# Editor notes — LumenDesk 2.0

## Scope and editorial choices

- This package treats the fictional launch as of October 18, 2026, the latest dated evidence supplied. The article now displays that source-snapshot date. “Now” refers to that snapshot, not a later live availability check. Dates and versions are authoritative only within the exercise.
- Read all supplied source files, including inputs/editorial-brief.md (“Assignment”) and inputs/voice-sample.md (“Original voice sample written for this exercise”). Used a direct, practical voice focused on checking answers and total work, without presenting suggested trial steps as completed tests.
- Neither the author nor Kingy tested the product. The article explicitly discloses this. DeskLab's work is attributed to DeskLab, not adopted as personal experience.
- Trial recommendations are editorial advice, not evidence of measured benefits. “Try” does not imply a free trial: no free-trial terms are supplied. No assumptions were made about waitlist timing, current OCR quality, or a guaranteed future release.

## Evidence and reconciliation

| Article point | Source and editorial handling |
| --- | --- |
| Existing Team customers in Canada/US; new-customer waitlist; browser only | inputs/S1-announcement.md, “Access.” Immediate access is not extended to new customers. |
| API and EU plans | inputs/S1-announcement.md, “Access.” Q1 2027 is a plan without a date or guarantee, not current access. |
| Formats, 50 MB file limit, citations, OCR requirement and limits on actions | inputs/S1-announcement.md, “Features.” Described as document assistance, not autonomous execution. |
| Current price and minimum | inputs/S2-pricing.md, “Current plans” and “Correction.” October 17's USD 29 replaces the erroneous USD 19 screenshot from October 16. No annual discount inferred. |
| Workspace economics | inputs/S2-pricing.md, “Current plans.” Three seats × USD 29 = USD 87 monthly; USD 87 + USD 12 OCR = USD 99. Both exclude tax and upload overages. The 500-upload allowance is workspace-wide, monthly, with no rollover; overage is USD 0.08/upload. These totals are calculations, not separate advertised plans. |
| Vendor speed evidence | inputs/S1-announcement.md, “Vendor test.” Attributed internal test, 40 procurement packets, medians of 50 and 20 minutes. Difference of medians = 30 minutes; this is not a measured median paired saving or a promise for each packet. Verification is excluded and raw task-level results unavailable. |
| Independent observation | inputs/S4-independent-review.md, “Method.” Six searchable procurement PDFs; five correct page citations and one wrong citation with correct answer text. Kept counts rather than extrapolating an accuracy percentage. Not a benchmark; no scans, OCR, privacy or team-workflow testing. |
| Training, storage and deletion | inputs/S3-data-policy.md, “Handling.” Base-model training exclusion is attributed to the policy, not expanded to every possible use. Default 30-day storage, early admin deletion, active-storage deletion within 24 hours and backups for up to 30 further days are distinct. |
| External processing and compliance limits | inputs/S3-data-policy.md, “Processing.” External contracted provider, no on-device mode, no zero-retention or guaranteed-compliance promise. No independent privacy audit is supplied. |

## Unsupported leads excluded from the article

All leads below come from inputs/S5-internal-notes.txt, “Unverified writer notes, October 18” (the file's opening label; it has no formal section headings). They are not established facts and were not repeated in the article as claims, even in the headline or metadata.

- **“Eliminates hallucinations”:** no supporting evidence. DeskLab's small observation does not establish universal reliability; a page citation was wrong (inputs/S4-independent-review.md, “Method”). Do not infer that all answer text is reliable from this sample either.
- **“GDPR certified”:** no certificate, named certifying body or verification supplied. The data policy makes no guaranteed-compliance claim (inputs/S3-data-policy.md, “Processing”). Do not substitute a legal conclusion.
- **Fully offline:** inconsistent with external model processing and no on-device mode (inputs/S3-data-policy.md, “Processing”); access is browser-app-only (inputs/S1-announcement.md, “Access”). Excluded.
- **Every team saves 60% of work time:** unsupported generalisation. (50 − 20) / 50 = 60% describes the reduction in the vendor's reported first-draft medians only. It does not measure whole-team work or verification time (inputs/S1-announcement.md, “Vendor test”). The article uses the underlying minutes and limitations instead.
- **Lead with USD 19:** superseded explicitly by inputs/S2-pricing.md, “Correction.” Mentioned in the article only to explain the correction, never as an available price.
- **Purported customer quote about firing a department:** no named source or permission supplied. Omitted completely from article and metadata; no replacement testimonial invented.

## Checks performed

- Read every supplied input and manually cross-checked factual article claims against the filename/section references above. Reviewed recommendations to separate them from observed results.
- Ran local standard-library validation (validate.py): JSON parses and has exactly headline, excerpt and search_description; article headline matches metadata; cited article filenames exist. The revision check also verified that each cited article section heading matches its source file.
- Word count by whitespace: **670** for article body including the source-snapshot line, subheadings and inline citations, excluding the H1; **648** with inline citations removed. Either convention is within 600–800 words. Metadata is separate.
- Headline: **61 characters**. Excerpt: **29 words**. Search description: **149 characters**. All meet the specified limits.
- Checked the USD 87 and USD 99 calculations locally. Manually reviewed access versus roadmap language, source attribution, price correction, privacy qualifiers and lack of hands-on claims.
- No product testing, browsing, publication, images, external research or independent verification was performed. Raw inputs were not changed.

## Remaining limitations and handoff

The packet supplies no raw vendor results, representative benchmark, privacy audit, legal certification, waitlist timetable or guaranteed roadmap. Those omissions limit conclusions; they do not prevent a source-led article. The article is explicitly dated to the October 18, 2026 source snapshot. A final editorial copy read remains; any later-dated publication would need a fresh availability and pricing check, which this package does not claim to provide. Estimated human finishing time: **5–10 minutes**, not measured; obtaining new evidence would require additional, unestimated time.

Deliverables: output/article.md, output/metadata.json and output/editor-notes.md. Original submission: output/original-submission/ (all three original deliverables, preserved byte-for-byte). Revision summary: output/change-log.md. No publication action is requested or completed.
