Example diff

What a diff actually looks like

This is not a mockup. Everything below is the real output FilingSmith produced for a real filing, copied from the record it was generated into, with nothing rewritten for the website.

We publish it for one reason: the entire claim of this product is that the numbers are checkable, and a claim like that is worth exactly as much as your ability to check it. So here is one, with its sources.

Accession
0000723125-26-000015
Prior accession
0000723125-26-000006
Filed
25 June 2026
Generated
14 August 2026

Micron Technology, Inc.

MU10-QQ3 FY2026Open on EDGAR ↗
Generated
14 August 2026
Compared against
Prior quarter 10-Q, as filed
Categories
margin · buyback · segment
MarginGross marginMagnitudegross margin up
Prior74.41%This filing84.56%+10.15pp

Gross margin expanded to 84.6% from 74.4% (up ~1,015bps) as revenue rose to $41.46B from $23.86B, driving operating income up to $33.32B and net income up to $28.24B.

Item 2 — Management's Discussion and Analysis · ¶ 2-p17 · us-gaap:GrossProfit
AI-driven memory and storage growth is outpacing industry supply. In the third quarter of 2026, we continued to benefit from substantial improvements in pricing and margins, reflecting strong demand growth, driven in large part by the continued advancement of AI. The AI-driven growth in the data cen…
BuybackRepurchase outlay, year to dateMagnitudeongoing
Prior$650.00MThis filing$650.00M$0

Weighted-average diluted shares rose slightly to 1,145,000,000 from 1,142,000,000; no shares were repurchased in the period, leaving year-to-date buybacks unchanged at $650.00M.

Item 2 — Management's Discussion and Analysis · ¶ 2-p149 · us-gaap:PaymentsForRepurchaseOfCommonStock
Our Board of Directors has authorized the discretionary repurchase of up to $10 billion of our outstanding common stock through open-market purchases, block trades, privately-negotiated transactions, derivative transactions, and/or pursuant to Rule 10b5-1 trading plans. Through May 28, 2026, we had…
SegmentCMBU revenue (largest segment)Magnitudesegment mix
Prior$7.75BThis filing$13.77B+$6.02B

All segments grew: CMBU revenue rose to $13.77B from $7.75B, CDBU to $11.52B from $5.69B, MCBU to $11.52B from $7.71B, and AEBU to $4.63B from $2.71B.

Item 1 — Financial Statements · ¶ 1-p562 · us-gaap:RevenueFromContractWithCustomerExcludingAssessedTax
We have the following four business units, which are based on market segments and our reportable segments:
Every figure in a published narrative is checked back against the set of values the diff engine computed. A single digit that cannot be traced rejects the whole narrative rather than triggering a fix-up.
The filing
CompanyMicron Technology, Inc.
FilingForm 10-Q, Q3 FY2026
Compared againstThe prior quarter’s 10-Q, as filed
Filed25 June 2026
Accession0000723125-26-000015
Prior accession0000723125-26-000006

How to check it

Every claim above carries a citation: the item, the paragraph, and the sentence it was drawn from. That is deliberate — a claim you cannot trace is a claim you have to take on trust, and trust is the thing this product is supposed to earn rather than request.

  1. Open the filing on EDGAR using the link above.
  2. Find the cited item and paragraph. The excerpt shown is a literal slice of the filed document, not a paraphrase.
  3. For any monetary or share figure, check it against SEC’s XBRL API for this filing — https://data.sec.gov/api/xbrl/companyconcept/CIK{cik}/us-gaap/{tag}.json returns every value the company tagged for that concept, each carrying the accession number of the filing it came from.

That last step is the one that matters, and it is the same step the system takes: the figures are read from that data, filtered to this filing’s accession number, before any prose exists.

What is generated and what is not

It is worth being exact about which parts of the block above a language model wrote.

ElementWhere it comes from
Every numberSEC’s XBRL API for this filing, or exact arithmetic over those values
The delta and direction of each metricComputed in code, from those values
The change categoryAssigned by the diff engine
The citation — section, paragraph, excerptLocated in the stored source document by byte offset
The sentence you readWritten by a language model, from the structured diff — and rejected outright if any figure in it fails to reconcile

The model never sees the filing. It sees the computed diff: categories, metric names, current and prior values, deltas. It cannot introduce a figure, because after it writes, every number in its output is checked against the set the engine computed, and an unsourced digit rejects the whole narrative rather than triggering a fix-up.

The source excerpt is deliberately withheld from the model as well. It is filing prose full of numbers the diff did not compute — showing it to the model is an invitation to reach for one.

Categories a diff can contain

CategoryHow it is detected
Margin movementArithmetic on reported revenue, cost of revenue and gross profit
Segment shiftsPer-segment revenue from the dimensional tags inside the filing
Buyback activityReported diluted share count and repurchase outlay
Guidance changesForward figures lifted from the filing’s prose, each verified as a literal span of the filed text
Risk-factor additionsThis year’s Item 1A compared paragraph-by-paragraph against last year’s
Executive changes and other 8-K eventsThe filing’s own item codes — 5.02, 2.02, 4.02 and others — read directly, no inference

The last two carry no figures at all, so no model call is made for them; their summaries come straight from the filing’s structure.

Not every filing produces a change in every category, and a category with nothing material to report is omitted rather than padded.