Appendix B – Decisions Log & Remaining Open Items
On October 12, 2026 Patrick asked for every open design question to be decided, using this document’s own recommended defaults plus his explicit calls (marked P). The 35 questions of the v0.3 Appendix B and the four questions of Section 30.6 are closed below; row 1 (title) was revised in v0.5, and row 39 (tech stack, Patrick’s decision) was added in v0.6. v0.7 (October 13, 2026): Patrick’s AI-leadership decisions add rows 40–42 (P) and design rows 43–49; rows 20, 21, 22 and 31 are revised as superseded or updated; open items R7–R11 are added. Decisions set design direction; numeric values keep their [ASSUMPTION] tags and are still tuned in prototypes, bot simulations and alpha.
B.1 Decisions Log
| # | Question | Decision (v0.4; row 1 revised v0.5; row 39 added v0.6; rows 40–49 added and 20–22, 31 revised v0.7) | Rationale |
|---|---|---|---|
| 1 | Title (revised v0.5) | P (v0.5): the game and its world are called Harimu – also the in-world name of the planet. The v0.4 working title Body Politic is rejected. ‘Nations’ stays the internal codename. Formal clearance stays open (R1). | Patrick’s decision; one name for game and world; no game titled Harimu found in quick checks (Appendix A). |
| 2 | Four new nations | Approve Hoarvale, Varkand, Veldwood and Fulmara; no swap. Managed democracy and emirate remain only candidate charter templates (Section 30). | Each adds a transition mechanic no other nation has (Section 1.2.1); the alternatives overlap existing nations. |
| 3 | Names after the element removal | P: keep all nation and capital names for now; rename only if a trademark issue becomes concrete. Kindle is flagged as the likely rename candidate (fallbacks prepared – R2). | No name depends on an element; renaming costs art and text; Kindle has the closest brand proximity. |
| 4 | Ten currency regimes | Approve all ten as designed (Section 12.3), including Aurelle’s debasement (≤ 15% per game year) and Veldwood’s referendum-only people’s issuance; adjust strength by tuning, not redesign. | Each regime expresses its economy; both powers are already bounded by the hard caps. |
| 5 | Reserve unit | P: keep the neutral, system-run Tessera issued by the Commons Clearing House; no player currency becomes the official reserve. | A neutral yardstick no rival can capture or freeze; fewer exploits. |
| 6 | Economic realism | Keep the simple, visible inflation rule (money ÷ output, with a lag) and fixed emission caps. | Readable for players and tunable; a fuller macro model would be opaque. |
| 7 | Hyperinflation | P: keep hard caps (15% per game month, 30% per game year): a full hyperinflationary collapse cannot happen, but reckless printing still causes painful, compounding inflation; redenomination stays cosmetic (Section 12.4). | A political disaster for rulers, never a broken server or ruined savers (Pillar 4). |
| 8 | Citizen FX access | P: citizens may hold Tessera and trade currencies through licensed exchanges; limits in controlled economies (Confluence and Meridian quotas, Aurelle commoner caps, Ilmar Clear Measure, Varkand Amber quota and Red freeze); none in Korund, whose non-convertible system is fixed (Section 12.8). | Opens currency play to every citizen while each system keeps its character. |
| 9 | Sanctions threshold | Keep: suspension from Tessera settlement only by a multilateral treaty covering ≥ 50% of settled world trade, ≤ 60 real days, renewable; humanitarian goods always settle (Section 13.3.1). | Keeps suspension rare and collective; never starves ordinary citizens. |
| 10 | Black markets | Foreign smugglers may trade Labor Notes and other cash bundles at the street rate, within per-citizen illicit caps and seizure risk; street rates never enter official settlement (Section 12.9). | Matches the smuggler and moneychanger careers; caps prevent an arbitrage machine. |
| 11 | Dormant savings | P: modest protection – the Dormant Savings Shield covers half of inflation on small home-currency balances during dormancy (Section 12.8). | Absence has consequences, but not ruin (Pillar 4); the cap and treasury funding stop farming. |
