Strategos Universalis

**MULTIPLAYER: 3–4 humans** (up from 1v1 in Alpha 0.9). Multi-peer host-as-relay (v6M4) — host acts as central relay; joiners connect once to host + host fans out state; deterministic strict-lockstep preserved across N peers with expectedPeerPacketsPerTick gating. N-peer save + rejoin with per-joiner hash verification + any-subset AI-takeover on missing rejoins (v6M4.5). Per-joiner 60s disconnect grace → deterministic mpAiTakeover command via CommandBridge (v6M4.6). End-of-game statistics screen (v6M5) with per-empire economy time-series graphs (credits / minerals / fuel / research / population / influence sampled every 60s) + 12-column counter table (ships + structures created / destroyed / lost; damage dealt / taken; planets captured / destroyed). Repair order for ships (v6M6) — routes to nearest friendly-or-allied Repair Yard in current system; formation members route via phantom to a single fleet-wide yard; auto-clears at full hull; alert if none in-system. Allied heal fix — Repair Yards now heal same-team ally ships (bug pre-v6M6). Player profile totals (v6M7) — Total games + Victories persisted per Clerk account via new Neon columns; Stats section + Reset button in the account-settings modal; guests skip persistence. Habitability-driven per-pop income (v6M8) — new formula `pop × base × planetTypeMod[biome] × habitabilityScoreMod(score)` with three orthogonal xlsx-authored lookup tables in Habitability.xlsx; score modifier is piecewise-linear interpolated between authored breakpoints. Range rings (v6M9 → MFix102) — three configurable ring types (solid green vision / dotted green scan / solid red weapon-max) around selected ships; multi-selection renders three aggregated union envelopes rather than per-ship rings; Alt-hold reveals per-ship rings for all ships in system as a peek mode; four settings toggles (player vision / scan / weapon + master allied toggle). Damage counter aggregation (v6M11) — floating shield / hull / miss counters batch on a rolling 500ms per (target, attacker) window instead of firing per-tick; single settings toggle for on/off. **Weapon-name simplification (v6M10)** — six weapon families with a Mk 4 rename retired to a single family name across all 6 Mks: Plasma Cannon / Fusion Beam / Gauss Autocannon / Megabolt Railgun / Concussion Missile (new name replacing both Pulson + Zeon) / Hellfire Torpedo. Numerous stability + polish fixes (MFix92–102 — worker.ts return-await for CORS-clean error paths; team vision sharing at planet + ship + movement-line level; settings modal offscreen fix at short heights; planet-terrain overlap rejection; formation follow-ups; drizzle-history hygiene). Save-format hard break from Alpha 0.9 — SAVE_FORMAT_VERSION 45 → 46, SAVE_BLOB_VERSION 43 → 44, no migration; legacy Alpha 0.9 saves reject at load with a version-mismatch message.

