What you will achieve
You will create a capture from one approved repository folder or bounded document set, freeze it, complete Survey → Plan → Extract → Discover → Infer → Ask, preview the generated knowledge, and verify one cited answer.
Before you begin
- Genesis desktop is installed and you can open the target project.
- Your account has the distribution-admin role.
- You have permission to copy and process the selected source.
- You have agreed whether the source may be processed by the configured AI model service.
- You know two or three questions the source should be able to answer.
Build the capture
- 01
Open SOCK
From the project launcher, open SOCK. If the tile is missing, verify that you are in the desktop app and have the distribution-admin role.
- 02
Create a named capture
Use a name that records source, purpose, and time, such as
payments-service-onboarding-july-2026rather thantest. This creates the record that will hold the work; it does not begin the Capture phase. - 03
Add one approved source
In the Sources phase, choose a repository, folder, file set, attributed knowledge note, or an approved PostgreSQL, MongoDB, or Oracle connection. For this tutorial, avoid sensitive database samples and large multi-source scopes.
- 04
Review the evidence boundary
Confirm the source, revision or path, and intentionally omitted evidence. This is your last opportunity to correct the source set.
- 05
Freeze and run Survey
Freeze the selected sources. Genesis copies or exports the evidence, fingerprints it, writes attestations, and then begins the Capture phase with Survey.
- 06
Run Plan
Continue when Survey completes. Review surfaced blockers and the safe execution plan before proceeding.
- 07
Run Extract
Allow deterministic extraction and AI-assisted comprehension to finish. Treat any warnings or gaps as part of the result.
- 08
Preview one extracted unit
Choose Preview on a completed source unit. Open its README, Evidence, or Known Unknowns file and confirm that claims link to citations and retain an evidence tier.
- 09
Run Discover
Continue to the higher-order pass. Review the cross-source surfaces grouped by dimension and lens. A dimension with zero surfaces is a valid result, not an error.
- 10
Run Infer
Allow SOCK to derive deeper candidate knowledge for the discovered surfaces. Confirm the knowledge-health view labels inferred output and keeps gaps visible. In releases with a combined Discover & Infer tab, both passes run within this view.
- 11
Open Ask
Ask one pre-agreed, bounded question such as “Which external interfaces are declared?” or “Where is retry behaviour configured?”
- 12
Inspect a citation
Open at least one cited source and check that it supports the answer. The citation is part of the answer, not optional decoration.

Open SOCK from the project launcher. This product-team standalone UX is illustrative; tile status, labels, and available areas can change between Beta releases. Labels, actions, and layout may vary by release.

Name the capture for the system, purpose, or evidence point in time so teammates can distinguish it later. Labels, actions, and layout may vary by release.

Review the complete source boundary before freezing. This example includes one repository and one read-only connected system. Labels, actions, and layout may vary by release.

After all sources are frozen, Survey becomes available and the lifecycle view shows the durable evidence set. Labels, actions, and layout may vary by release.

A completed Plan hands a ratified execution sequence to Extract; review blockers and gaps before continuing. Labels, actions, and layout may vary by release.

Extraction completion exposes per-unit Preview actions, knowledge health, recorded gaps, run activity, and the live Claude terminal. Labels, actions, and layout may vary by release.

Discover catalogues cross-source surfaces; Infer derives deeper candidates. The Beta UI may group both passes in one tab while retaining their separate responsibilities. Labels, actions, and layout may vary by release.

Preview at least one generated artefact and verify that claims, citations, tiers, status, and known unknowns remain inspectable. Labels, actions, and layout may vary by release.

Ask answers from the capture, marks inference and unresolved boundaries, and provides numbered evidence links for verification. Labels, actions, and layout may vary by release.
The 21 dimensions you will review
Discover evaluates the capture against the complete internal dimension catalogue below. Infer then derives deeper candidate knowledge from the surfaces Discover finds. A dimension can legitimately return zero surfaces when the frozen evidence does not support a finding.
| # | Dimension |
|---|---|
| 1 | as-is-archaeology |
| 2 | business-journey |
| 3 | business-rules |
| 4 | call-chain |
| 5 | capability-map |
| 6 | compliance-footprint |
| 7 | contract-surface |
| 8 | data-lineage |
| 9 | data-relation |
| 10 | dead-code |
| 11 | dependency-blast-radius |
| 12 | domain-model |
| 13 | infra-inventory |
| 14 | integration-topology |
| 15 | interface-mappings |
| 16 | legacy-estate-map |
| 17 | migration-seams |
| 18 | ops-inventory |
| 19 | reliability-map |
| 20 | security-posture |
| 21 | ui-flow |
For definitions, examples, and the review question behind each identifier, open the [complete 21-dimension glossary](/sock/reference#dimensions).
Verify the result
- The source set shows as frozen and cannot be edited.
- Survey, Plan, Extract, Discover, and Infer show durable completion or a clearly named recoverable blocker.
- At least one generated artefact can be previewed and its claims trace to citations, evidence tiers, and status.
- The Discover and Infer result exposes all applicable surfaces across the 21-dimension catalogue without implying that every dimension must contain a finding.
- Ask cites captured evidence for a supported answer.
- A question outside the supplied evidence returns a scope limit or known unknown rather than invented certainty.
Next steps
Review all 21 dimensions
Use the full glossary to understand what each dimension looks for and which review question it supports.
Read nextLearn the evidence model
Understand source access modes, credentials, AI processing, and evidence tiers.
Read nextConnect a database
Review database versions, bounded row reads, sampling risk, and the connector workflow.
Read nextRun a meaningful evaluation
Define questions and success criteria before capturing a larger product slice.
Read next