WIP limits, pull system, no fixed iterations (bib)
• 2039590758597251341 — Kanban: continuous flow, no fixed iterations
• 2039596911204966407 — Kanban is the most radical step: cycle time = single work item
• 2039597131431047679 — AI-native companies: management feedback in seconds
• 2039597531123032444 — companies running 90-day loops cannot compete with 90-second ones

Empirical process control

• Empirical Process Control — defined vs empirical; three pillars: transparency, inspection, adaptation (bib)
• 2039587798358090117 — execute small, get feedback, adapt (transparency, inspection, adaptation)
• 2039588285211885798 — empirical process control = feedback control applied to work processes

Iteration & tinkering

• 2039604195364647207 — reduce everything to cycles: hypothesis → action → feedback → adjust
• 2039604487682486628 — cycle frequency matters more than step quality
• 2039604691202662534 — re-strategize every few hours; weekly is already too slow
• 2039604827538550874 — the cycle applies to agents, products, companies, and life
• 2039603982247870958 — Taleb: tinkering beats planning in complex systems
• 2039606964775444504 — convex tinkering: small errors, large potential gains
• 2039607159978328172 — hypothesis → action → feedback = Scrum’s empirical process control
• 2039607374038880401 — shortening cycles buys more attempts per unit of time

Related MOCs

• Moc Continuous Assembly — feedback control applied to physical assembly
• Moc AI Agents — AI agents as the fast comparator/actuator closing the loop
• Moc Product and Business — iteration and risk in business context
• Cybersecurity — AI compresses the security feedback loop; near-zero cost creates motivation