See which AI tools are actually earning their license.

GitClear attributes every line of code to the model that wrote it via Claude, Cursor, Copilot, Codex, Augment or Gemini. Then durable output is scored against rework, defects, review time, and more.

One comprehensive scorecard. Ten minutes to get started. No sales call required.

No credit card Live scorecard in minutes SOC 2 Type II

gold.model_task_roi — enriched from your bronze tables

databricks unity catalog · airflow on astronomer · 5,200 ai work episodes

bronze → silver → gold
rebuilt nightly · 12 min p95

Bronze

bronze.ai_assistant_events

bronze.git_commits

bronze.pull_request_reviews

as your vendors ship it

conform

Silver

silver.ai_work_episodes

silver.change_lineage

one row per AI work episode

relate

Gold

gold.model_task_roi

gold.durability_curve

question-shaped marts

01 Task–Model ROI Matrix

“For this kind of task, which model has historically delivered the best value?”

best value

lowest all-in cost

fastest result

lowest revision risk

least steering

prompt category	Apex-R 2.4 premium reasoning	Apex-F 2.4 premium fast	Core 3.1 balanced workhorse	Lite 1.9 cheap, verbose	Local-32B self-hosted
implement_feature outcome: durable Diff Delta @ 90d	1.85× $79.2 / success 80% @30d · n=262	1.70× $79.8 / success 78% @30d · n=87	0.97× $98.9 / success 74% @30d · n=612	0.37× $212 / success 60% @30d · n=349	0.11× $436 / success 49% @30d · n=146
root_cause_bug outcome: verified diagnosis, no recurrence	2.34× $81.9 / success 82% @30d · n=131	1.62× $109 / success 71% @30d · n=44	0.81× $154 / success 64% @30d · n=306	0.19× $558 / success 44% @30d · n=175	0.04× $1381 / success 34% @30d · n=73
dry_cleanup outcome: duplication removed, retained @ 90d	1.63× $52.1 / success 90% @30d · n=66	n=22 insufficient	1.00× $55.5 / success 90% @30d · n=153	0.60× $75.8 / success 86% @30d · n=87	0.17× $156 / success 70% @30d · n=36
explain_code outcome: explanation accepted, next task lands	1.60× $21.0 / success 97% @30d · n=112	1.74× $18.0 / success 97% @30d · n=37	1.00× $21.5 / success 94% @30d · n=262	0.51× $34.8 / success 83% @30d · n=150	0.15× $70.0 / success 69% @30d · n=62
address_pr_feedback outcome: thread resolved without reopen	1.71× $50.3 / success 89% @30d · n=103	1.78× $44.7 / success 92% @30d · n=34	0.98× $57.1 / success 87% @30d · n=240	0.42× $110 / success 73% @30d · n=137	0.11× $238 / success 59% @30d · n=57
build_failing_test outcome: fails before fix, passes after	1.66× $32.1 / success 94% @30d · n=47	n=16 insufficient	1.04× $33.6 / success 91% @30d · n=109	0.48× $59.9 / success 78% @30d · n=62	0.14× $117 / success 65% @30d · n=26

prompt category

Apex-R 2.4

premium reasoning

Apex-F 2.4

premium fast

Core 3.1

balanced workhorse

Lite 1.9

cheap, verbose

Local-32B

self-hosted

implement_feature

outcome: durable Diff Delta @ 90d

1.85×

$79.2 / success

80% @30d · n=262

1.70×

$79.8 / success

78% @30d · n=87

0.97×

$98.9 / success

74% @30d · n=612

0.37×

$212 / success

60% @30d · n=349

0.11×

$436 / success

49% @30d · n=146

root_cause_bug

outcome: verified diagnosis, no recurrence

2.34×

$81.9 / success

82% @30d · n=131

1.62×

$109 / success

71% @30d · n=44

0.81×

$154 / success

64% @30d · n=306

0.19×

$558 / success

44% @30d · n=175

0.04×

$1381 / success

34% @30d · n=73

dry_cleanup

outcome: duplication removed, retained @ 90d

1.63×

$52.1 / success

90% @30d · n=66

n=22 insufficient

1.00×

$55.5 / success

90% @30d · n=153

0.60×

$75.8 / success

86% @30d · n=87

0.17×

$156 / success

70% @30d · n=36

explain_code

outcome: explanation accepted, next task lands

1.60×

$21.0 / success

97% @30d · n=112

1.74×

$18.0 / success

97% @30d · n=37

1.00×

$21.5 / success

94% @30d · n=262

0.51×

$34.8 / success

83% @30d · n=150

0.15×

$70.0 / success

69% @30d · n=62

address_pr_feedback

outcome: thread resolved without reopen

1.71×

$50.3 / success

89% @30d · n=103

1.78×

$44.7 / success

92% @30d · n=34

0.98×

$57.1 / success

87% @30d · n=240

0.42×

$110 / success

73% @30d · n=137

0.11×

$238 / success

59% @30d · n=57

build_failing_test

outcome: fails before fix, passes after

1.66×

$32.1 / success

94% @30d · n=47

n=16 insufficient

1.04×

$33.6 / success

91% @30d · n=109

0.48×

$59.9 / success

78% @30d · n=62

0.14×

$117 / success

65% @30d · n=26

best in row tied with best (CI overlap) n < 25, no ranking shown

02 Token-to-Durable-Production Yield Curve

“Where does each model's apparent productivity disappear?”

