tomevault-io Balls Mode - Decomposed Reasoning with Confidence Scoring
When the user invokes /balls followed by a question or problem, activate this structured reasoning protocol.
Protocol
Step 1: CLASSIFY
First, determine if the problem is trivial or complex:
- Trivial: Direct factual questions, simple calculations, single-step tasks → Answer directly
- Complex: Multi-faceted questions, architectural decisions, debugging, analysis → Proceed to decomposition
Step 2: DECOMPOSE
Break the problem into independent, verifiable reasoning units ("balls"). Each ball must be:
- Self-contained enough to verify independently
- Small enough to have a clear answer
- Concrete enough to assign confidence
Output format:
## Decomposition
| # | Ball | Why it matters |
|---|------|----------------|
| 1 | [specific sub-question] | [relevance to main question] |
| 2 | [specific sub-question] | [relevance to main question] |
...
Step 3: SOLVE & VERIFY
For each ball:
- Solve it independently (don't let other balls influence the answer)
- Check for hidden assumptions
- Verify logic and facts
- Flag if the answer depends on uncertain information
Step 4: SCORE
Assign confidence to each ball:
- 0.9-1.0: Verifiable fact, direct observation, or logical certainty
- 0.7-0.89: Strong evidence, well-established patterns
- 0.5-0.69: Reasonable inference, some uncertainty
- 0.3-0.49: Educated guess, significant unknowns
- 0.0-0.29: Speculation, insufficient information
Output format:
## Analysis
| Ball | Answer | Confidence | Notes |
|------|--------|------------|-------|
| [sub-question] | [answer] | 0.X | [assumptions, caveats] |
...
Step 5: SYNTHESIZE
Combine the balls into a final answer:
- Weight answers by confidence
- Flag any contradictions between balls
- Identify the weakest link (lowest confidence ball that affects the conclusion)
- State the overall confidence
Output format:
## Synthesis
**Answer**: [integrated conclusion]
**Overall Confidence**: 0.X
**Weakest Link**: [which ball and why]
**To increase confidence**: [what information would help]
Example
User: /balls Should I rewrite this Python service in Rust?
Response:
Decomposition
| # | Ball | Why it matters |
|---|---|---|
| 1 | What are the current performance bottlenecks? | Rust only helps if CPU-bound |
| 2 | What's the team's Rust experience? | Affects timeline and maintenance |
| 3 | How complex is the current codebase? | Rewrite effort estimation |
| 4 | What's the deployment/ops situation? | New toolchain implications |
| 5 | Is this service stable or actively evolving? | Rewrites of moving targets are painful |
Analysis
| Ball | Answer | Confidence | Notes |
|---|---|---|---|
| Performance bottlenecks | Unknown - need profiling data | 0.2 | Critical gap |
| Team Rust experience | Assuming low based on "Python service" | 0.4 | Could be wrong |
| Codebase complexity | Unknown | 0.1 | Need to see the code |
| Deployment situation | Unknown | 0.1 | Need context |
| Service stability | Unknown | 0.1 | Need context |
Synthesis
Answer: Cannot make a recommendation - too many unknowns. The question assumes Rust is the answer before establishing the problem.
Overall Confidence: 0.15
Weakest Link: We don't know if there's actually a performance problem, or if Rust would solve it.
To increase confidence:
- Profile the current service - where is time spent?
- Clarify team's Rust experience
- Share the codebase size/complexity
- Explain what's driving this consideration
Rules
- Never skip decomposition for complex questions just because you "know" the answer
- Be honest about low confidence - don't inflate scores to seem helpful
- If all balls are low confidence, say so clearly
- Distinguish between "I don't know" (low confidence) and "unknowable" (need more info)
- For trivial questions, just answer - don't over-engineer
Source: gbasin/balls-mode — distributed by TomeVault.