| name | triz-analysis |
| description | Generic TRIZ analysis for identifying technical and physical contradictions in software architecture and pipelines. |
When to use
Use during the initial phase of any major architectural change, feature implementation, or optimization task. TRIZ analysis helps surface fundamental tradeoffs (contradictions) that might be overlooked.
Core Process
- Identify the Improving Parameter: What are we trying to improve? (e.g., speed, accuracy, modularity).
- Identify the Worsening Parameter: What gets worse if we improve the first parameter? (e.g., CPU usage, code complexity, latency).
- Formulate the Technical Contradiction: If we increase [Improving Parameter], then [Worsening Parameter] becomes unacceptable.
- Formulate the Physical Contradiction: A single element must have two opposite properties (e.g., the buffer must be large for accuracy but small for latency).
Project-Specific Contradiction Examples
1. VAD Throughput vs. False Positives
- Improving: VAD throughput (frames/sec).
- Worsening: False positives (misidentified speech).
- Contradiction: Faster processing often requires simpler heuristics or larger windows, which can lead to missing short speech segments or misclassifying noise.
2. Review Quality vs. Pipeline Latency
- Improving: Review quality (interactive features, high-res spectral analysis).
- Worsening: Pipeline latency (time to first report).
- Contradiction: Higher quality reviews require more pre-computation and data extraction, delaying the point at which the user can start reviewing.
3. Deterministic Output vs. Self-Learning Calibration
- Improving: Deterministic pipeline output (reproducibility).
- Worsening: Self-learning calibration (adaptive thresholds).
- Contradiction: For reproducibility, we want fixed parameters. For better performance across diverse media, we want the system to learn and adapt its thresholds dynamically.
4. Modularity vs. CI/Runtime Complexity
- Improving: Modularity (separating logic into multiple crates or fine-grained modules).
- Worsening: CI/Runtime complexity (longer build times, complex dependency graph).
- Contradiction: More modules improve code organization and reusability but increase the overhead of managing dependencies, compilation times, and cross-module interfaces.
Output
All TRIZ analysis findings must be written to analysis/triz/ with a descriptive filename (e.g., analysis/triz/feature-x-contradictions.md).
Related Skills
- triz-solver: For resolving the identified contradictions.
- triz-audio-timeline: Domain-specific TRIZ for audio segmentation.