# PR5.49 — Hemingway common-corpus results

Observed 2026-10-14 against independently inspected clean HEAD `a8c8ad0da3f51c5b5bd47eab4375add458dfff8f`. Owner screenshot shows green Production #67. Accepted state remains 15 products, seven approved routes and three remaining; Issue #5 open. No live database/deployment query.

## Actual progress

**Five of fifteen runnable product cases observed. Ten remain not-run.** PWA Premium lifetime's five separate record slots remain not-entitled. This is five completed before/recheck observations, not five passed features or three qualified alternatives.

The new record uses the exact five before/manual-after pairs committed in PR5.48. Existing protocol and prior research stay unchanged. No graph, claims, engine, memberships, taxonomy, SQL or publication changes; zero new qualified candidates.

## Executed session

[Hemingway anonymous free web editor](https://hemingwayapp.com/) in the Codex in-app browser, Edit mode, weakener details expanded. No account or Plus trial enrolled. Only the original short corpus was entered. Plus grammar/AI/rewrite/file upsells were visible but not invoked.

For each case, the exact before input was verified from the visible editor, DOM counters and a screenshot captured, then the exact supplied manual reference input entered and rechecked. The active control remained unchanged. These are human-authored reference revisions, not accepted vendor or AI suggestions.

| Case | Before | Recheck after supplied manual edit | Outcome |
| --- | --- | --- | --- |
| passive | 2 passive uses; Grade 4 | 0 passive uses; Grade 2 | Observed flags cleared |
| wordiness | 1 weakener and 1 simpler-word item; Grade 6 | Both 0; Grade 5 | Observed flags cleared |
| redundancy | No related counters/highlights; Grade 7 | No related counters/highlights; Grade 6 | No redundancy feedback for this sample |
| complexity | 1 very hard sentence; Post-graduate label | 0 hard/very hard; Grade 7 | Observed difficult-sentence flag cleared |
| active-control | No related counters; Grade 0 | Same unchanged text and counters | No related feedback; control only |

The wordiness screenshot shows the phrase `In order to` highlighted; the entire before sentence's wording is not thereby proven to have every issue detected. The redundancy grade changed after manual shortening, but a grade change is not evidence of a redundancy rule or suggestion.

The literal `Post-graduate` UI label is preserved without assigning a numerical grade. Displayed word counts are stored as UI observations, not treated as an independent word-count benchmark.

Ten screenshots are in ZIP `evidence/`, paired by case and phase. Exact text, UI values, dispositions and hashes are in `seed/research/hemingway-common-corpus-2026-10-14.json`. Screenshot viewport captures show the main passage/summary; expanded-counter values also came from the observed DOM. Initial parser failures for a nonnumeric readability label and singular weakener label were corrected before final evidence capture. Missing values were never recorded as zero.

## What remains untested

### Grazie Free — Chrome

The available browser inventory exposed the in-app browser and MCP apps only; no connected Chrome or accessible Grazie extension session was present. The owner's Chrome choice remains recorded. Actual extension readiness was asked, but no installed/login state is assumed.

All five Grazie cases remain not-run. No installation, account creation, extension permission change or vendor prose submission was performed. Free proofreading quota remains unresolved.

### Grammarly Free

A visit to app.grammarly.com reached the public homepage rather than a usable authenticated editor. The visible Log in link led to [the existing-account sign-in page](https://www.grammarly.com/signin). The agent entered no account details; a sign-in handoff was prepared.

All five Free anchor cases remain not-run. Public marketing or an anonymous checker cannot substitute for the chosen authenticated Free surface. After login, verify the displayed Free plan and record actual paid/locked feedback without enabling a Pro trial.

### PWA Premium lifetime

The owner has no target license. Keep all five slots not-entitled and preserve the documentary research. No purchase or silent substitution to Free, another paid tier or a trial.

## Decision and evidence limits

The observed passive, simpler-word and difficult-sentence flags support this bounded anonymous web session. They do not establish identical rule coverage, accuracy, unlimited use, offline behavior or comparative privacy.

The exact redundancy sample did not demonstrate a redundancy detector. Its result cannot support a broad shared redundancy operation. General absence across the entire product is likewise not established.

A full common operation still needs Grazie's exact Free session and the selected Grammarly Free anchor evidence; PWA remains documentary-only. No new qualification/import/publication proposal follows from these five observations.

## Continuing the worksheet

ZIP `OWNER_RESULTS_PARTIAL.json` fills the five Hemingway records and retains ten not-run plus five PWA not-entitled. Copy it for further observations; it is an evidence attachment outside repository `files/`.

Use the unchanged `SAMPLES.txt` for Grazie Free Chrome and Grammarly Free. Record plan/version/surface/English/style settings, completed before feedback, the exact manual after input, completed recheck, any paid/quota prompt and relevant screenshot references. Unknown meter values remain unknown. This worksheet is evidence intake, not automated qualification or publishing approval.

The shipped checker pins the result data, checks exact matching corpus text and audits the accepted graph. It does not execute browser cases, refetch vendor pages, verify screenshot bytes after copying `files/` alone, or approve a filled worksheet. Delivery attachment hashes are in MANIFEST.