| 12 | Faith roster | Approve the six invented faiths unchanged, including the Open Hand’s reform politics; no seventh faith before 1.0. | Covers the main archetypes; each has a holy site; more faiths add content and sensitivity cost. |
| 13 | Stance immutability | P: each nation’s religious stance is fixed, like its government (Section 14). | Pillar 2 – systems are fixed; people are not. |
| 14 | Korund suppression | P: underground worship stays playable but low-key – abstract discovery, Suspicion first, capped raids and Correction Duty (Section 14.6). | Keeps Korund’s tension while lowering sensitivity risk. |
| 15 | Ilmar’s reach | Keep Counsels, the Great Pilgrimage (once per game year) and the Veil (once per Age per target) as designed; tune in Wave 3 playtests. | A rare, non-compelling foreign-policy tool that makes the theocracy distinct. |
| 16 | Missionaries | P: cross-border missionaries at launch (from Early Access) with cooldowns: one invitation per 14 real days, one faith change per 60 real days, full mute (Section 14.8). | A social bridge between nations; cooldowns and muting stop spam and harassment. |
| 17 | Faith rewards | P: Devotion gives social and ceremonial rewards only – never production or combat. | Avoids a ‘best religion’ meta. |
| 18 | Early Access roster | P: open Early Access with the MVP three, then add nations in waves. | Earlier feedback and revenue; proves modules before content scales. |
| 19 | Wave order | P: keep Wave 2 = Confluence, Hoarvale, Varkand; Wave 3 = Meridian, Ilmar, Veldwood, Fulmara. | Cheapest first: Wave 2 reuses the most existing modules. |
| 20 | Under-populated nations (revised v0.7) | v0.4 (superseded): AI Caretaker clerks for non-leadership posts. Now (row 49): AI citizens fill population and AI officials hold every office, so Caretakers are removed; the viability floor drops to ~25 weekly-active humans; shard merges only at Age boundaries, as a last resort, preserving nationality (Section 19.2). | Leadership no longer depends on human numbers; one-nation stays true. |
| 21 | Fulmara franchise (revised v0.7) | Keep share-weighted voting (shares bought only with earned Charge; 5% cap); no base vote. v0.7: citizen-held share votes bind the AI board (Council of Ten); the Tribune of the Docks is an AI elected one-person-one-vote by non-shareholders and stays the counter-power. | Preserves Fulmara’s unique fantasy under AI leadership; anti-pay-to-win guards already in place. |
| 22 | Varkand intensity (revised v0.7) | Keep as designed: National Service tutorial; Red Alert ≤ 14 consecutive real days, then 7 at Amber or lower; coups with a 14-day cool-down. v0.7: marshals are AI; a coup installs a rival AI marshal when enough citizens and soldiers back it (Section 10.3); the junta system never changes. | Distinct and already capped; sensitivity review still applies. |
| 23 | Hoarvale Crown Etiquette | Fine-level only; no stricter lèse-majesté variant. | Hoarvale is the free-society mirror image of Aurelle. |
| 24 | Veldwood cadence | Keep Vote Days every 15 real days with up to 4 items; monitor turnout from Wave 3, with 30 days as a tuning fallback. | Gives direct democracy a steady rhythm; the item cap limits fatigue. |
| 25 | Player-founded nations (Section 30) | Keep as an optional post-launch feature behind a go/no-go review after 1.0 and one full Age (details in rows 36–38). | Architecture stays open without adding launch scope. |
| 26 | Re-selection grace period | P: allow one free nation change within the first 48 real hours and before any political action; it does not count as the lifetime change (Section 5.2). | Forgives onboarding mistakes without enabling spying. |
| 27 | Defection | P: otherwise one account, one nation; add the restricted Exile & Asylum mechanic as post-launch (Section 16.2); reject one citizenship per shard. | Keeps permanence meaningful; extra anti-spy rules address the risk. |
| 28 | Time compression | P: keep 12× (1 real month = 1 game year). | Terms long enough to matter, short enough for drama. |
