Skip to content

Metrics

Marty McEnroe edited this page Feb 5, 2026 · 1 revision

AgentOS Metrics

"Taxation, gentlemen, is very much like dairy farming. The task is to extract the maximum amount of milk with the minimum amount of moo." — Lord Vetinari


Executive Summary

Metric Value
Total Issues Created 207
Total Issues Closed 159
Currently Open 48
Closure Rate 76.8%
Active Days 27 (Jan 10 - Feb 5)
Average Velocity 5.9 closed/day
Peak Day (Closes) 55 (Feb 3)
Peak Day (Opens) 55 (Feb 2)

Velocity: Issues Opened vs Closed

xychart-beta
    title "Daily Issue Velocity (Central Time)"
    x-axis ["1/10", "1/11", "1/14", "1/16", "1/17", "1/21", "1/23", "1/28", "1/29", "2/1", "2/2", "2/3", "2/4"]
    y-axis "Issues" 0 --> 60
    bar [2, 3, 3, 7, 10, 0, 6, 11, 12, 10, 55, 27, 44]
    line [0, 2, 3, 2, 8, 12, 1, 1, 4, 7, 23, 55, 31]
Loading

Legend: Bars = Opened, Line = Closed


Cumulative Progress

xychart-beta
    title "Cumulative Issues Over Time"
    x-axis ["1/10", "1/14", "1/17", "1/21", "1/28", "2/1", "2/2", "2/3", "2/4", "2/5"]
    y-axis "Total Issues" 0 --> 220
    line [2, 13, 31, 31, 56, 68, 123, 150, 194, 207]
    line [0, 5, 15, 27, 29, 36, 59, 114, 145, 159]
Loading

Legend: Top line = Total Created, Bottom line = Total Closed


The Burn Rate

Jan 10-17:  ████████░░░░░░░░░░░░  30 opened, 15 closed (warmup)
Jan 21-29:  ██████████░░░░░░░░░░  37 opened, 26 closed (steady state)
Feb 01-05:  ████████████████████ 136 opened, 118 closed (hyperdrive)

The inflection point: Feb 2, when 55 issues were created in a single day. This marked the transition from "building AgentOS" to "AgentOS building itself."


Velocity by Day of Week

Day Opened Closed Net
Friday 12 10 +2
Saturday 55 23 +32
Sunday 27 55 -28
Monday 44 31 +13
Tuesday 69 40 +29

Observation: Weekends are productive. Sunday Feb 3 closed 55 issues—more than most weeks.


Issue Lifecycle

pie title Issue Status Distribution
    "Closed" : 159
    "Open" : 48
Loading

Average time to close: Most issues close within 24 hours of creation. The governance workflow (LLD → Gemini → Implementation → PR) typically completes in 2-4 hours for straightforward features.


Cross-Project Usage

AgentOS is used by multiple repositories:

Repository Status Issues Processed
AgentOS Active 207
RCA-PDF Active TBD
dispatch Planned -

Note: Cross-project metrics aggregation is tracked in #329


Future Metrics (Roadmap)

Governance Metrics

Metric Description Status
LLD First-Pass Rate % of LLDs approved on first Gemini review Planned
Revision Count Average revisions before approval Planned
Gate Time Time spent in each governance gate Planned
Block Reasons Categorized reasons for BLOCK verdicts Planned

Cost Metrics

Metric Description Status
Gemini API Cost Daily/weekly Gemini usage Planned
Claude API Cost Token usage by workflow Planned
Cost per Issue Total cost / issues closed Planned

Quality Metrics

Metric Description Status
Test Coverage Aggregate coverage across modules Planned
Regression Rate New failures per release Planned
Reopen Rate % of issues reopened after close Planned

Agent Metrics

Metric Description Status
Concurrent Agents Peak simultaneous agent count Planned
Agent Success Rate % of agent tasks completing Planned
Permission Friction Prompts per hour by agent Logged

The Vetinari Index

"The Patrician moved a small marker on his wall map."

Lord Vetinari tracks everything. The Vetinari Index is our composite health score:

Vetinari Index = (Closure Rate × 0.3) +
                 (First-Pass LLD Rate × 0.3) +
                 (Test Coverage × 0.2) +
                 (Agent Success Rate × 0.2)

Current Index: 0.77 × 0.3 + (estimated 0.65 × 0.3) + (0.80 × 0.2) + (0.90 × 0.2) = 0.77

A Vetinari Index above 0.75 indicates a well-functioning city. Above 0.85, the Patrician permits himself a thin smile.


The Clacks Overhead

Every metric travels the Clacks. Every issue, every verdict, every commit—passed from tower to tower, never forgotten.

GNU Terry Pratchett
GNU AgentOS Contributors

Historical Notes

The Great Audit of Feb 2

On February 2nd, 2026, a comprehensive audit identified 55 issues in a single day:

  • Orphaned tests
  • Mock-heavy test suites
  • Unimplemented flags
  • Silent failures

This wasn't a crisis—it was visibility. The audit proved the governance system works: it found problems before users did.

The Sunday Sprint

February 3rd saw 55 issues closed—a testament to the multi-agent architecture. While The Great God Om slept (briefly), agents continued their work.


Related


"The city worked. It was a kind of miracle, really. One of Vetinari's miracles." — Terry Pratchett, Making Money

Clone this wiki locally