# Tara scheduling and reminders · delivery receipt · 8 September 2026

The [fifteen-slide guide](../../decks/tara-scheduling.html) and
[PDF](../../decks/tara-scheduling.pdf) follow an evening return through candidate
selection, local clocks, quiet hours, reminders, storage and delivery evidence.
Native decisions, time tables and record specimens explain the mechanisms.
[Authoring](../../authoring/v1-tara-scheduling.md) records the story and source
boundaries. The implementation baseline is
`c0001c6c3f5f26db07f90bb30569d314579ffacf`; the source pin remains
`1ceae52c7896a59ac74a4429d410f3218f0b91c6`.

All 644 earlier slides, script entries, narration manifest entries and MP3 bytes
are preserved. Fifteen new tracks add 1,336.191 seconds (22.3 minutes). The current
library has 64 guides, 659 unique slides, 682 guide-slide instances and 386 cited
paths. No source unit is assigned merely because it is cited: reviewed coverage
remains 80, with 10,018 structural and 103,246 unassigned units. The combined
ritual/steps/scheduling feature unit still needs its remaining practice and
assembly explanation.

## Verification actually performed

- Sixteen immutable-source VM cases execute the complete scheduler and complete
  AST-selected reschedule function. Controlled clocks, explicit storage and
  actual scheduler dependency characterize the numerical examples and failure
  cases. Node 24.13.1 / ICU 78.2 / timezone data 2025c is the recorded runtime.
  These are source characterizations, not live deliveries or predictions about
  future timezone legislation. Targeted ESLint and Ruff pass.
- The settled presentation unit/preservation checks pass (108 cases), including
  all earlier assembled slides and speech. Generated artifact and coverage
  freshness pass. The five cited source files are byte-identical from the pin
  through the implementation baseline and this delivery.
- The final [browser report](browser/report.json) passes 111 checks across
  desktop, compact desktop, 390/320-pixel mobile, expanded/collapsed/mobile center
  embedding and six enlarged-text cases. Sixty captures, 45 embedded axe checks
  and three deliberately rejected diagram mutations are retained. No document
  overflow, fit finding or runtime error remains.
- The complete focused [center review](center.json) passes all fourteen checks,
  covers navigation/search/history/mobile focus and embedded fit, preserves
  63 previous guide HTML files and four shared UI assets, checks the first/last
  audio boundary and exports the complete fifteen-page PDF. Thirty embedded
  captures are retained. Browser audio output was disabled.
- The [PDF text report](pdf-visual-text.json) checks exact visible-body labels
  and order for every page. Model optical review is recorded separately below;
  text extraction alone does not prove visible ink or legibility.
- The fifteen new recordings receive encoded-audio/pause validation during
  generation and a final scoped decode check. The prior 644 tracks reuse their
  baseline decode evidence after exact manifest, script and MP3 comparison,
  recorded in [speech continuity](speech-reuse.json). All 659 final file hashes
  are checked. This is not a fresh decode of every earlier track.

## Optical review and corrections

All fifteen compact-desktop draft slides were optically inspected. Six received
composition polish: duplicate authored arrow glyphs were removed, clock text
was made easier to scan, and long record/query labels were shortened. All six
revised captures were re-inspected at original size. The full final matrix then
found a 320-pixel overflow in the plan's long type-name heading. It now reads
Schedule plan; the exact type remains in the explanation. Notes and speech did
not change. Failed and draft attempts remain in [attempts](attempts/).

The final record-heading desktop, selected mobile layouts, complete PDF sequence
and representative full-size PDF pages were reviewed by the model. Their exact
files and review scope are recorded in [optical review](optical-review.json).
This is model optical inspection, not an independent human comprehension study.

## Reproduction and remaining work

Run `python3 docs/presentations/presentation-center/verification/v1-tara-scheduling-2026-09-08/verify-receipt.py`
from the repository root. The verifier binds committed inputs, artifacts,
captures, retained raw logs and every recording. Later changes invalidate this
frozen delivery receipt; use its publication revision for historical evidence.
The final passing command set is recorded in [state](state.json); the failed
320-pixel attempt is retained separately and is not counted as a pass.

The [full V1 program](../../authoring/v1-completion.md) remains active. Tara
practice/player, companion, mentor and workbench depth are unfinished, as are
the other V1 workstreams in that ledger. Normal-speed human listening, human
reader review, live-member notification/device acceptance and production
deployment validation were not performed. This receipt closes this technical
presentation increment only.