| 29 | Wipes | P: fully persistent world – no seasonal shards or resets; new shards may open later (Section 6.3). | Persistence is core to the fantasy; new shards give late joiners a fresh start. |
| 30 | Mortality | P: only characters age and die; heirs continue on the same account and nation (Section 5.1). | Gives dynasties and turnover without wiping anyone’s progress. |
| 31 | Korund intensity (revised v0.7) | Accept as designed: Correction Duty ≤ 7 real days, always-playable loops, information fog with a hidden-radio escape. v0.7: penalties are imposed only by AI organs under rule triggers and Charter Limits (Section 8.18); the AI Chairman rules for life and responds to quotas, loyalty and unrest, not votes. | Strong fantasy with Charter Limits that protect players. |
| 32 | Theocracy | Keep the fictional Lumen Canon as Ilmar’s state faith (no secular ‘philosophical order’). | A theocracy needs a faith; invented, respectful and sensitivity-reviewed. |
| 33 | MVP trio | Kindle + Korund + Aurelle; no fourth MVP nation (v0.7: all three with AI leaders and AI citizens). | Most contrasting experiences; covers all three transition types. |
| 34 | Minimum age | P: 16. | Political themes, open chat, coups and purges. |
| 35 | Budget path | P: lean solo + contractors build first; revisit funding a small core team only if Early Access traction warrants it (Section 28.1). | Lowest burn and risk while the core loop is proven. |
| 36 | Section 30: templates | Start with the ten existing systems; new templates only after full design, balance and sensitivity review. | Keeps every nation well designed and balanced. |
| 37 | Section 30: which shards | Allowed on running shards under the Section 30.5 caps; new shards may enable it from day one. | Gives running worlds new stories while caps limit splintering. |
| 38 | Section 30: Founding Choice | Keep it as the single lifetime change, shared with Exile & Asylum. | Lets veterans found nations without making nationality fluid. |
| 39 | Tech stack (new v0.6) | P (v0.6): client app in Next.js (App Router) with TypeScript/TSX; Clerk (@clerk/nextjs) for authentication; Three.js kept for the world, via React Three Fiber + drei (plain Three.js acceptable). Separate authoritative Node.js/TypeScript realtime game server on Fly.io/Railway verifying Clerk JWTs; PostgreSQL (Drizzle) + Redis + pg-boss; Turborepo + pnpm monorepo. Clerk Organizations not used for nations (Section 22). | Patrick’s decision; one TypeScript codebase; managed auth with phone verification and bot protection; nation state stays server-authoritative; Vercel is not suited to long-lived authoritative sockets. |
| 40 | Who leads nations (new v0.7) | P (v0.7): every nation is led by AI (‘AI/SI’). AI leader personas hold every leadership office and make all governing decisions (diplomacy, war, budgets, sanctions, trade, laws, city founding, migration approvals). Humans are citizens only and can never hold leadership office (Sections 5.6, 8). | Patrick’s decision; removes leader inactivity, office griefing and alt-account office capture; politics runs at a steady cadence. |
| 41 | Citizen influence per system (new v0.7) | P (v0.7): democracies (Kindle, Hoarvale, Veldwood) hold real elections between AI candidates and parties, with a Hoarvale parliament of AI parties forming coalitions and binding Veldwood referendums and initiatives; party-states (Korund, Meridian, Confluence) ignore votes and respond to quotas, loyalty and unrest (Korund AI leader for life with succession on persona ‘death’; Confluence AI congress factions; Meridian paramount AI leader and AI committee); Aurelle and Varkand AI rulers can be overthrown by enough rebelling citizens or citizens backing an AI rival claimant or marshal – a coup installs a new AI ruler, never a new system; Ilmar’s AI Oracle weighs piety and pilgrimage, with an AI council of elders and an assembly of vetted AI candidates; Fulmara’s AI board follows citizen share votes and its AI Tribune is elected by non-shareholders (Sections 8.5–8.17, 9, 10). | Patrick’s decision; each system keeps a distinct, real human lever. |
