Hold The Horde — Progress
Updated 2026-09-08 17:32 Jerusalem (UTC+3). PM and GPT work continues; only the two Claude roles are paused. All work remains design only.
21 tracked design packages: 19 preparation-closed, 2 open. Equipment20 closed400 at17:29; new outcome21 Traps and Encounter Progression was explicitly assigned401. Shared integration04 remains dependent. The published Design Coverage map is separate management work, not another counted package. These are bounded written outcomes, not complete game domains or an overall completion percentage.
“Closed” means the bounded written design/review outcome is complete. It does not mean a feature is implemented, every detail approved or fun validated. This board covers the current design batch; it is not a complete campaign or production backlog, and the closed/total ratio is not an overall game-completion percentage. Stable IDs stay fixed; added/reopened work must be recorded explicitly.
Availability correction —8 Sep: the owner paused only Claude, and explicitly confirmed PM and the existing GPT Systems designer should continue. PM’s broader stop was an incorrect inference and has been withdrawn. Systems is assigned the next trap/encounter outcome, a temporary GPT reviewer has saved fresh criteria, and PM has published the full-v1 coverage map. Neither permanent Claude role will be woken or replaced before owner confirmation. Active one-minute fallback monitoring and15-minute progress/forecast reports continue.
Forward forecast — baseline F1
Planning answer: a coherent core design this week, a complete first-pass game-design wiki in about two weeks, and a detailed campaign specification in roughly two to three months. These are PM estimates to manage against, not promises or measured confidence intervals. They will change as the content inventory and owner choices become concrete.
All dates are2026; times are Jerusalem. Ranges are cumulative from8 September, not durations to add together. Baseline F1 was set8 September. Dates assume ordinary weekdays, about6–8 productive agent/PM operating hours per working day, functioning local hosts, and most consequential owner choices within a working day. That does not mean6–8hours of the owner's time or guaranteed overnight operation.
| Milestone | Target | Current likely window | Confidence | What will be ready |
|---|---|---|---|---|
| M0 — current puzzle/rescue decision pack | Achieved8 Sep15:59 | Original target16:30; range16:30–17:00 | Actual completion | Both written comparisons reviewed and closed; owner choices remain separate |
| M1 — whole-game outline and scope bands | Achieved8 Sep16:20 | Original18:30; revised17:00 | Actual completion | Draft369, focused review370, response374 and PM integration378 complete; architecture, scope bands and missing-work inventory prepared |
| M2 — coherent core-design baseline | Fri11 Sep | 10–14 Sep | Medium-low | One compatible proposed/selected loop: opening, builds, combat, oxygen, death/shift, rescue and home; old conflicting rules replaced and consequential choices dispositioned |
| M3 — complete first-pass game-design wiki, v1 | Tue22 Sep | 18–29 Sep | Low-medium | Every major gameplay and campaign area has an actual design, a scoped content inventory, dependencies and a usable development handoff outline; major categories are no longer only menus of alternatives |
| M4 — detailed campaign/content specification | Tue10 Nov | 20 Oct–1 Dec | Low | The approved finite inventory of regions, rooms, puzzles, clues, encounters, rewards, cast scenes and endings is specified through its state changes and recovery paths |
Will the full wiki be ready today? No. Today's decision pack and whole-game outline are prepared; they are not a complete design wiki. The working target for the first complete design wiki is22 September. The later detailed-content milestone expands that design into the actual authored campaign. Neither milestone claims that a written plan has proved fun, balance, performance or30/90/100–200hours of play.
For M3, “complete first pass” includes the story's causal chain and ending structure; station layout/shifts/persistence; equipment acquisition and progression; combat/horde/trap families; oxygen/resources; rescue/colony/cast arcs; both language-learning arcs and clue distribution; journal, controls and accessibility; content budget, production constraints and the later validation plan. Numerical tuning can remain explicitly provisional. A major domain marked only “TBD” does not satisfy completion.
M4 has the weakest estimate because no approved content inventory yet exists. It assumes a substantial authored campaign matching the brief's ambition and two additional specialists. It is approximately30–60 active working days; with the current team alone, the initial audit estimated40–80. A smaller approved campaign could be much quicker; a dense secret/language campaign or a changed global rule could exceed the range. I will revise M4 after the inventory and three representative detailed content packages are written and checked. Final gameplay validation belongs to a separately authorised implementation/playtest phase; it has no honest calendar endpoint yet.
