The OODA loop stands for Observe, Orient, Decide, Act - a decision-making framework originally developed by military strategist John Boyd for combat operations, now widely applied to understand commercial operations, competitive strategy, and information behavior.
The Four Stages
Observe: Gather information from your environment. This involves collecting raw data, identifying relevant signals, and maintaining awareness of changing conditions. The observation phase connects to broader patterns of information seeking and environmental context awareness.
Orient: Analyze and synthesize the observations within your existing mental models, cultural traditions, and previous experiences. Orientation is the most critical phase - it’s where you make sense of what you’ve observed by filtering it through your understanding of the world. This relates closely to metacognition and how we apply our existing knowledge frameworks to new situations.
Decide: Choose a course of action based on your oriented understanding. This is where reversible decisions versus high-stakes choices become relevant - the decision phase should account for both the quality of your orientation and the reversibility of potential actions.
Act: Execute your decision and implement the chosen course of action. This creates new conditions in your environment, which feeds back into the observation phase, creating a continuous loop.
The Loop as Feedback System
The OODA loop is fundamentally a feedback system - your actions change the environment, which you then observe, creating an iterative cycle of learning and adaptation. The speed at which you can complete OODA cycles relative to competitors or changing conditions determines effectiveness.
Boyd’s key insight: Getting inside your opponent’s decision loop - cycling through OODA faster than they can - creates confusion and advantage. When your actions consistently outpace their ability to observe and orient, their decisions become based on outdated understanding.
This iterative, context-responsive approach mirrors Christopher Alexander’s concept of unfolding with context, where you “pay attention to the context, removing things that frustrated me, and expanding things that made me feel alive” through repeated cycles of observation and adjustment.
Eight Information Activities
The OODA loop encompasses eight distinct information-seeking behaviors that occur across its phases:
- Starting: Initial information gathering that kicks off the observation phase
- Chaining: Following initial findings to explore connected information
- Browsing: Casual, non-intentional searching that may reveal unexpected insights
- Differentiating: Grouping and categorizing information sources by relevance and reliability
- Monitoring: Tracking developments over time to detect patterns and changes
- Extracting: Pulling relevant materials from the information environment
- Verifying: Confirming accuracy and validity of gathered information
- Ending: Determining when sufficient information has been gathered to move to decision
These activities aren’t strictly sequential - they interweave throughout the observe and orient phases, supporting both networked exploration and systematic analysis.
Application to AI and LLMs
The OODA framework maps remarkably well to how language models can be prompted to solve complex problems through iterative “thought/act/observation” loops. Rather than treating an LLM as a single-shot question-answering system, we can structure interactions as repeated OODA cycles:
Observe: The model examines available information, searches documentation, or queries tools Orient: The model reasons about what it has found in relation to the problem Decide: The model determines what action to take next Act: The model executes a tool call, generates output, or requests more information
This approach transforms LLMs from fixed computation into potentially unbounded computation - each cycle can spawn new observations and decisions, similar to how multi-agent research systems use iterative exploration to solve complex problems.
Example Prompt Structure
“Thought: Let’s think step by step. I need to find out X and then do Y. Act: Search Wikipedia for X…”
This explicit structuring of the model’s process through observe-orient-decide-act phases enables more sophisticated problem-solving. The model can:
- Recognize when it lacks necessary information (observe)
- Reason about what information would be most useful (orient)
- Choose appropriate tools or approaches (decide)
- Execute and evaluate the results (act), then loop back
The pattern appears in various forms of prompt engineering, particularly chain-of-thought prompting and agent-based architectures. By making the OODA cycle explicit in prompts, we enable models to exhibit more adaptive, context-aware behavior rather than rigid, predetermined responses.
Relationship to Decision-Making and Agency
The OODA loop fundamentally relates to agency - particularly volitional agency, which involves “actively making decisions guided by internal desires” and continuously evaluating actions and their consequences. Each cycle through OODA represents an exercise of agency: gathering information, making sense of it through your values and understanding, choosing a path, and acting on that choice.
The framework emphasizes that effective decision-making isn’t about making perfect choices but about maintaining high decision velocity while incorporating feedback. This aligns with the principle that speed compounds when you manage downside - rapid OODA cycles let you learn faster, correct mistakes earlier, and adapt to changing conditions before they become critical.
Your milieu - the configuration of people, information flows, and ideas around you - fundamentally shapes each phase of your OODA loop. What you can observe, how you orient to that information, which decisions seem possible, and what actions you can take are all influenced by your environment. Changing your milieu changes your OODA capabilities.
Limitations and Considerations
Orientation Bias: The orient phase is vulnerable to cognitive biases and outdated mental models. If your framework for understanding the world is flawed, faster OODA cycles may just reinforce errors more quickly.
Information Overload: Too much observation can paralyze orientation and decision. The framework requires judicious filtering - knowing what information matters and what constitutes noise.
Context Dependency: What works in one domain (military tactics) may not transfer directly to another (business strategy, personal development). The speed and nature of feedback loops vary dramatically across contexts.
Over-Optimization for Speed: While cycling faster than competitors provides advantage, rushing through orientation and decision can lead to poorly considered actions. Balance speed with deliberate thinking when stakes are high.