How it works

the close-out pipeline, end to end

Peterson Technology PartnersAnsco & Associates · close-out, end to end
LEE WRIGHT · DERICK DAVIS · CHRIS HAMILTON
RULE SETS: PROVISIONAL — PENDING AT&T SPEC

A close-out photograph proves nothing.

It could be from any job, any day, any phone — which is exactly why someone has to look at every one of them. So we moved the proof to the moment of capture, where it costs nothing, and the package arrives already verifiable.

1683photographs captured on the phone
37/46packages that needed nobody
0cloud calls to judge a photo
01

Before the shutter

The phone checks he is on the line segment the engineer assigned — before he takes the photograph, not after the package is rejected.

100%of placed captures sat on their assigned run (1680 of 1683)
100% inside the engineer's corridor

GPS accuracy is subtracted before anything is called off-segment. Accusing a technician because the fix was poor is how a tool loses a crew.

02

At the shutter

Focus, exposure and legibility are judged on the device in about forty milliseconds. No network, no cloud call, no cost per photo.

511captures where the phone read the text it was asked to read

At 150,000 photos a month a cloud call per photo is a budget line. Arithmetic on the phone is free, and it works at the bottom of a vault.

03

The stamp

Every photograph is bound at capture to where, when, which device and which requirement — then hash-chained to the one before it.

1683captures chained across 46 packages and 6 devices
1683 records · break one link and every link after it fails

This is the one property that cannot be added later. A photograph taken without it is unverifiable forever — which is precisely why today's close-out photos have to be inspected by a person.

04

Off the phone

He presses done and puts the phone in his pocket. The upload continues while the app is closed, killed, or the phone reboots.

46packages delivered, largest 92 photos across 3 sectors

A foreground request dies in the vault and nobody finds out until the package is late. This is a background transfer the system owns.

05

The desk

The chain is re-verified here, server side — the phone's own word for it is worth nothing — and a package only reaches a person when something actually needs one.

37/46packages that needed nobody to look at them · 46/46 chains intact
1683 photographs capturedjudged on the phone — offline, freechain re-verified at the desk9 of 46 packages reached a person

Verified means nothing was changed after capture. It is deliberately not a judgement that the work passes — conflating those is how a tool starts quietly approving things nobody approved.

open the review desk →
06

The same evidence, years later

Nothing new is captured for any of this. Claims, presence and safety all read the records close-out already created.

4questions answered from the same capture records
onecapturea claim, two years onwho was on sitesafety conditionswas he ticketed

A claim arrives two years after the work. The records place a subcontractor seven metres from the reported point, working outside their locate ticket, with a bore six inches under spec.

run those engines →
07

The paperwork, separately

AT&T's own documents: construction drawings against contracts and invoices, the wireless document set against itself, and CD PDFs into GIS.

$7.37Mover-billing exposure found on one job, from a rate written per 100 LF and invoiced per LF

Also $8,262 Ansco never invoiced on the same job. An over-billing is caught by the customer. An under-billing is only ever caught by the contractor, and only if someone looks.

run the document engines →
See it yourself

The rule sets behind this are our best reading of AT&T’s requirements, not their published spec, and the app says so on screen. Ansco’s own rejection history is what calibrates where the line falls — it refines the judgement, it does not create it.