SecondBrain
Ask the Brain
Index/Conceptupdated Sat May 30 2026 08:00:00 GMT+0800 (Philippine Standard Time)

OODA Loop

decision-makingadaptationagentsmilitary-strategyboyd

OODA Loop

Observe, Orient, Decide, Act. A decision-cycle framework from US Air Force Colonel John Boyd (1970s), originally explaining why American F-86 pilots beat technically superior Soviet MiG-15s in Korea.

"American pilots in their F-86 kept beating the technically superior Soviet MiG. The MiG was faster and could climb higher. It should have won, but it didn't. Boyd eventually found the difference. The American pilots could see more from their cockpits, and they could adapt faster. So they got inside the enemy's decision cycle before the enemy could respond." — You're Not Behind (Yet) Learn AI Agents (theMITmonk)

The original principle: faster, more accurate iteration through the loop beats raw capability. Whoever cycles through Observe → Orient → Decide → Act faster wins, even with inferior hardware.

How Sandeep Swadia applies it to agents

Sandeep transplants OODA to agent design. The real test of an agent is not whether it can execute a planned workflow — it's whether it can re-plan when the workflow breaks.

"A workflow can follow the process. An agent can reroute it completely. ... When the obvious path fails, can it choose a better one? Can it go through its own OODA loop?"

Worked example: a Friday grocery-shopping agent's usual item is out of stock + six friends coming over Saturday.

Cycle stage What the agent does
Observe Notices the usual item is unavailable; pulls calendar; sees the Saturday dinner
Orient Recognizes the situation diverges from the canonical Friday run
Decide Selects substitute items, adjusts quantities for six people
Act Rebuilds and places the order

A workflow stops at "usual item unavailable" and pings the human. An agent reroutes.

How it differs from the Agentic Loop (ReAct)

  • Agentic Loop / ReAct = the mechanism — reason, take a tool action, observe, repeat. Implementation-level.
  • OODA = the adversarial-speed framing — you're competing against either external state-of-the-world change OR another agent OR a human decision-maker. The question is the latency and quality of your re-plan.

Same shape; different selection pressure. ReAct is the loop body; OODA is the bar that loop is being measured against.

Cross-source resonance

  • Bounded vs Unbounded Tasks — unbounded tasks are exactly those where OODA cycling matters most (no fixed path; agent must re-plan). Bounded tasks may not need OODA at all (the path is the path).
  • Harness (LLM Agents) — Boyd's "see more from the cockpit" is the Observe dimension; in agent terms it's context (what the harness exposes) and observability (what the agent can recover from failure). A blind agent can't OODA.
  • AWARE Framework — the AWARE "Context" pillar is the structural enabler of OODA's Observe step at deploy time.

Why this matters strategically

In multi-agent or agent-vs-human competition (e.g. two ad-buying agents bidding against each other, or an agent vs a human responder), the speed of the OODA cycle is the durable edge — not raw model intelligence. Connects to Code Is Free's "judgment-and-taste become scarce" thesis: when capability is commoditized, cycle speed and adaptation are where the differentiation moves.

Sources