CONTEXT(CLONE)
Sign in

Cross-check defect coding in pipe inspection video

For: Drainage or pipeline inspection manager at a utility or survey contractor

Pattern: Cross-examinationNeeds physical AI: plannedDesigned for 20 to 1,500 agents
Get early accessSign in to use

The pain today

Operators code defects from camera footage by eye against a standard code list. Coding varies from survey to survey, and rehabilitation budgets are set from those codes.

The ask

“I attached the pipe inspection videos, the operator's defect log and the defect coding standard. For each pipe length, check independently what the footage shows against the standard, and show me where the two checks and the logged codes disagree.”

Plain words, as you would say it to a colleague. Edit it to fit your case before you send it.

What you attach or connect

  • Pipe inspection videos by pipe length
  • Operator defect logs
  • Defect coding standard
  • Network map references

The unit of work

One worker task per one pipe length's footage and log entries.

Why a swarm fits

Each pipe length is coded alone against the same standard. Two perception model families code it independently, so a disagreement with the log is not just one model's habit.

Not for

Structural assessment or rehabilitation design. It checks coding consistency against the footage.

The decision tree

6 typed decisions, each with an action for every answer

At fixed moments in a run, the engine puts one narrow question to a decision model. The decision model never writes text: it answers yes or no with a probability, picks from listed options, or gives a score, about a small slice of the material. The engine then does exactly what this tree says, which is what makes the run auditable. The thresholds are the template's design values, not measured results.

  1. Planner, while planning

    Scope checkA choice among options

    Before work starts on a unit

    Does the written description of this pipe length's footage say the view was clear enough, and complete enough, to code against the standard?

    Sees only: Planned design: one pipe length's footage quality description and survey header

    Why: Planned: footage that cannot be coded is reported as such, not guessed.

    • Clear and completethenAccept
    • Survey abandoned part waythenMark unresolved
    • Too poor to codethenSkip this unit

  2. After workers, the judge checks

    Evidence checkYes or no, with a probability

    After a worker answers

    Does the written description of the footage at this distance meet the coding standard's own criterion for the defect code assigned?

    Sees only: Planned design: one code's criterion from the standard and the footage description at one distance

    Why: Planned: each code must be earned from the standard's wording, since budgets follow the codes.

    • Yes: 0.85 or higherthenAccept
    • Unsure: 0.50 up to 0.85thenEscalate to a strong model
    • No: below 0.50thenReject and retry

  3. Reconciler, while merging

    Conflict checkA choice among options

    While reconciling

    How do the two independent codings and the logged code compare for the defect at this distance?

    Sees only: Two independent codings and the log entry for one distance in one pipe length

    Why: Planned: a logged code is disputed only when both independent codings differ from it.

    • All three agreethenAccept
    • Both codings differ from the logthenMark unresolved
    • Codings differ from each otherthenEscalate to a strong model
    • Defect seen but not loggedthenMark unresolved

  4. Run control, between rounds

    Retry or stopYes or no, with a probability

    After a rejection or low confidence

    Does the coding standard's text distinguish the two codes clearly enough to settle the difference from the descriptions alone?

    Sees only: The two codes' criteria and both footage descriptions

    Why: Planned: borderline grades the standard cannot separate go straight to the certified inspector.

    • Yes: 0.85 or higherthenEscalate to a strong model
    • Unsure: 0.50 up to 0.85thenMark unresolved
    • No: below 0.50thenMark unresolved

  5. After workers, the judge checks

    Evidence checkYes or no, with a probability

    After a worker answers

    Is this disagreement reported by defect and pipe length only, with no operator named or compared?

    Sees only: One drafted disagreement entry

    Why: Planned: the check is about coding consistency, never about rating an operator.

    • Yes: 0.90 or higherthenAccept
    • Unsure: 0.60 up to 0.90thenReject and retry
    • No: below 0.60thenReject and retry

  6. Accountable person, before anything is settled

    Person decidesYes or no, with a probability

    Before anything is reported as settled

    Does this pipe length have a disputed code, an unlogged defect or footage that could not be coded?

    Sees only: One pipe length's agreed and disputed codes

    Why: Planned: a certified inspector settles every disputed code and signs the survey report.

    Accountable: A certified inspector settles every disputed code and signs the survey report.

    • Yes: 0.20 or higherthenAsk a person
    • Unsure: 0.05 up to 0.20thenAsk a person
    • No: below 0.05thenAccept

The fleet: who does what

Model tiers by role, not brands: you choose the models. Strong reasoning models plan and reconcile, small fast models do the wide work, and the judge is a decision model from a different family, so it does not share the workers' blind spots.

  1. Planner

    A strong reasoning model turns the coding standard into observable criteria and pairs footage with log entries.

    Decisions here:1. Scope check

  2. Workers

    Two perception models for video and images from different families, planned, each coding the same footage independently.

    Designed for 20 to 1,500 agents, one worker task per one pipe length's footage and log entries. Each worker receives only its own unit.

  3. Judge, from a different model family

    A decision model from a third family compares both codings with the log and keeps unresolved differences open.

    Decisions here:2. Evidence check5. Evidence check

  4. Reconciler

    A strong reasoning model lists agreed codes, disputed codes and defects seen but not logged, per pipe length.

    Decisions here:3. Conflict check4. Retry or stop

  5. Accountable person

    A certified inspector settles every disputed code and signs the survey report.

    Decisions here:6. Person decides

Checked before anything is accepted

  • Every code cites the footage distance or timestamp and the standard's criterion
  • A logged code is disputed only when both independent codings differ from it
  • Footage too poor to code is reported as such, not guessed
  • Disagreements are reported per defect, never per operator

What comes back

  • Per pipe length: agreed codes and disputed codes, with footage references
  • Defects seen in footage but absent from the log
  • Logged defects neither check could see
  • Footage that could not be assessed

What to measure

  • Share of disputed codes a certified inspector changes
  • Agreement between the two independent codings
  • Inspector hours per survey reviewed
  • Cost per pipe length checked

Names of measures only. No result is claimed for this template.

Templates open in the workspace chat with the ask filled in. Nothing runs until you send it.

Get early accessSign in to use

More in Physical AI and field operations

Check site photos against the construction programme

For: Construction project manager or owner's representative tracking progress across sites

Progress is reported as a percentage in a spreadsheet.

Pattern: Map, verify, reduceNeeds physical AI: planned5 decisionsDesigned for 30 to 800 agents

Roll up drone inspection footage across an asset network

For: Asset integrity manager for overhead lines, pipelines or telecom towers

Drone flights return footage for every span and tower.

Pattern: Hierarchical decompositionNeeds physical AI: planned6 decisionsDesigned for 50 to 3,000 agents

Review plant walkdown video from four specialist angles

For: Site HSE manager or operations manager at a process or manufacturing plant

Walkdowns are filmed but reviewed by one person with one checklist.

Pattern: Specialist panelNeeds physical AI: planned5 decisionsDesigned for 8 to 300 agents