Estimated work timeline
This is a planning schedule, not historical delivery data. “Buffer” is allowed revision/decision time, not an unassigned promise to polish indefinitely.
| Workstream | 8 Sep | 9–11 Sep | 14–18 Sep | 21–29 Sep | Oct | Nov–1 Dec |
|---|---|---|---|---|---|---|
| Current decision pack | Done15:59 | |||||
| Whole-game outline and scope | Done16:20 | Use for domain design | ||||
| Core rules and shared integration | Finish / buffer into14 Sep | |||||
| Campaign, puzzles, cast, progression and languages | Start | Develop | Integrate and review | |||
| Complete wiki v1 | Draft coverage | Target22 Sep; buffer to29 Sep | ||||
| Detailed authored content | Representative packages | Expand approved inventory | Main authoring | Target10 Nov; buffer to1 Dec |
The critical sequence is core decisions and interfaces → whole-game architecture/content budget → domain designs → cross-domain integration → v1. Content detail proceeds in parallel only where interfaces are stable. Extra writers do not eliminate the PM/review bottleneck.
Next deliveries: Traps and Encounter Progression draft target18:00 today(range17:45–18:30), reviewed outcome18:45(range18:15–19:15). Equipment Catalogue closed17:29 against its original reviewed18:30 target; PM’s Design Coverage was published17:27 against17:45. Core11 Sep(10–14), complete first-pass wiki22 Sep(18–29), and specialist setup9–10 Sep remain unchanged.
Extra agents — when and why
Plan to add two agents tomorrow,9 September, with bounded briefs prepared today. The current team can finish today's work. We will begin the two specialist streams as the relevant scopes become ready, aiming to have both available by10 September rather than waiting a week.
| Additional role | Planned start | First bounded responsibility | Reason for adding it |
|---|---|---|---|
| Campaign and puzzle designer | 9 Sep | Develop the outline's discovery dependencies and two representative regional/puzzle chains, keeping owner mysteries revisable | Most existing work is local examples; the campaign and its deep discoveries need a dedicated author |
| Technical design and production planner | 9–10 Sep | Audit proposed scope against Unity/Steam, state/save complexity, reusable content needs and available free-asset categories; identify constraints and likely expensive systems | The largest forecast uncertainty is how much authored content and engineering complexity the design implies |
These are design roles. No gamecode, prototypes, simulations, tests, purchases or AI art are authorised by this staffing plan. The owner has not yet created these sessions; no staffing completion is claimed. PM will guide setup one step at a time. Existing Systems retains the equipment economy; if that becomes the actual bottleneck, the second specialist brief can shift to economy/content planning rather than accumulate an idle role. No additional permanent critic is currently planned.
The independent schedule audit proposed campaign/puzzle and economy specialists after the core baseline. PM keeps its point that extra agents do not accelerate core decisions, but brings onboarding forward to9 Sep so a bounded content/production audit can begin alongside integration. The production role is chosen first because the absent content inventory and feasibility assumptions dominate the long-range estimate. This is a reversible staffing judgment, not a demonstrated speedup.
Why these estimates, and when they change
Recent focused comparison/review packages took about16–49 elapsed minutes. That supports today's short delivery windows; it does not establish campaign-authoring throughput. The independent remaining-scope audit estimated2–4 active days for a coherent core,6–12 with two specialists or8–15 with the current team for a whole-game first pass, and30–60 with two specialists for detailed content. PM has allowed calendar/integration buffer around the v1 target.
- Every15-minute progress report includes the next milestone's current target/range and whether it moved. The full forecast is reviewed at each day's close, with the next substantive re-estimate by9 Sep18:00 after the scope outline.
- Keep the original target visible when revising it. Record old estimate, new estimate and the concrete cause; do not silently move dates.
- A changed core mechanism or late owner story change may add roughly1–3 working days to the core/v1 forecast; quantify the actual affected work when it happens.
- A blocked decision moves its dependent work, not every parallel stream. Host outages are reported with elapsed delay. If authors produce a queue waiting over a working day for review/integration, reduce parallel intake before adding more agents.
