Navigara

How Top Engineering Teams Actually Perform

A live index of the industry's leading engineering orgs. Every commit is weighted by the work it represents, not the lines it touched, and reported as ETV per engineer.

+1,798engineers' worth of capacity,
nobody hired

2,371 engineers needed at the pre-AI baseline
Engineering capacity added per org
real engineerscapacity, not headcount1 mark = 10 engineers
Rows come to +1,862, the index reads +1,798. They are not meant to agree. Why?
Each org is measured against its own first reading and the index against its own, so the parts do not add up to the whole. The index also divides pooled output by the pooled engineer set, deduplicated across orgs and covering every team in the index, including ones without a public row.
Capacity added since the index's own first reading, measured from shipped output. How to read it
Every org row is measured against its own first reading, not a shared one. The figures come from shipped output, not from AI usage: teams also changed size, tooling and composition over this period.

How fast did it happen?

ETV per engineer, rolling 90 days. Each series starts from its own pre-AI baseline, so read a curve against where it started, not against its neighbours.

Show methodology

Where does each org stand?

Same measure, one card per org. Headcount is not the story, the multiple per engineer is.

OpenAI

Capacity vs pre-AI

19.2x

+491 engineers

14.45 ETV / dev / mo

Avg. dev performance / month (90-day MA)

+81.3%

Work mix

Features

26%

Maint

8%

Tests

41%

Docs

6%

Fixes

19%

Company breakdown

Microsoft

Capacity vs pre-AI

4.8x

+479 engineers

6.51 ETV / dev / mo

Avg. dev performance / month (90-day MA)

+100.5%

Work mix

Features

32%

Maint

11%

Tests

37%

Docs

3%

Fixes

17%

Company breakdown

Vercel

Capacity vs pre-AI

3.3x

+113 engineers

5.04 ETV / dev / mo

Avg. dev performance / month (90-day MA)

+56.5%

Work mix

Features

31%

Maint

16%

Tests

30%

Docs

7%

Fixes

16%

Company breakdown

Google

Capacity vs pre-AI

2.9x

+185 engineers

2.74 ETV / dev / mo

Avg. dev performance / month (90-day MA)

+29.4%

Work mix

Features

31%

Maint

15%

Tests

33%

Docs

6%

Fixes

15%

Company breakdown

Meta

Capacity vs pre-AI

2.7x

+303 engineers

1.94 ETV / dev / mo

Avg. dev performance / month (90-day MA)

+74.1%

Work mix

Features

32%

Maint

21%

Tests

26%

Docs

3%

Fixes

19%

Company breakdown

Cloudflare

Capacity vs pre-AI

3.9x

+290 engineers

1.65 ETV / dev / mo

Avg. dev performance / month (90-day MA)

+61.4%

Work mix

Features

30%

Maint

12%

Tests

33%

Docs

13%

Fixes

12%

Company breakdown
See all 7 orgs

Speed tells you how much a team ships. It doesn't tell you what.

Direction

Faster at what?

The same amount of output can be new features, or fixes to work that already shipped. The mix is the decision.

Performance Composition

How each org's ETV splits across features, maintenance, tests, docs, and fixes.

Patterns

Where does the output come from?

The same measure one level down, at the codebases the output is actually made in. Ranked by per-engineer ETV, this week against the full quarter.

Top 5 Repositories · last 7 days

Highest per-engineer Engineering Throughput Value (ETV) shipped in the last 7 days. Not normalized to a month.

  1. 1

    turborepo

    Vercel · 1 dev

    26%23%39%1%10%

    22.9

    ETV / dev · 7d

  2. 2

    agents

    Cloudflare · 2 devs

    19%0%46%6%28%

    20.0

    ETV / dev · 7d

  3. 3

    PowerToys

    Microsoft · 4 devs

    55%8%34%2%2%

    7.8

    ETV / dev · 7d

  4. 4

    vscode

    Microsoft · 38 devs

    27%9%50%1%13%

    7.4

    ETV / dev · 7d

  5. 5

    openai-agents-python

    OpenAI · 2 devs

    22%0%56%14%8%

    4.7

    ETV / dev · 7d

Top 5 Repositories · last 90 days

Highest per-engineer Engineering Throughput Value (ETV) over the trailing 90 days, normalized to a 30-day calendar month.

  1. 1

    openai-agents-js

    OpenAI · 1 dev

    17%3%50%9%21%

    84.9

    ETV / dev / mo

  2. 2

    openai-agents-python

    OpenAI · 2 devs

    17%2%46%13%22%

    41.1

    ETV / dev / mo

  3. 3

    turborepo

    Vercel · 4 devs

    35%28%16%3%19%

    16.8

    ETV / dev / mo

  4. 4

    zerocopy

    Google · 2 devs

    53%4%19%21%3%

    14.2

    ETV / dev / mo

  5. 5

    vscode

    Microsoft · 57 devs

    32%11%39%2%16%

    11.2

    ETV / dev / mo

Your turn

Where does your team stand?

The index measures public repositories. The same measure runs on private ones.

Benchmark your orgRead the methodology