Realm development phases and build plan

• Read and reconciled all design documents (docs 01–12)
• Wrote five worked business examples using only the 9 primitives
• Chose the v1 simulation stack (Python engine, Next.js frontend)
• Set up the project repository with a clean directory structure
• Wrote the one-pager pitch and the canonical glossary (doc 15)

Can you describe, in one paragraph each, exactly what 5 different player-invented businesses look like — including which primitives they use, what their daily activities are, and how they make money? If yes → Phase 1. If no → the primitives spec is incomplete. Revise the primitives and try again.

• Tick-based deterministic simulation loop (advance_tick; RNG via (tick, purpose) pairs — no wall-clock in sim)
• World generation: Frontier bootstrap grid; terrain via biome noise; plot terrain, hidden subsurface composition, ownership
• Materials (Primitive 2): 12 definitions including spoiled_grain; party-wide storage unit cap with field_stockade bonus; grain → spoiled_grain spoilage transform (conserved, Law 1)
• Capital (Primitive 5): accounts, atomic double-entry transfers, conservation enforced at the ledger layer
• Production (Primitive 6): 5 recipe templates (e.g., timber → lumber, ore → metal); 40% of recipe labor_cents paid to distinct stub_hires employees per batch; tool_cache / watch_hut lower labor cost on the producing plot (BPS)
• Movement (Primitive 4): shipping with tick-based transit; fee = BASE + manhattan × PER_TILE; deliveries respect storage cap
• Order books (Primitive 7b): asks + bids, escrow, iceberg clips, counterparty reputation minimum, FIFO by price then order_id
• P2P trade (Primitive 7a): atomic exchange with optional idempotency key and stable structured error codes
• Basic contracts (Primitive 8): supply FSM with buyer deposit and liquidated damages; employment with recurring wages
• Reputation (Primitive 9 placeholder): honor/breach counters; order book filters on min_counterparty_honored

• 6 Tier 1 behavioral archetypes: grain consumer, lumber buyer, timber relister, coal buyer, clay buyer, electricity buyer

• Next.js shell with Bloomberg-style tab panels
• SVG organic mesh world map (no Pixi.js yet — that is Phase 2)
• Plot detail, market order book + depth chart, inventory, build menu, hire menu, action log, caravans, contracts, Bazaar (P2P) tabs

• SQLite save/load including order books, P2P idempotency, and market history

Get 3–5 strangers to play the prototype for 1 hour. After their session, ask them: “would you keep playing if I gave you another hour?” 3 of 5 must say yes. Watch for: when they’re confused (first-hour script needs work), when the world feels empty (increase agent activity), when cause and effect are invisible (add visible price reasoning). If they call it a spreadsheet rather than a game — that’s acceptable in Phase 1. You’re testing whether the economics are fun, not the visuals.

• Pixi.js v8 world map: colored terrain, plot boundaries, ownership shading, GL/SVG toolbar toggle
• Schematic plot view (PlotSchematicPanel.tsx): drag-drop production flow editor; validates against engine recipes and party inventory via POST /plots/{id}/schematic/validate
• Real charting (Recharts): MarketHistoryChart.tsx with axis labels, Bazaar symbol sync, ask + bid series
• Polished UI: Cmd/Ctrl+K command palette, tab strip, notification toaster (realmToast.tsx)
• Settings modal: pause/speed control, scenario selector, save management

• Tier 2 optimizing agents: 5 archetypes in agents_tier2.py (market-maker, logistics, production-planner, employer, and a fifth); distinct from Tier 1 schedules
• 27 materials (up from 12), 20 recipe templates (up from 5)
• Contract stubs: loan, equity, service-subscription FSM with tick-based lifecycle and API routes
• Surveying: survey_plot reveals terrain and recipe_ids; cost flows to system reserve; supports an optional tradable survey intel market
• Decay (Law 5): tick_building_decay; maintain_building fee = max(1_000, cost // 5); labor bonus gated by building condition
• Information cost (Law 6): purchase_market_intel; free tier shows last FREE_MARKET_HISTORY_TICKS ticks; paid tier extends visibility; fee → system reserve

• 4 scenarios: Frontier (default), The Bootstrapper ($5,000 start, 32×24 grid), **The Speculator** ($20,000 start, 40×30 grid), The Cartel (grain-side pressure overlay)
• Scenario selection UI in settings modal

• Snapshot version bumped to v2; additive fields for scenario ID, market intel expiry, building instance ID and condition BPS

Get 5 strangers to play for 5+ hours each (across multiple sessions, ideally a week of self-directed play). At the end: at least 3 of 5 say “I’d buy this for $30 today.” Bonus signals: testers post screenshots organically, tell friends, ask when multiplayer is coming. Those are stronger than the survey answer alone.