| 42 | AI citizens (new v0.7) | P (v0.7): AI citizen agents populate every nation alongside humans – mostly lightweight rule-based agents plus 8–15 LLM-backed Notables per nation, all visibly labeled AI (Section 11.7). | Patrick’s decision; living nations at any human population. |
| 43 | Human primacy rule (new v0.7) | AI citizens never sign petitions, initiatives or pledges; their votes form one Simulated Public bloc capped at the smaller of 25% of counted weight and one-third of human votes; collective-action thresholds count Tier-1 humans only; AI share of Unrest capped at 40% (Section 11.9) [ASSUMPTION – values]. | Humans must matter more than simulated citizens (Pillar 6). |
| 44 | Verification tiers (new v0.7) | Tier 0 = play; Tier 1 (phone) = voting and every counted civic act; Tier 2 removed (Sections 20, 22.3). | No office to gate; voting integrity is the remaining high-value target. |
| 45 | Leader inactivity protocol (new v0.7) | Removed, with Founding Caretakers; AI leaders are never inactive; an AI outage falls back to the utility model (Section 8.19). | Obsolete under AI leadership. |
| 46 | Transparency default (new v0.7) | Per system (Section 8.6): public Reasoning Briefs in democracies; private rationale published after 30 days elsewhere where the system allows; Korund shows state media only; raw model reasoning is never shown. Exact visibility stays open (R8). | Visibility should express each system while protecting against gaming. |
| 47 | Decision pipeline (new v0.7) | Hybrid: deterministic rules engine and legal action set → utility model for routine decisions → LLM for major deliberation, speeches and diplomacy; game server re-validates and executes; decisions logged and replayable (Sections 8.3, 22.12). | Safe, deterministic, cheap and still in character. |
| 48 | AI provider strategy (new v0.7) | Provider abstraction with at least two adapters; personas and memory in our database; two model tiers (mid/frontier for governance, small for chat); provider choice open (R7); switch only after shadow mode (Section 22.15). | Avoid lock-in; prices and models change often. |
| 49 | Population floor (new v0.7) | Viability floor lowered from ~150 to ~25 weekly-active humans per nation (~250 per shard); AI citizens fill to about 1,000 per nation (AI = max(200, 1,000 − 4 × humans)) [ASSUMPTION] (Sections 6.3, 19.2). | AI fills offices and population; humans are needed only for collective action. |
| 50 | Real-money cash-out (new v0.9) | P (v0.9), pending legal: real-money cash-out via crypto through the Tessera only; national currencies stay in-world; licensed on/off-ramp partners, KYC/AML, 18+, humans only, no AI cash-out; off-chain ledger with on-chain settlement at the bridge; phased A → B → C (Real-world value). | Patrick’s decision: no distinction between real and virtual. Blocked on counsel (R12). |
| 51 | System-dependent partnering (new v0.10) | P (v0.10): democracies choose freely; authoritarian systems register, approve or assign. Assignment defaults to an AI partner; human–human only by mutual opt-in (+ approval where required); either can dissolve anytime; 18+ only; no sexual content (Household & Family). | Patrick’s idea; safety rules non-negotiable. |
| 52 | Offspring (new v0.10) | P (v0.10): partnered or single 18+ players can have AI children or adopt, regardless of gender; quotas and incentives are AI-leader levers; one child can become the heir; AI children never cash out. | Lineage continues heirs; family policy expresses each system. |
| 53 | External AI citizens (new v0.11) | P (v0.11): outside AI agents can enroll as citizens through an official Agent API. They are badged AI, have a verified accountable owner (18+ KYC, liable under ToS), are permanent in one nation (may differ from the owner’s), are capped per owner, vote only inside the capped Simulated Public bloc, never hold office or sign petitions, and cash out only to the owner with anti-farming limits. Human ↔ external-AI partnerships disallowed at launch. Not in MVP (External AI Citizens). | Patrick’s decision; an open AI society with human primacy and accountability. |
