THE HEARING / CLASS WRITING AND HOMEWORK SUBMISSIONS
30 October 2026 | Created by Yoav Duman

DECISIONS ON THE USER'S PROPOSALS
1 REQUIRED: class mode now blocks paste and text-drop insertion into editable text fields, including entry. Just for fun allows it. Ordinary typing/composition, copying out, authorized backup import and built-in frames remain available. A short live notice explains blocked input; the code does not read clipboard contents.
2 ADOPTED: aim 60-85 words remains a concision target; the main explanation completes at 50-120. Both bands appear in briefing, editor/live count, inclusion checks and grading summary. The 30-45-word comparison and every evidence/reasoning requirement remain.
3 ADAPTED: Print / Save as PDF accurately names the browser dialog, with upload/text-entry guidance. It is not labelled as a direct PDF download. Save backup file (.json) explicitly identifies a resume backup rather than a readable grading file. Full HTML/plain-text records remain available.
4 ADOPTED: the internal failed-search-service sentence was removed from verification_note. Original source identities, URLs, dates, locators, substantive caveats and disclosed uninspected judgments remain.
5 ADAPTED: three optional country-specific target buttons insert neutral event/opportunity frames with learner-filled slots. No target is silently auto-filled from an opening hypothesis. Free text remains available; unfinished helper slots do not satisfy the target check.
6 ADAPTED: optional rival/evidence writing-frame buttons put persistent stems into the fields. Students replace the bracketed instructions. Existing text is preserved unless they explicitly confirm replacement. The same fields persist across Hearing, Compare and Finish. Unfinished frames cannot complete either required response.
7 ADOPTED: What the record supports and Where the claim stops are on each card face outside collapsed details. Eight source-grounded limits fill the previously missing card-specific boundaries. All original card bodies, source records, exact numerical audit, original links and reading controls remain. Seeing face guidance does not pretend a source record was opened.
8 ADOPTED: Copy grading summary provides exact learner text, name, mode, solo/shared-pair status, dated case, target, spine, main paragraph, evidence board/complication reason, source-limit callout/self-check, both question attempts/last readings and applied decisions, rival, discriminating request, second case/comparison and current Draft/Complete status. It does not generate analysis, award a grade, or replace the full source/history report.

CLIPBOARD HARDENING
A cloud-browser delay exposed a possible blocked/pending permission path. The selected readonly summary now appears immediately on the same click, independent of clipboard completion. Manual Control+C was actually verified. Automatic clipboard success, explicit denial and unresolved permission are tested in the engine harness.

RIGOR RETAINED
Two distinct supports, one complication and reason, target, actor/action/affected option, full used-source review, both practice attempts and reasoned decisions, a rival about a shared uncertain link, different expected evidence, a cited second-case comparison and human assessment remain. Main word-count tolerance changes technical acceptance, not the expected content. Inclusion checks cannot determine whether an explanation is warranted.

VERIFICATION
1843 assertions pass: 1247 new focused checks plus 596 prior checks. Prior boundary expectations were updated to 49/50/60/85/120/121; old analytical omission/comparison/fidelity assertions remain. All 24 historical card bodies, 12 active and 12 legacy answer banks/feedback are retained. Exact original source JSON is preserved except the specified verification-note sentence. Source/build parity covers all cases/stages/reports/summaries. Test counts describe assertions, not learners or measured learning.

The six-page facilitator guide was updated and every page visually inspected. The latest ZIP is checked for CRC/integrity, exact standalone-build bytes and preserved historical review folders. Browser scope/limitations are separate in Round4_Retests. Two new expert reviews are AI static/code assessments, not first-use student trials. Previous first-use review PDFs describe their dated versions; this record supersedes their old hard word-limit instructions.


THE HEARING / TARGETED BROWSER RETEST
30 October 2026

This was an implementation check by an AI agent, not a new blind reviewer or human classroom trial. The isolated QA page had its own storage key and did not overwrite a learner's real workspace. Correct title and all four country choices were verified.

OBSERVED
- For class entry-name paste PASTE_TEST_ENTRY was blocked, leaving the field empty with a visible live notice. A normal t key entered successfully.
- In Connection, the Russia Ballot entry button inserted a bracketed target frame and selected a learner-filled slot. Finish kept that target To review.
- In Finish, separate rival and evidence-request buttons inserted their full writing frames. Unfinished helper slots remained To review.
- PASTE_TEST_RESEARCH did not replace the evidence-request frame in class mode; the typed frame remained unchanged and a live notice explained the rule.
- The main editor and inclusion check exposed both aim 60-85 and accepted 50-120. Comparison remained 30-45.
- Submission controls accurately displayed Print / Save as PDF and Copy grading summary, with browser-print and Canvas text-entry guidance.

CLIPBOARD FINDING AND FIX
Automatic copying did not complete in this cloud browser: the prior test string remained in its clipboard after both keyboard and direct click activation. This was not reported as a universal browser failure. The implementation was hardened so the full selected readonly summary appears synchronously before the clipboard promise resolves.

After rebuilding, keyboard activation showed the complete readonly summary immediately, with an honest DRAFT / NOT FILED COMPLETE label and all current learner fields. Manual Control+C copied the expected 2528-character draft summary; the browser clipboard readback started THE HEARING / GRADING SUMMARY and contained the rival/comparison sections. The final metadata-only addition labels mode and solo/shared-pair work; it was covered by the final engine suite, not another full browser replay.

LIMITS / INFRASTRUCTURE
The browser intermittently reported capture changes, mouse/key protocol timeouts and extension-metadata errors. Keyboard/DOM snapshot alternatives worked for the confirmed manual-copy route. Later navigation and screenshot capture could not be completed reliably. No screenshot or failed operation is represented as a successful final-page capture.

Actual fun-mode paste, text-drop delivery, final PDF generation/download, Canvas submission and a newly filed complete record were not verified in this browser round. Handler/state/markup tests separately cover mode policies, drop/preinput routes, source-face placement, all completion boundaries, import/history, print handling, both clipboard failure and unresolved permission, and exact report/summary content. Physical-device/accessibility use, classroom time and learning gains remain outside this check.


TARGETED REVISION PDF VISUAL QA - PASS

Read-only visual inspection completed for all six revised facilitator-guide pages:
- guide-1.png
- guide-2.png
- guide-3.png
- guide-4.png
- guide-5.png
- guide-finalpage.png, the fresh render of unchanged PDF page 6 supplied by root.

No visible overflow, clipping, overlapping elements, widows, orphan section headings, unreadable text, or footer collisions. Tables retain clear columns, row padding, readable wraps and consistent alternating fills. Typography and hierarchy remain restrained and readable.

Page 2: the revised 60-85-word aim and 50-120-word completion range fit cleanly in the existing paragraph. The closing-question section remains complete with clear separation above the footer.

Page 6: the added optional-frame instruction and PDF/text submission versus JSON-backup guidance fit cleanly. Both tables and the final independent-check note remain legible; there is no excessive wrapping or collision at the page ending.

Render-file exception: original guide-6.png was truncated at 57,344 bytes; view_image could not process it and Pillow reported Truncated File Read. This was not treated as a PDF fault. The replacement guide-finalpage.png was inspected and passes.

Scope: targeted visual QA of the revised guide only. This is not a new blind review, human trial, or learning-outcome claim. No source edits or regeneration performed by this reviewer.