- Forecasts must be useful even when confidence is low. Uncertainty is a reason to widen a range and state assumptions, not to omit an estimate or wait for certainty.
| Forecast revision | Recorded | Change and reason |
|---|---|---|
| F1 | 8 Sep | First forward baseline added at owner's request; prior board had actuals but omitted estimates. M3 target22 Sep; M4 target10 Nov. |
| F1.1 | 8 Sep16:04 | M1 reviewed outline moved from18:00–20:00 to16:30–17:30 because the actual draft arrived16:01 and Reviewer had already fetched it. Review/rework still required. Core, full-wiki and detailed-content dates unchanged. |
| F1.2 | 8 Sep16:20 | M0 actual15:59 and M1 actual16:20 recorded. Both prepared ahead of forecast. M2–M4 dates unchanged: the outline is a scaffold, and scope/unit reconciliation is still required before rolling up its partial111–208 role-hour inventory. |
| Availability correction | 8 Sep17:10 | Only Claude paused; PM/GPT work resumed after PM incorrectly stopped broader intake. Equipment Catalogue20 now targets reviewed18:30 today. M2–M4 dates unchanged pending coverage reassessment. |
| F1.3 | 8 Sep17:32 | Catalogue actual17:29 and coverage actual17:27 recorded ahead of their18:30/17:45 targets. Next reviewed trap/encounter outcome targets18:45(18:15–19:15). M2–M4 unchanged: the coverage audit still identifies eight partial domains and three missing contracts; remaining-work estimates need deduplication. |
What changed in this report
- Equipment Catalogue20 is preparation-closed400 after the actual draft396, independent GPT review and one author response399. Twelve reusable parts and three legal rig examples are prepared. The splitter now has an explicit provisional payload rule; targeting, readiness and physical effects are clearer.
- New outcome21 gives Systems five trap/mechanism families and a four-region encounter-pressure map. It continues useful design work under the Claude-only pause.
- Design Coverage maps fifteen domains: four with usable first-pass rules, eight with examples/alternatives and three missing dedicated contracts. Those categories are not a game-completion percentage.
- No new owner choice, game implementation or permanent agent creation is claimed. Full-wiki forecast remains22 Sep(18–29).
Open and closed outcomes
All milestone times below are Jerusalem time. “Updated” separates later owner-rule integration from the original preparation closeout.
| ID | Outcome | Preparation status | Decision status | Latest milestone | Next step / remaining scope |
|---|---|---|---|---|---|
| 01 | Reusable equipment combinations | Closed | Selected direction | Sep7 22:19; updated Sep8 13:45 | Catalogue and numerical rules remain proposed |
| 02 | Oxygen visibility and journal pause | Closed | Selected direction | Sep7 20:55; updated Sep8 13:45 | Exact oxygen amounts remain open |
| 03 | Carried property returns on death | Closed | Selected rule | Sep8 13:45 | Shift and survivor ordering tracked separately |
| 04 | Shifts, persistent work and historical records | Open — dependent | Binding direction selected | Sep8 14:49 direction recorded | Integrate final station contract after15; older fixed-map clauses are superseded |
| 05 | Ordinary combat and horde encounters | Closed | Selected direction | Sep8 10:51; updated13:45 | Automatic ordinary fire, manual specials; balance remains open |
| 06 | First-survivor comparison | Closed | Short rescue selected; person unchosen | Sep8 response364/disposition368 | Yeva versus Aster ready; short-route condition and death/refuge timing remain explicit limits |
| 07 | Practical discovery and optional help | Closed | Selected direction | Sep8 11:23; updated13:45 | Current routes cannot be inferred from historical records |
| 08 | Cooperation and useful work aboard | Closed | Tentative preference | Sep8 12:00; updated13:45 | Equipment loans tentative; full colony/cast still open |
| 09 | Two-language expedition and journal use | Closed | Reviewed proposal | Sep8 13:45 | Detailed language system unselected;16 is a distinct evidence scene |
| 10 | How equipment is acquired and improves | Closed | Reviewed alternatives | Sep8 13:51 | Find working components versus observe/fabricate; owner choice later |
| 11 | Three distinct horde creatures | Closed | Reviewed proposals | Sep8 14:07 | Roster and origins unselected |
| 12 | Hospital story and its causal alternatives | Closed | Reviewed alternatives | Sep8 14:47 | Owner hospital/godhood draft leads; exact story/ending remain revisable |
| 13 | Player resurrection and corruption | Closed | Reviewed alternatives | Sep8 14:38 | No player decay or death budget adopted |
| 14 | Opening expedition: reusable machinery | Closed | Option A selected | Prepared Sep8 15:14; selected15:30 | Preferred direction recorded; exact starting items/rooms remain proposed |