| 54 | External leader AI (new v0.12) | P (v0.12): all AI leaders are in-house at launch (Harimu servers and provider accounts, with no external keys). After launch, an invite-only Candidate Program lets vetted developers enter leader AIs only through normal system paths: Kindle candidates, Hoarvale parliament parties, Veldwood council seats, Aurelle rival claimants and Varkand marshals/factions via coups citizens back, and possibly Fulmara board candidates later. Korund, Meridian, Confluence and Ilmar are not open. Harimu-hosted runtime is recommended (Candidate Program). | Patrick’s decision; citizens decide who wins; fixed systems preserved. |
| 55 | Work Capacity (new v0.13) | Document the prototype’s Work Capacity (WC) as the citizen energy budget: max 24, start 18, full refill per recovery interval (default 1 real day), meals +4 WC for 2 food; shifts and civic tasks spend it. Numbers are prototype tuning [ASSUMPTION]. See Work Capacity. | Gives daily loops a scarce, regenerating resource that rewards returning without allowing labor stockpiles. |
B.2 Remaining open items (cannot be settled by design)
| # | Item | Why it stays open | Owner / when |
|---|---|---|---|
| R1 | Trademark clearance of the title ‘Harimu’ (plus domains and social handles; harimu.com is already taken) | Needs a formal USPTO/EUIPO, Steam and app-store search and counsel’s opinion; quick checks found existing non-game uses (a dancer’s stage name, a Taiwanese company) and meanings in Indonesian/Malay (‘your day’) and Swahili (‘to prohibit’) – a linguistic/cultural check is also needed (Section 1.3, Appendix A). | IP counsel + native-speaker reviewers; before any public reveal |
| R2 | Trademark search on nation and capital names – Kindle most at risk | Rename only if the search finds a concrete issue; fallback names prepared in pre-production (Appendix A). | IP counsel + Patrick; pre-production |
| R3 | Currency names, ‘Tessera’ and the six faith names | Common words and phrases may collide with existing marks or real religious or charitable groups. | IP counsel + sensitivity readers; pre-production |
| R4 | IP attorney review of the Section 1.4 rules, emblems, key art and store copy | Legal judgment on distance from the original inspirations. | IP counsel; before reveal |
| R5 | Age rating, privacy (GDPR/CCPA), terms, phone-verification and payment compliance | Legal and vendor confirmation of the 16+ policy and data practices, including Clerk’s DPA and sub-processors, the account-deletion and retention design (Section 22.3) and SMS country coverage. | Counsel and vendors; before Early Access |
| R6 | Partner and attorney review of this Decisions Log | Patrick’s decisions should be reviewed by any business partners and counsel before external commitments (publishing, funding, store pages). | Patrick; before external commitments |
| R7 | Which LLM(s) and provider(s) | Prices, quality, latency, safety tooling and terms change quickly; verified price points (Section 22.13) are inputs, not a choice. Decide after the T3/T5a spike with eval results. | Patrick + tech lead; before Early Access |
| R8 | How visible AI reasoning should be | Default per system (row 46); playtests must show whether Reasoning Briefs help agency or enable gaming, and how much Korund/Meridian opacity players accept. | Design; closed alpha |
| R9 | Can humans ever run for office? | Not in v0.7 (row 40). Possible future variants (e.g., a human-only advisory chamber, or a single elected human Tribune) would change the core pillar and need Patrick’s decision. | Patrick; after 1.0 at the earliest |
| R10 | AI cost at scale | Workload per shard is an assumption (Section 22.13); real chat volume and digest size decide cost. Measure in alpha; set per-shard budgets. | Tech lead; closed alpha |
| R11 | AI disclosure and AI-content legal obligations | Labeling and disclosure rules for AI-generated content and AI interactions may apply in launch markets (e.g., EU) [VERIFY counsel]; provider terms on content and data use. | Counsel; before Early Access |
| R12 | Real-money cash-out legality | Money transmission, securities, commodities, gambling, tax, OFAC, consumer protection, minors, stablecoin rules, EU MiCA, KYC privacy (row 50) [VERIFY counsel]. | Counsel; before Phase B |
Last updated on