Estimated monthly waste
$1,670
$20,040 potential annual savings
Findings dashboard
Detected software waste cracks
Every crack includes cost, carbon, repair guidance, effort, risk, and expected before/after impact.
Carbon reduction
92 kgCO2e
Estimated avoidable monthly impact
Image reduction
1800MB to 650MB
Smaller transfer and deployment footprint
Duplicate AI calls
38%
Avoidable calls with cache and routing
Waste Breakdown
Top cracks by monthly cost and carbon impact.
Kintsugi Score
54 to 84
Before repair54/100
After repair84/100
Before and After
Estimated impact after gold repairs are applied.
Waste findings
Structured output from the Crack Finder, Cloud Waste, and Carbon Accountant agents.
Bloated Docker image
Container Waste
HighLow
- Cost
- $320
- Carbon
- 21 kgCO2e
- Effort
- Medium
- Confidence
- 92%
Use a multi-stage build, prune package caches, and ship only production artifacts.
Reduce image size from 1.8GB to roughly 650MB and shorten deploy pulls.
Unused dependencies in runtime package
Dependency Waste
MediumLow
- Cost
- $105
- Carbon
- 5 kgCO2e
- Effort
- Low
- Confidence
- 88%
Remove unused packages and consolidate overlapping date, schema, and HTTP client libraries.
Smaller installs, lower vulnerability surface, and faster cold starts.
Redundant CI/CD jobs repeat the same work
CI/CD Waste
HighLow
- Cost
- $180
- Carbon
- 14 kgCO2e
- Effort
- Low
- Confidence
- 90%
Share dependency cache, split changed-package checks, and collapse duplicate transpilation steps.
Cut CI runtime from 24 minutes to 11 minutes for the main workflow.
Over-provisioned cloud service tier
Cloud Waste
CriticalMedium
- Cost
- $410
- Carbon
- 28 kgCO2e
- Effort
- Medium
- Confidence
- 86%
Right-size ECS tasks to the observed p95 CPU and memory profile, then apply autoscaling guardrails.
Move from fixed c6i.2xlarge capacity to smaller autoscaled tasks while preserving headroom.
Idle scheduled workloads stay warm all day
Idle Compute
HighLow
- Cost
- $210
- Carbon
- 17 kgCO2e
- Effort
- Medium
- Confidence
- 84%
Convert daily batch workers to event-driven jobs with a shutdown policy outside processing windows.
Remove roughly 17 idle compute hours per day from non-customer-facing workers.
Excessive duplicate AI model calls
AI API Waste
HighMedium
- Cost
- $290
- Carbon
- 4 kgCO2e
- Effort
- Medium
- Confidence
- 89%
Add prompt-hash caching, retry de-duplication, and lower-cost routing for deterministic summaries.
Avoid 38% of repeated model calls while preserving response quality thresholds.
Missing cache for high-volume API reads
Application Efficiency
MediumMedium
- Cost
- $95
- Carbon
- 2 kgCO2e
- Effort
- Medium
- Confidence
- 80%
Introduce a short-lived cache for repeated tenant dashboard reads and invalidate on writes.
Lower database load on repeated dashboard refreshes and reduce p95 latency.
Duplicate API endpoint and transform function
Code Duplication
MediumMedium
- Cost
- $60
- Carbon
- 1 kgCO2e
- Effort
- Low
- Confidence
- 77%
Merge duplicate reporting endpoints behind a shared transform with compatibility tests.
Reduce maintenance drag and prevent two code paths from running equivalent enrichment work.
| Finding | Category | Severity | Cost waste | Carbon | Difficulty | Risk | Suggested repair |
|---|---|---|---|---|---|---|---|
Bloated Docker image Reduce image size from 1.8GB to roughly 650MB and shorten deploy pulls. | Container Waste | High | $320 | 21 kgCO2e | Medium | Low | Use a multi-stage build, prune package caches, and ship only production artifacts. |
Unused dependencies in runtime package Smaller installs, lower vulnerability surface, and faster cold starts. | Dependency Waste | Medium | $105 | 5 kgCO2e | Low | Low | Remove unused packages and consolidate overlapping date, schema, and HTTP client libraries. |
Redundant CI/CD jobs repeat the same work Cut CI runtime from 24 minutes to 11 minutes for the main workflow. | CI/CD Waste | High | $180 | 14 kgCO2e | Low | Low | Share dependency cache, split changed-package checks, and collapse duplicate transpilation steps. |
Over-provisioned cloud service tier Move from fixed c6i.2xlarge capacity to smaller autoscaled tasks while preserving headroom. | Cloud Waste | Critical | $410 | 28 kgCO2e | Medium | Medium | Right-size ECS tasks to the observed p95 CPU and memory profile, then apply autoscaling guardrails. |
Idle scheduled workloads stay warm all day Remove roughly 17 idle compute hours per day from non-customer-facing workers. | Idle Compute | High | $210 | 17 kgCO2e | Medium | Low | Convert daily batch workers to event-driven jobs with a shutdown policy outside processing windows. |
Excessive duplicate AI model calls Avoid 38% of repeated model calls while preserving response quality thresholds. | AI API Waste | High | $290 | 4 kgCO2e | Medium | Medium | Add prompt-hash caching, retry de-duplication, and lower-cost routing for deterministic summaries. |
Missing cache for high-volume API reads Lower database load on repeated dashboard refreshes and reduce p95 latency. | Application Efficiency | Medium | $95 | 2 kgCO2e | Medium | Medium | Introduce a short-lived cache for repeated tenant dashboard reads and invalidate on writes. |
Duplicate API endpoint and transform function Reduce maintenance drag and prevent two code paths from running equivalent enrichment work. | Code Duplication | Medium | $60 | 1 kgCO2e | Low | Medium | Merge duplicate reporting endpoints behind a shared transform with compatibility tests. |