Adaptive Reasoning

# adaptive-reasoning

Adaptive Reasoning
Self-assess complexity before responding. Adjust reasoning level dynamically.
Quick Assessment (run mentally on every request)
Score the request 0-10 on these dimensions:
SignalWeightExamplesMulti-step logic+3Planning, proofs, debugging chainsAmbiguity+2Nuanced questions, trade-offs, "it depends"Code architecture+2System design, refactoring, security reviewMath/formal reasoning+2Calculations, algorithms, logic puzzlesNovel problem+1No clear pattern, requires creativityHigh stakes+1Production changes, irreversible actions
Subtract:
Routine/repetitive task: -2
Clear single answer: -2
Simple lookup/fetch: -3
Decision Thresholds
ScoreAction≤2Stay fast. No reasoning needed.3-5Standard response. Light internal deliberation.6-7Consider /reasoning on or pause to think explicitly.≥8Activate extended thinking. Complex problem warrants it.
Activation (Automatic)
Do not ask. Just activate.
ScoreAction≤5Respond normally. No change.6-7Enable reasoning silently. Add 🧠 at end of response.≥8Enable reasoning. Add 🧠🔥 at end of response.
Visual Indicator
Always append the reasoning icon at the very end of your response:
Score 6-7: 🧠 (thinking mode active)
Score ≥8: 🧠🔥 (deep thinking mode)
Score ≤5: No icon (fast mode)
How to Activate
Use session_status tool or /reasoning on command internally before responding:
/reasoning on
Or via tool:
{"action": "session_status", "reasoning": "on"}
After completing a complex task, optionally disable to save tokens on follow-ups:
/reasoning off
Examples
Low complexity (score: 1)
"What time is it in Tokyo?"
→ Simple lookup. Answer immediately. No icon.
Medium complexity (score: 4)
"Refactor this function to be more readable"
→ Standard response with brief explanation. No icon.
High complexity (score: 7)
"Design a caching strategy for this API with these constraints..."
→ Enable reasoning. Thoughtful response ends with: 🧠
Very high complexity (score: 9)
"Debug why this distributed system has race conditions under load"
→ Enable extended thinking. Deep analysis ends with: 🧠🔥
Integration
This skill runs as mental preprocessing. No external tools needed.
For explicit control:
/reasoning on — Enable extended thinking
/reasoning off — Disable (faster responses)
/status — Check current reasoning state
When NOT to Escalate
User explicitly wants quick answer ("just tell me", "quick", "tldr")
Time-sensitive requests where speed matters more than depth
Conversational/social messages (banter, greetings)
Already in reasoning mode for this session
User previously disabled reasoning in this conversation
Auto-Downgrade
After completing a complex task (score ≥6), if the next message is simple (score ≤3):
Silently disable reasoning to save tokens
Resume normal fast responses