• LLM agent runtime (Anthropic API integration)
• Memory system: rolling window + long-term summary
• 5 named characters with distinct personalities (per the AI agent design doc): examples include Margaux the Industrialist, Rico the Speculator
• Cost monitoring and per-session budget caps

• Stable build pipeline, crash reporting, opt-in telemetry
• Save file forward-compatibility (saves from v1.0 work in v1.1+)
• Auto-update mechanism

• Steam page and/or itch.io and/or self-hosted
• Trailer and launch video, marketing copy, screenshots
• Demo version (limited scenario, no save)

• One-time purchase, recommended $25–$50
• Payment processing and email/newsletter collection

• Documentation site (this site)
• Discord community
• Bug report workflow

Launch publicly. Within 30 days of launch: at least 200 paying users with >70% day-7 retention, and a Steam (or equivalent) rating averaging 80%+. Soft-launch to beta testers first. Fix critical issues. Then public launch. If retention is below targets, the design has a flaw at scale — address it before moving to Phase 4. If reviews flag the same issue repeatedly, that issue is the priority.

• Lua sandbox: sandboxed interpreter with restricted stdlib, deterministic, resource-metered
• Read API and Write API (per the user-code layer spec)
• Service definition, publication, and versioning
• Resource metering: CPU and storage budgets billed in in-game currency
• Service marketplace UI

• Visual drag-drop block editor for non-programmers
• Compilation from blocks to Lua
• Beginner templates: “auto-restock,” “price alerts,” and similar

• Pre-shipped services that AI agents publish (so the marketplace isn’t empty at launch)
• Services that AI agents subscribe to from the player
• Services that the player can create and sell to AI agents

• Code editor (Monaco or CodeMirror)
• Live preview and test environment
• Debugging tools: logs, breakpoints if feasible

At least 30% of returning players (week-2 cohort) deploy at least one service or use the block editor non-trivially. At least 50% of testers can describe a service they wish existed (showing they understand the platform’s value). Track via telemetry: who opens the IDE, who writes code, who deploys, who subscribes. If fewer than 30% engage, the UX is too hard — add more templates, simplify the block editor, and improve tutorials.

• Authoritative server architecture
• Postgres-based state (migrated from SQLite for multi-user concurrency)
• Real-time WebSocket layer
• Auth: email + password, OAuth optional
• Identity and account management
• Anti-cheat (mostly automatic via the determinism and audit-log architecture)

• Genesis bonuses: starting-capital boosts for early players who fill needed roles (shipping, finance, agriculture), phasing out as roles become saturated
• NPC seed economy: consumers, laborers, baseline banks and utilities
• Frontier zone reveal mechanic

• Invite-only registration
• Season management: start, run, end, ranking, and archive
• Live leaderboard
• Spectator mode for non-participants

• Monitoring and alerting, backup and disaster recovery
• Customer support workflow and moderation tools

Closed Cohort 1 runs for its full 90-day season with: less than 5% participant dropout, less than 2 hours of unplanned downtime total, no exploits affecting more than 5% of player state, and at least 60% of participants saying they’d join Cohort 2. Run a 7-day pre-season stress test with 20 participants before the full season. If the cohort falls apart — mass exit, major exploits, infrastructure failure — do not open public mode. Run Cohort 2 first.

• Cohort 2 (with lessons from Cohort 1)
• Cohort 3 (larger, ~150 players)
• Public open beta (1 shard, free or low-cost entry)
• Tournament and in-season event tooling
• Streamer and content-creator tools: replay export, embed widgets
• Mobile companion app v1: read-only access plus the key quick-action flows from the mobile spec

Public open beta runs for 90 days with: 1,000+ daily active users at peak, 40%+ day-30 retention, no critical exploits, and a steady-state economy (no hyperinflation, healthy market depth across major asset classes). Monitor KPIs daily. Address economic dysfunction immediately — hyperinflation or market collapse is a design problem, not just a numbers problem. If the open beta shows systemic problems, fix and re-run before opening permanent multiplayer.

• Multi-shard architecture with shard selection UI
• Cross-shard player profile: single account, multiple worlds
• Subscription or updated monetization model for multiplayer
• Mobile companion v2: full functionality per the mobile spec
• Player-issued currencies (Primitive 5 v2)
• Advanced financial primitives: derivatives, insurance, bond markets
• Player governance and corporations: multi-player businesses with shareholders