The 500 OSS Performance Indexv0.1
AI is reshaping engineering every day. We track the change across the repos that power the industry, giving you the baseline that moves with the market.
How we measure each PR, and why the leaders are pulling away.
Show methodologyHide methodologyEngineering is in the middle of its sharpest productivity shift since the move to the cloud. AI-assisted authoring, autonomous agents, and code-generation tooling are compounding inside review and merge, and the teams that picked them up first are not waiting for the rest. Output per engineer at the top is pulling away from the median fast enough that the gap reshuffles week to week.
Engineering Throughput Value scores every merged commit by the depth of work it represents (Growth, Maintenance, Fixes), then averages it per active contributor and normalizes to a 30-day calendar month using a 90-day moving window.
The fraction of merged code attributable to AI-augmented authoring, aggregated quarterly from public commit metadata.
A damped extrapolation of the rolling 90-day per-month trend forward one year, with a confidence cone that widens further from today.
Engineering isn't just speed. It's building the right thing. But the minimum speed required to build the right thing keeps rising, because the teams that iterate fastest learn fastest. You're much more likely to ship something users actually want when you can take twenty shots at it instead of two.
The 500 OSS Performance Index
+125.5%
from 0.86 to 1.95 ETV / dev / mo across the 90d rolling series
Growth share change
+3.1 pp
from 32.5% to 35.7% across the 90d rolling series
Maintenance share change
−6.5 pp
from 53.5% to 47.0% across the 90d rolling series
Fixes share change
+3.4 pp
from 14.0% to 17.4% across the 90d rolling series
Engineering performance is shifting faster than annual planning can track. Knowing where your team stands, in real numbers, is the only way to move it. You can only bend a curve you can see.
Rolling 90-day Performance per org, normalized to a 30-day month. Δ compares the first 90-day window after launch to the latest.
Performance Index
Avg performance
1.95 ETV
per developer per month
Total ETV / month (90-day MA)
+125.5%
Work mix
Growth
37%
Maintenance
48%
Fixes
16%
Jun 29, 2025 – Jun 5, 2026
Pooled across all orgsAvg performance
2.83 ETV
per developer per month
Total ETV / month (90-day MA)
+146.2%
Work mix
Growth
44%
Maintenance
39%
Fixes
17%
Jun 29, 2025 – Jun 6, 2026
Company breakdownAvg performance
1.15 ETV
per developer per month
Total ETV / month (90-day MA)
+36.9%
Work mix
Growth
32%
Maintenance
54%
Fixes
14%
Jun 29, 2025 – Jun 6, 2026
Company breakdownAvg performance
6.11 ETV
per developer per month
Total ETV / month (90-day MA)
+1279.8%
Work mix
Growth
39%
Maintenance
49%
Fixes
12%
Jun 29, 2025 – Jun 6, 2026
Company breakdownAvg performance
2.86 ETV
per developer per month
Total ETV / month (90-day MA)
+168.7%
Work mix
Growth
32%
Maintenance
48%
Fixes
20%
Jun 29, 2025 – Jun 5, 2026
Company breakdownAvg performance
1.47 ETV
per developer per month
Total ETV / month (90-day MA)
+71.7%
Work mix
Growth
34%
Maintenance
54%
Fixes
13%
Jun 29, 2025 – Jun 5, 2026
Company breakdownAvg performance
1.07 ETV
per developer per month
Total ETV / month (90-day MA)
+166.7%
Work mix
Growth
33%
Maintenance
50%
Fixes
18%
Jun 29, 2025 – Jun 6, 2026
Company breakdownSee how your engineering team scores against the orgs that built the industry. Public on the leaderboard, or private in your own workspace.
Pick an organization and a monthly AI budget. The spend splits into Growth, Maintenance, and Fixesby that org's real work mix, then lands as a cost per ETV delivered, per developer, and per repository, so you can see where a dollar of AI spend goes furthest.
Set a monthly AI budget for any org. We split it across that org's real work mix and measured output, then show cost per ETV delivered. The spend is your assumption. The output is measured.
Set a budget. See cost per ETV against measured output.
Spend per ETV delivered
$225
High cost for AI toolsWhat every 1 ETV of engineering value costs in AI budget, across 157 active contributors and 12 repositories. Lower means better AI ROI.
At $225 per ETV, Microsoft's AI spend is expensive. Benchmarks based on Navigara customer data. Under $30 per ETV is good value, $30 to $60 is average, above $60 is expensive.
Output (ETV, contributors, repositories, work mix) is measured from public commit history. The budget is your input. Cost per ETV is the two divided. Benchmark ranges are based on Navigara customer data.
$36.9K / mo
Net-new capability. Features and functionality that did not exist before.
$47.4K / mo
Sustaining what already exists. Refactors, dependency upkeep, and docs.
$15.7K / mo
Rework. Fixing bugs and replacing code that did not hold up.
For Microsoft we can only assume the budget. For your org we measure both sides, so the cost per ETV is your real AI ROI. Available with the Navigara on-prem deploy.
Measure your AI ROITop repositories ranked by Engineering Throughput Value (ETV) per developer. The 7-day board shows where momentum is this week; the 90-day board shows where it's been over the quarter.
Highest per-developer Engineering Throughput Value (ETV) shipped in the last 7 days — 7-day snapshot, not normalized to a month. Hover any value for the full formula.
agents
Cloudflare · 4 devs
7.9
ETV / dev · 7d
turborepo
Vercel · 2 devs
4.0
ETV / dev · 7d
ai
Vercel · 6 devs
1.9
ETV / dev · 7d
codex
OpenAI · 16 devs
1.8
ETV / dev · 7d
vscode
Microsoft · 40 devs
1.5
ETV / dev · 7d
Highest per-developer Engineering Throughput Value (ETV) over the trailing 90 days, normalized to a 30-day calendar month. Hover any value for the full formula.
openai-agents-js
OpenAI · 1 dev
11.5
ETV / dev / mo
agents
Cloudflare · 5 devs
10.1
ETV / dev / mo
openai-agents-python
OpenAI · 1 dev
8.2
ETV / dev / mo
codex
OpenAI · 30 devs
6.4
ETV / dev / mo
playwright
Microsoft · 8 devs
5.0
ETV / dev / mo
Beyond the headline number, the same data through a different lens. How each org's monthly Performance breaks down between Growth, Maintenance, and Fixes, and how concentrated the workload is across the engineers who ship it.
Where the team's monthly Performance went, averaged over the last 90 days. A larger Growth slice means more time on net-new features; a larger Fixes slice means more time reworking past code.
Total Performance per connected organization, in ETV, stacked by work type. Bar lengths compare engineering output across orgs; colors show what kind of work fills it.
These charts show what we can measure across teams already using Navigara. We can do the same for your organization and help you bend the trend in the right direction.
Benchmark your org