# Editor notes — revised LumenDesk 2.0 package

## Scope and voice

This is a fictional exercise, assessed as of October 18, 2026, the latest dated source observation in the packet. “Now” means that launch window; no later access or policy changes are established. The article serves small teams deciding whether to try a document assistant. [inputs/editorial-brief.md — Assignment; inputs/S4-independent-review.md — Method]

The supplied brief and voice sample were read again during review. The article uses direct, practical advice about known answers, citations and verification work without reproducing the sample's sentences. Neither the author nor Kingy tested the product. The proposed evaluation is advice, not an action performed. “Trial” means a customer evaluation, not an advertised free trial; the packet establishes no free-trial offer, waitlist timing, cancellation terms or refund rights. [inputs/editorial-brief.md — Assignment; inputs/voice-sample.md — Original voice sample written for this exercise; inputs/S1-announcement.md — Access; inputs/S2-pricing.md — Current plans]

## Evidence and correction decisions

| Source and section | Treatment |
| --- | --- |
| `inputs/S1-announcement.md` — Access | Existing Team customers in Canada and the US have announced browser access. New customers have a waitlist, not immediate access. API and EU availability are Q1 2027 plans, without a guaranteed rollout or exact date. |
| `inputs/S1-announcement.md` — Features | Included supported formats, 50 MB per-file limit, separate OCR requirement and action limits. Did not describe an autonomous operator. |
| `inputs/S2-pricing.md` — Current plans; Correction | October 17 correction controls: USD 29 per seat monthly, not the erroneous USD 19 screenshot. Three-seat minimum; 500 uploads per workspace monthly, no rollover; USD 0.08 overages and USD 12 workspace OCR add-on. Prices exclude tax. No annual offer inferred. |
| `inputs/S1-announcement.md` — Vendor test | Attributed 40-packet result to LumenDesk. Restricted the 60% calculation to reported median first-draft time. Verification is excluded and raw task-level results are unavailable. No end-to-end savings claim. |
| `inputs/S4-independent-review.md` — Method | Attributed six-PDF observation to DeskLab. Reported five correct page citations and one wrong page despite correct answer text. Did not infer a general accuracy rate or testing of scans, OCR, privacy or team workflow. |
| `inputs/S3-data-policy.md` — Handling; Processing | Preserved the base-model training statement, default 30-day retention, admin deletion option, active-storage deletion within 24 hours and possible 30 further days of backup retention. Included external processing and no on-device mode. Did not equate no base-model training with no storage or no third-party processing. |

The price correction is not evidence of an increase from USD 19. DeskLab's observation addresses citations, not speed, so it does not validate the vendor's speed result. The article's factual claims cite local filenames and exact headings. No original input file was edited.

## Calculations and metadata

- Minimum subscription: 3 × USD 29 = **USD 87 per month**, before tax and overages.
- Minimum with OCR: USD 87 + USD 12 = **USD 99 per month**, before tax and overages.
- Draft-time difference: 50 − 20 = **30 minutes**. Relative reduction: 30 ÷ 50 × 100 = **60%** of the reported median draft-production time. This is not average individual savings or a reduction in the full working day.
- Inputs: `inputs/S2-pricing.md` — Current plans; `inputs/S1-announcement.md` — Vendor test.

The revised excerpt explicitly qualifies USD 87 as before tax and overages. Its access caveat is supported by `inputs/S1-announcement.md` — Access. The search description adds no product claims beyond the article topics.

## Unsupported leads excluded

Every lead below comes from `inputs/S5-internal-notes.txt` — “Unverified writer notes, October 18” (its opening label, not a formal Markdown heading).

| Lead | Reason excluded as a claim |
| --- | --- |
| “Eliminates hallucinations” | No supplied evidence establishes this. DeskLab reports a page-citation error, although that is not evidence of wrong answer text. [inputs/S4-independent-review.md — Method] |
| “GDPR certified” | No certificate or supporting evidence is supplied. The policy does not guarantee regulatory compliance. [inputs/S3-data-policy.md — Processing] |
| “Runs fully offline” | Unsupported and inconsistent with external model processing and no on-device mode. [inputs/S3-data-policy.md — Processing] |
| “Saves every team 60% of their work time” | Improper generalization from vendor median draft time excluding verification. Only the qualified calculation is included. [inputs/S1-announcement.md — Vendor test] |
| “Lead with USD19” | Superseded by the explicit USD 29 correction. [inputs/S2-pricing.md — Correction] |
| “We fired our paperwork department” | No named source or permission; no evidence for the implied staffing outcome. No customer quotation was fabricated or used. [inputs/S5-internal-notes.txt — Unverified writer notes, October 18] |

## Review changes

- Removed “rather than treating scans as standard uploads.” That wording could imply an unsupported distinction in upload counting. The replacement says only to budget for OCR rather than assuming it is included in the base subscription. [inputs/S1-announcement.md — Features; inputs/S2-pricing.md — Current plans]
- Qualified the base price with tax and overages in both article and excerpt.
- Added practical evaluation advice on setting a cost-justifying time-saving target and involving the usual document checker. This is editorial advice, not evidence of measured savings.
- Changed the final subheading to “A sensible first evaluation” to reduce the chance of implying a free-trial offer.
- Strengthened word-count validation to exclude all headings and citations as well as metadata. The original passed its stated count including subheadings; the revision also passes the stricter paragraph-only count.
- Made the validator resolve paths relative to its own location and read UTF-8 explicitly. Added JSON string-value checks, archived-file integrity verification and a saved results report.

## Checks actually completed

Re-read the assignment, brief, voice sample and all five evidence files. Compared article claims with the source passages, focusing on source conflicts, access timing, scope of the speed measurement, citation errors, OCR and retention. All numerical calculations above were checked again.

Executed the revised `output/check_package.py` successfully. Latest results are in `output/check-results.json`:

- Article: **625 paragraph words**, excluding all headings, citations and metadata; **659** including subheadings and Markdown heading markers but excluding headline and citations; **705** including inline citations but excluding headline. All meet 600–800.
- Headline: **61 characters**, within 50–70 and identical in article and metadata.
- Excerpt: **35 words**, within 45.
- Search description: **151 characters**, within 140–160.
- Metadata parses as JSON with exactly the required keys and nonempty string values.
- All **13 article citation groups** reference existing files and headings.
- All three required deliverables are readable UTF-8 files.
- All **four archived original files** match their recorded SHA-256 hashes. They were copied and byte-compared before edits.

Citation validation proves that files and headings exist, not that the prose is supported; the source comparison was a separate editorial review. Arithmetic checks are not product tests. No browser, rendered-layout inspection, external research, product testing, publication or image work was performed.

## Limitations and handoff

The packet cannot establish general answer accuracy, end-to-end savings, OCR quality, privacy implementation or regulatory suitability for a particular business. Raw vendor results are absent and DeskLab's observation is small and informal. [inputs/S1-announcement.md — Vendor test; inputs/S4-independent-review.md — Method; inputs/S3-data-policy.md — Processing]

Revised deliverables: `output/article.md`, `output/metadata.json`, `output/editor-notes.md`. Review summary: `output/change-log.md`. Reproducible checks and report: `output/check_package.py`, `output/check-results.json`. Original submission: `output/original-submission/`, containing all four originally delivered files plus a hash manifest. The one-time `output/revise_package.py` records preservation and the initial text edits; it intentionally refuses to overwrite an existing archive.

Estimated human finishing time: **5–10 minutes** for final copy review and house-style citation handling. This is an estimate, not a measured result. Additional product evaluation or reporting would require separate scope.
