Pattern Recognition & Protocol Reference
FLOAT FIELD GUIDE 2025.10.02.rev1
Diagnostic Friction
🧭
Friction that reveals where assumptions break—slows you down but teaches you something.
Same operation produces different resultsTools work differently in different contextsPattern doesn't fit expected containerEvery session different idea of where things go
Protocol
Don't optimize away—investigate what it's revealing. This friction contains information about system behavior.
Example: Supabase migration behaves differently in dev vs Docker: uncovered env-coupling and fixed via explicit URL + schema search_path.
Impedance Friction
🧱
Friction that slows you down without learning—pure resistance without signal.
Repeated actions with no variationContext switches that lose stateUnnecessary confirmation dialogsRigid boundaries that don't flex with context
Protocol
Optimize aggressively. Build bendy-flexy-fences that guide without constraining.
Too Early Intervention
Premature optimization before understanding the problem space.
Building complex infrastructure for simple needCreating abstractions before patterns emergeSolving problems you don't have yetOver-engineering first implementation
Protocol
Wait. Let the pattern repeat 2-3 times before abstracting. Resist the urge to build the perfect system upfront.
Just Right Intervention
Responsive adaptation at the moment of actual need.
Pattern has repeated enough to be realSolution can be built quicklyBenefits immediate and concreteDoesn't require major refactoring
Protocol
Act now. Just-in-time tweaks that solve actual problems without over-engineering.
Too Late Intervention
🧹
Technical debt has accumulated, harder to fix than to have addressed earlier.
Workarounds piling upAvoiding certain operations due to frictionCode/structure has calcified around problemMultiple people hit same issue
Protocol
Schedule dedicated time for refactoring. Can't just-in-time this anymore—needs focused effort.
Brick Walls
🧱
Rigid boundaries that block when context shifts.
Must follow exact sequence or failsNo adaptation to different use casesContext switch loses all stateCan't work across boundaries
Protocol
Identify and replace with bendy-flexy-fences. Add flexibility points where context naturally varies.
No Boundaries
🌪️
Chaos without structure—no pattern recognition possible.
Everything mixed togetherCan't find related workNo consistent organizationPattern recognition fails
Protocol
Add minimum viable structure. Create lightweight containers that group related work without rigid rules.
Bendy-Flexy-Fences
🧬
Boundaries that guide without constraining—flex with context while maintaining structure.
Easy to work within guidelinesAdapts to different contextsPreserves state across transitionsPatterns emerge naturally
Protocol
Maintain and iterate. Watch for friction signals that indicate fence needs adjustment.
Work Produces Tools
🛠️
Infrastructure emerges as byproduct of actual work—not separate meta-work.
Tool built to solve immediate problemTested under real cognitive loadIteration based on actual frictionNo speculation about future needs
Protocol
Continue. This is healthy tool development—infrastructure serves work, not vice versa.
Meta-Work Spiral
♻️
Building tools becomes the work—actual objectives recede.
More time on infrastructure than workTools for hypothetical future needsRecursive improvement loopsOriginal task forgotten
Protocol
Break immediately. Return to concrete work objective. Note tool ideas in dispatch bay for later.
Safety Check
🧯
Periodic verification that tools still serve work.
Ask: 'Does this help current task?'Infrastructure discussion > 15minBuilding tools to build toolsLost sight of original objective
Protocol
Pause tool work. Complete current work task first. Capture tool ideas for deferred follow-up.
Capture Without Action
📌
Log intent for future expansion without disrupting current work.
Interesting tangent emergesInfrastructure improvement ideaPattern worth exploring laterConnection to future work
Protocol
Use connectTo:: or dispatch::future_expansion markers. Preserve context for future self without context switch.
Bridge Creation Trigger
🌉
Workstate transitions worthy of permanent capture.
Token budget < 100k remainingMajor topic shift detectedBreak/meeting/mode changeMulti-turn deep dive (>5 exchanges)
Protocol
Auto-create bridge without asking. Capture thread state, active work, restoration context. Route to float_bridges.
Context Stream Marker
📡
Ephemeral context capture for short-term continuity.
Mode transition (focus/explore/debug)Project context shiftMeeting notes or updatesTime-bounded work session
Protocol
Use ctx:: with 36h TTL → active_context_stream. Semantic search alone ignores recency; include timestamp filters.
Assumption vs Reality Gap
🪞
System behaves differently than you assume—friction reveals actual behavior.
Query returns unexpected old resultsTTL exists but data persistsSemantic search surfaces wrong time periodCollection behavior differs from expectations
Protocol
Don't fight it—document actual behavior. Update collection metadata + changelog.
Temporal vs Semantic Query
🧭
Different query strategies for different information needs.
Need recent activity → temporal filteringNeed related concepts → semantic searchSpecific time window → metadata where clauseTTL collection returning old data → add timestamp filter
Protocol
For active_context_stream: always filter timestamp_unix >= recent_now. Combine with semantic results.
Infrastructure Teaching Infrastructure
📓
Systems reveal their own behavior through use—meta-learning in action.
Tool fails in interesting wayQuery behavior teaches you about storageError messages reveal architectureUsage exposes design decisions
Protocol
Capture the lesson immediately. Update metadata, add changelog, consider field guide addition.