Deriving verdicts from stored observations…

Aurora Lamp Stand

example.com/aurora-lamp-standlast run 2026-08-14 09:00replayedplan v1 (seed)

Completedok3/3

Valueunverified

Extractiondrifted

Causeelement-relocated

every run on one axis, oldest first · bar height is completeness, ink is the extraction verdict, a caret marks a run where the judges disagreed

How to read the four verdicts

Completed: How much of the extraction plan produced a value. Says nothing at all about whether the value is right.
Value: Whether the number is believable against this field’s own history. A price can be believable and still be the wrong price.
Extraction: Whether the place the value was read from has moved. This is the judge nothing else has, and the only one that catches a correct-looking number read from the wrong element.
Cause: What the evidence says is actually wrong — a redesign, a removed element, a real-world price change — and how confident that reading is.

What happened

price · 80% confidence in this cause

2026-08-13div.productdiv.pricingspan.price899 INR

via .price

2026-08-14div.productdiv.price-stackspan.price899 INR

via .price

The value did not change. Where it was read from did.

2026-08-14899 INRdriftedunverifieddiv.product > div.price-stack > span.pricev1

The band under the line is the element this value was read from. The run where the band changes and the line does not is the warning.

price still reads the correct value and no longer reads it from the same place: div.product > div.pricing > span.price -> div.product > div.price-stack > span.price. Nothing is wrong with the data yet.

Run history

Each run judged as it arrived. A marked row is one where the judges disagreed.

Recent runs, newest first, with all four verdicts.
Run	Completed	Value	Extraction	Cause	Plan
2026-08-14	ok3/3	unverified	drifted	element-relocated	v1
2026-08-13	ok3/3	unverified	trusted	healthy	v1
2026-08-12	ok3/3	unverified	trusted	healthy

Replacement selectors the evidence supports

Candidate selectors, ranked by confidence, with what each one reads on the page.
Selector	Confidence	Reads	Why the evidence supports it
[data-testid="product-price"]	95%	“₹899”	Captured on the page during the run being diagnosed, holding "₹899". It is a test id, which exists to be depended on and survives restyling.

Every candidate above was observed on the page. A model may pick from recorded evidence; it may never add to it.

Aurora Lamp Stand

What happened

Run history

Signals on the latest run