Grasp · Diligence Report · Public sample from engine v0.3, kept as an illustration of the deliverable. Contributor identities are anonymized; client reports name them when the client elects it. Both the figures and the layout predate the current engine; current reports rename Departed to Inactive and add the self-reporting signals described on the methodology page.

Who still understands pallets/click?

Comprehension audit: windowed churn analysis, windowless survivorship analysis, and review-depth forensics. Metadata only; no source code contents were read.
generated 2026-07-26 · churn view: line-weighted, 3-year window · review coverage: full

Churn Grasp Score

living knowledge in recent work; where risk is accumulating now

Survivorship Score

living authorship of the standing inventory; the asset as it exists today

Standing inventory

39,329

surviving lines attributed across 165 files

Active humans

34 / 462

contributors active in last 90 days vs. all seen

Preliminary findings

Auto-drafted from thresholds; each finding is verified, contextualized, or struck during the analyst pass before final delivery.

Standing dark inventory: docs: 6,612 surviving lines, only 43% with a living author.
Key-person concentration: Contributor A holds 36% of all living-authored surviving code. Succession/retention consideration.
Key-person concentration: Contributor B holds 35% of all living-authored surviving code. Succession/retention consideration.
Review-process signal: 68% of sampled merged PRs show no recorded human review (sample of 91, stratified toward the largest merges).

Where risk is accumulating: Churn view (3-year window)

Module	Risk	Living knowledge	AI-attr.	Departed	Bus	Heat/180d
docs	LOW 37	60%	0%	40%	2	67
src/click	LOW 12	87%	2%	11%	2	140
tests	LOW 10	89%	0%	11%	2	138
(root)	LOW 2	98%	0%	2%	2	130
.github	LOW 0	100%	0%	0%	3	19
requirements	LOW 0	100%	0%	0%	2	0

The standing asset: Survivorship view (windowless)

Every surviving line attributed via git blame; a line written fifteen years ago that still ships counts fully. Sorted least-living first.

Module	Living authorship	Departed	Bus	Lines	Top holders
examples	36%	64%	1	1,780	Contributor B (36%), Contributor A (0%)
docs	43%	57%	2	6,612	Contributor C (18%), Contributor A (11%)
src/click	56%	44%	2	12,629	Contributor B (31%), Contributor A (13%)
tests	68%	32%	2	14,321	Contributor A (43%), Contributor B (12%)
.github	89%	11%	2	319	Contributor B (65%), Contributor A (15%)
(root)	100%	0%	2	3,668	Contributor B (45%), Contributor C (43%)

Key-person concentration

Share of all living-authored surviving code held per person: the succession and retention picture.

Contributor	Share of living code
Contributor A	36%
Contributor B	35%
Contributor C	15%
Contributor D	2%
Contributor E	2%
Contributor F	2%
Contributor G	2%
Contributor H	2%

Review-depth forensics

No recorded human review

68%

of sampled merged PRs

AI-reviewed only

reviews from AI bots, none from a human

Comment-free approvals

21%

of human-reviewed sampled PRs

Large fast merges

300+ lines merged in under 30 minutes

PR size	Count	Median time to merge
1-50 lines	66	34,183 min
51-300 lines	24	286,710 min
301-1000 lines	1	37,991 min

The review record has known blind spots (pairing, pre-coordinated work, and review outside the platform are all invisible to it). These are process signals rather than verdicts on individuals.

Method, scope, and limits

Computed from repository metadata with the published Grasp methodology (graspscore.com/methodology). Living authorship is a proxy for comprehension: a potential indicator rather than a diagnosis. AI attribution is a minimum. Nothing here asserts misconduct or lack of competence by any contributor, and nothing is a claim about software quality, security, or fitness. The accompanying evidence bundle (comprehension.json, survivorship.json, review_depth.json) allows every figure in this document to be independently re-derived.