# Local launch draft — HOLD for external release

**Status:** local work produced, not published, not human-approved, not native
Hive activation. This is original draft copy and an approval-gated handoff.
No audience was contacted; no purchase, campaign, sign-up or analytics
collection occurred.

## Draft headline

**Make room. Not promises.**

Agenda Pocket is an offline planner for small agendas that tells you when
your commitments cannot fit—and proves a smallest permitted change when its
bounded exact search can do so.

### Three factual feature statements

1. **Keep every commitment.** Model task durations, availability, dependencies
   and fixed starts on one track. No task or break is silently omitted.
2. **See why it fails.** Proven conflicts name the constraints. Interrupted
   searches say unknown, rather than claiming impossibility.
3. **Negotiate explicitly.** Compare proven minimum-cost proposals under
   “fewest changed fields, then fewest minutes.” Approve a proposed cutoff or
   duration change before seeing the revised complete schedule.

### Draft CTA

With the owner's permission, open the local repository copy and try the
**synthetic Pocket Arcade tournament**. No sign-up or sharing is requested.
This invitation does not authorize a public post, hosting, contact or tracking.

## A truthful 30-second local demonstration

- Load Pocket Arcade: 85 mandatory minutes, a 75-minute initial cutoff, two
  preserved breaks and a fixed check-in.
- Read the proved overload and the disclosed objective.
- Choose a one-field, 10-minute permitted proposal. The cutoff can move from
  15:15 to 15:25 within already declared synthetic availability, or one round
  can be explicitly renegotiated to its allowed floor.
- See all seven commitments, then export the labeled local JSON report.
- Load the midnight case with a one-node bound to show **unknown**, not a
  manufactured impossibility.

This is a text walkthrough, not a recorded demonstration, generated video,
voiceover, actual tournament, observed user trial or proof of usefulness.

## Limitations that belong beside the CTA

- At most eight mandatory, uninterrupted tasks; one resource/track.
- Declared discrete time grids, at most 120 slots through the full repair
  envelope, one day, no calendar or date interpretation.
- “Smallest” is relative to the explicit repair policy and lexicographic
  objective. It is not best value, highest priority, or least work lost.
- Shorter durations require real scope/timebox agreement. Availability is
  user-declared, not externally verified.
- A bounded search may return unproven; unsupported input is not solved by
  dropping fields.
- No autosave; export explicitly. No guaranteed browser/extension privacy.

## Evidence-linked decision recommendation

| Gate | Current evidence/state | Recommendation |
|---|---|---|
| Declared product boundary | README, API contract and task mapping | Local draft produced; no owner approval invented |
| Numerical behavior and minimality | Actual Node logs, independent oracles, adversarial fixture outputs in `evidence/` | Use recorded results, not this prose, as pass/fail evidence |
| Packaging and attribution | Generated-bundle equivalence checks, source hashes, MIT source data | Local candidate only |
| Browser, keyboard, errors and layout | `manual-checks.json` | **Not observed**; mandatory release gate remains open |
| Screen reader/accessibility | `manual-checks.json` | **Not observed**; no certification claim |
| Support | FAQ and optional synthetic issue-intake proposal | Unreviewed draft; no staffed service |
| User usefulness | No participants, responses or research | **Not validated** |
| Distribution | No approved channel/audience/publication content | **HOLD** |
| Native organization/team activation | Initialization read as inert metadata only | **Not performed or authorized** |

An unresolved mandatory gate means no external go-ahead. Work-produced is not
reviewed, accepted, activated or launched.

## Reversible future delivery proposal — not executed

The only proposed initial surface is an **owner-reviewed local file bundle**
for a bounded internal walkthrough. Before any later public surface, separately
name the exact content, audience, owner and approval. Keep tracking off. Verify
the same artifact hashes and behavior after delivery. If verification fails,
withdraw that copy rather than inventing success evidence.

Public static hosting, outreach, paid channels, purchases and media production
each require their own explicit authorization; none is requested or executed
by this workstream. A future publication owner would also need an actual
withdrawal/takedown procedure for the selected surface. There is no current
hosted URL or publication receipt to report.

## Hypothesis and disconfirmation, not traction

Hypothesis: people planning a small meeting or game session may benefit from
seeing a contradiction and a bounded renegotiation instead of silent omission.
No research establishes that audience or benefit.

In a separately authorized, consented synthetic walkthrough, disconfirm the
interaction if a participant cannot distinguish infeasible from unproven,
believes a duration reduction completes unchanged work automatically, or
mistakes a proposed repair for an already-booked schedule. Comprehension,
task completion and perceived usefulness should be asked separately. No
responses, savings, conversion rates or customer quotes have been collected.