Diff Delta is lineage-aware: moves, renames and reformatting keep their lineage, so code the developer rewrites stops counting toward the model.

A clipped section of two gold tables from our design study, drawn in HTML/CSS so it scales. Synthetic data : model names are placeholders and no real measurement is implied.

Databricks data engineering

GitClear specializes in enriching Databricks bronze tables — raw AI assistant telemetry, git history, pull request events, issue trackers — into the silver and gold tables that relate facts nobody could join before: which model earned its inference spend on which kind of task, and how much of its output was still in production 90 days later. The two panels at the top of this page are those tables.

We build these pipelines on Apache Airflow with Astronomer, the same orchestration we run for billion-dollar enterprises. Bronze ingest → silver work episodes → gold marts, rebuilt nightly in Unity Catalog, against your own warehouse — your data never leaves your Databricks account.

Medallion modeling (bronze → silver → gold) in Unity Catalog, with lineage that survives an audit
Containerized, idempotent Airflow DAGs on Astronomer — backfill a year of history without babysitting it
Diff Delta as the unit of output, so moves, renames and reformatting never inflate a model's numbers
Episode-grain facts, so cost, durability and revision risk can all be asked about the same row

The product

Four surfaces. One defensible ROI score.

Every AI stat in GitClear originates from deep analysis of code changes — so when a number doesn't look right, you can always drill into the code that produced it.

01 · Line-level attribution

Every line tagged with the model that wrote it.

GitClear cross-references your Git history with vendor AI usage APIs and agent telemetry hooks to produce commit-grade provenance — no guessing, no aggregate estimates.

Claude, Copilot, Cursor, Codex, Augment and Gemini APIs supported out of the box
Attribution precision maximized via telemetry hooks
Access via a robust API, for your own analysis or internal reporting

src/api/payments/checkout.ts

authored_by_llm · 90d view

COPILOT

const result = await validatePayment (req.body);

COPILOT

if (!result.ok) return res. status ( 400 ). json (...);

HUMAN

CLAUDE

try { await chargeWithRetry (result.token, 3); }

CLAUDE

catch (err) { logger. error (err); throw err; }

CURSOR

const audit = await logTransaction (result, req.user);

HUMAN

return res. json ({ ok: true , id: audit.id });

28%

Copilot

14%

Cursor

29%

Claude

29%

Human

02 · AI hotspot directories

Find the folders where AI is creating more work than it saves.

Not every directory responds to AI the same way. GitClear surfaces the folders where AI-assisted code has elevated defect and duplication rates — so you can coach, gate, or restrict tool access before it compounds.

Per-directory AI %, defect Δ, duplication Δ
Risk score normalized against your own baseline
Exportable as quarterly engineering review artifact

AI hotspot directories — defect & duplication risk last 90d

src/api/payments/ 68% +4.1% 3.2×

lib/auth/oauth/ 54% +2.8% 2.4×

app/models/user/ 47% +1.2% 1.9×

src/components/ui/ 71% +0.3% 1.4×

test/integration/ 82% -0.1% 0.8×

03 · Cohort comparison

See human vs. LLM code, measured by the same yardstick.

GitClear's Diff Delta metric works the same way whether a line came from Claude or a senior staff engineer. Compare durable change velocity, rework rate, and review time across cohorts — without apples-to-oranges caveats.

Cohort views by team, repo, or AI tool usage level
Side-by-side weekly trends — AI power users vs. non-adopters
Statistical significance flags on every delta

Durable change · AI-assisted vs. human-authored 12 wk

AI-assisted

11 devs

Diff Delta / wk +18%

Rework rate (30d) 12%

PRs merged / wk 34

Lead time 1.4d

Human-authored

3 devs

Diff Delta / wk baseline

Rework rate (30d) 9%

PRs merged / wk 28

Lead time 2.2d

Weekly durable change

AI Human

The methodology

Inspired by Google DORA. Built for the AI era.

Three inputs, one defensible number — so finance, your board, and your own engineers can all read the same scorecard without arguing about what it means.

Attribution

AI usage APIs plus commit heuristics plus agent telemetry hooks — not survey estimates. Every line traceable to the model that wrote it.

Output quality

Diff Delta quantifies durable change vs. churn. Human and LLM code measured with the same metric, across the same time window.

Developer experience

Self-reported hours saved and satisfaction scores. Productivity gains don't count if your best engineers are walking.

Industry Leading
AI Code Quality Research

211M

lines of code analyzed across three longitudinal studies. Cited by MIT Tech Review, TechCrunch, and The New Stack.

8×

increase in duplicate code blocks since AI coding assistants became mainstream in enterprise codebases.

9×

higher code churn from AI power users — who also produce 4–10x more code volume.

Integrations

Works with the tools your team already pays for.

GitClear plugs into your Git host and your AI vendor APIs directly — no proxies, no middleware, no code changes. First scorecard renders in under ten minutes.

GitHub GitLab Bitbucket Azure DevOps GitHub Copilot Cursor Claude Code Anthropic API Gemini Code Assist Augment

See what your AI spend is actually returning.

Connect your repos. Get your scorecard in under ten minutes. No credit card, no sales call — unless you want one.