| 15 | Whole-station mechanism comparison | Closed | Goal selected; mechanism conditional | Sep8 15:51 disposition363 | Physical connection candidate prepared; owner payoff choice and detailed content remain open |
| 16 | Testable alien-language interaction | Closed | Reviewed proposal | Sep8 15:38 | Protocol discovery prepared; does not prove request semantics or a god |
| 17 | Whole-game design outline and scope bands | Closed | Reviewed planning scaffold | Sep8 16:20 disposition378 | Use finite bands and dependencies for domain designs; story/ending/count choices and estimate coverage remain open |
| 18 | Player information and accessibility | Closed | Reviewed proposed information contract | Sep8 16:35 disposition387 | Full cue/screen structure and accessibility specification remain later work; no usability claim |
| 19 | Survival and loadout | Closed | Reviewed provisional resource contract | Sep8 16:45 disposition392 | Detailed variants and balancing remain later work; integrate compatible core rules after owner decisions |
| 20 | Equipment catalogue | Closed | Reviewed12-part planning sample; acquisition unchosen | Sep8 17:29 disposition400 | Exact thresholds, multi-actuator controller rules, acquisition and later finite catalogue remain open |
| 21 | Traps and encounter progression | Open — authoring | Five families/four-region SMALL proposal | Sep8 17:29 assignment401 | Draft18:00(17:45–18:30); one focused GPT review/response; reviewed target18:45(18:15–19:15) |
Recent delivery timeline
These are observed calendar spans from assignment to preparation closeout, including handoffs and review. They are historical measurements, not estimates of labour or promised future turnaround.
| Outcome | Assigned Sep8 | Preparation closed Sep8 | Elapsed |
|---|---|---|---|
| Equipment progression comparison | 13:17 | 13:51 | 34 minutes |
| Three horde creatures | 13:51 | 14:07 | 16 minutes |
| Hospital-story comparison | 13:58 | 14:47 | 49 minutes |
| Resurrection/corruption comparison | 14:20 | 14:38 | 18 minutes |
| Opening comparison | 14:40 | 15:14 | 34 minutes |
| Alien-language interaction | 15:06 | 15:38 | 32 minutes |
Work sequence
| Stream | Now | Next |
|---|---|---|
| Equipment catalogue | Preparation closed400; final737b3cfd | Use in trap/encounter examples; detailed controller and acquisition integration later |
| Whole-v1 coverage | Published398; fifteen-domain map checked | Use unique deliverables to reconcile scope/forecast; PM validation-handoff plan can proceed |
| Trap/encounter progression | Systems401 receiving/source reads verified17:31; GPT criteria saved | Inspect actual submission, one focused review/response, PM disposition |
| Shared integration04 | Current owner rules known; some mechanisms/choices pending | Integrate confirmed rules and explicitly isolate unresolved decisions |
| Claude Lead/Reviewer | Paused by owner for token limits | Resume only when owner confirms; GPT work continues meanwhile |
| Specialist setup | Two bounded briefs ready | Planned9–10Sep; no new permanent sessions created |
The available team continues useful work. No manual message forwarding is needed, and a paused provider does not suspend other authorized agents.
Decisions for the owner
Already answered: opening A and persistent room/group connections across shifts, culminating in connecting the entire station.
Open creative question: should completing the entire connection puzzle finally stabilise the layout, or should rooms keep moving while every established connection survives? PM recommends stabilisation as a reward for mastery. This is a proposed late payoff, not an answer inferred from the opening choice; the team can continue its comparison while it is pending.
The first-survivor choice is also already with the owner: Yeva, the pressure technician preserving ward measurements (PM recommendation), versus Aster, the supply clerk suspicious of the arrival and requesting one reversible door closure. Neither is selected. Progression, creature, story and corruption alternatives are prepared for later reactions; the owner is not expected to review the whole wiki at once.
Reporting agreement
PM sends a compact report every15 minutes while useful work is active, plus meaningful completion, disagreement, blocker or decision updates. The Claude-only pause does not stop reporting while PM/GPT work is active. Use quiet hourly checks only when no useful authorized work can continue. Each report gives the timestamp, changes since the previous report, numeric open/closed preparation counts, real worker activity, blockers, next deliverable with ETA, current v1 target and any forecast change, and a link here. An unchanged interval is stated honestly. No page/message counts or invented completion percentages stand in for progress. Clearly labelled estimated dates and ranges are required; never present them as known facts or guaranteed deadlines. The local monitor requires the computer and apps to remain available.