**MULTIPLAYER — two humans can play a full match together.** 1v1 PvP or co-op vs AI, connect via 6-digit match code, WebRTC peer-to-peer with deterministic strict-lockstep simulation. Full session-state persistence: host saves broadcast to joiner over the data channel so both peers hold byte-equal saves; load-then-rejoin flow creates a fresh match code + hash-verifies both peers' saves match before resuming from the exact saved tick. Mid-match peer drop → 30-second grace banner then abandoned-match end screen; MP-deterministic concede + victory/defeat outcomes derived per-peer. Production authentication (v6M1) — Clerk production instance signed against strategosuniversalis.com custom domain with passwordless email + Google + Microsoft OAuth; Neon production Postgres branch persisting accounts + profiles; GDPR-compliant data export + account deletion; first-time terms-acceptance modal. Signaling (v6M2) — Cloudflare Worker (sg-mp-backend) hosts 6-digit-code match creation + HTTP-polled ICE/SDP signal exchange; WebRTC data channel opens ordered + reliable between two browsers. **In-browser authoritative simulation (v6M3.1)** — CommandBridge routes both peers' inputs through a strict-lockstep 10 Hz loop; both peers broadcast per tick + apply only when both packets present; deterministic sort keeps simulations byte-identical. Full setup UI (v6M3.2) matches the SP layout with per-empire race + colour + team pickers, host-controlled galaxy params, and configurable AI opponents. Multi-human-empire runtime (v6M3.3) — Empire.controllerUserId tags each empire with its Clerk user; AI dispatch skips human-controlled empires per peer. Save + rejoin (v6M3.4) via hash-verified handshake. Disconnect + match-end (v6M3.5). Individual pause via CommandBridge (v6M3.5.d) — either peer can pause; sim halts on both under strict lockstep. Known limitations: 1v1-only (2v2 + 3–4-human FFA planned for v7); no lobby browser / matchmaking / chat (v7); background-tab rAF throttling can freeze lockstep — workaround is split-screen visible windows; **WebRTC connect on two-machine LANs can be intermittent under Google-STUN-only signalling — reliable smoke fallback is two browser windows on the same computer (localhost bypasses external NAT + mDNS entirely). Reliable production MP across arbitrary networks requires a TURN relay, planned for v7.** Underlying single-player game unchanged from Alpha 0.8 (SP saves from Alpha 0.8 load unchanged; no format changes for SP).

AI THAT ACTUALLY WINS — Brutal AI now beats the developer in playtest for the first time. The full v5 design surface ships in this release. Influence Diplomacy (v5M5 — 5-tier ladder Contact → Friend → Partner → Ally → Assimilating; per-target allocation rate; Embassy income; tier-driven trade/vision/migration bonuses; per-tier diplomatic overlay). Ship Orders Pass 1 (v5M7 — Patrol with waypoints single + group; Guard target planet-or-ship; Explore mode; Standing Orders defaults per archetype; Hold Fire engagement mode). Per-empire AI archetypes (v5M9 — 7-archetype roster Aggressor / Defender / Expander / Balanced / Random / Unpredictable / Race-Aligned; per-race overlay system; objective catalogue with 12 templates; Brutal-difficulty advanced strategies). Formations (v5M11 — 4 types Standard Battle / Sniper / Stealth / Shield Wall; per-fleet multi-select-driven; anchor-tracked positioning). AI Military Conquest (v5M12 — MV-weighted attack-fleet + raiding-fleet objectives; ship commitment + dispatch + engagement passes; retreat behaviour; difficulty-scaled dispatch). Pre-release polish (v5M13 — defeat alert + always-visible remaining-enemies HUD chip; all distances + speeds rendered in km / ly; galaxy-view ownership symbol redesign; build<N>:<kind> cell syntax for AI build orders). AI cell-expression authoring (MFix76.3 + MFix79 + v5M13.4 — per-planet build orders via xlsx with biome / habitability conditions; upgrade<N>:<kind> + build<N>:<kind> forms; data:build-time validator). Planet structure decommissioning (MFix70 — 0% refund, manual flip). Per-chassis ship quota (MFix78.5 — Scout:3 rolling-maintain syntax). MFix80 data-driven AI ship-bucket classifier — 15 RACE-BB/CC combat chassis previously silently excluded from attack-fleet commitment, now included. MFix84.1B radiation belt hazard avoidance for AI ships. MFix84.1C AI in-system scouting sweep + opportunistic planet bombardment after winning fleet engagement. **MFix85.1 multi-hop combat dispatch — the load-bearing AI aggression fix; replaces the v5M12.3 direct-jump-only adjacency check with BFS multi-hop pathing.** Numerous stability fixes across MFix42–85 (test infrastructure, lint baseline, formation drift, alerts, terrain features, weapon audio, planet ownership model, stepper input cache, xlsx narrative-prefix converter, defect resolution). Save-format hard break from Alpha 0.7 — SAVE_FORMAT_VERSION 34 → 45, no migration.

Advanced AI behaviour models (v4M8 — 10 sub-checkpoints: compositional archetype × difficulty × race-overlay framework with new Brutal difficulty tier; threat-weighted target eval; multi-hop A* pathing; fleet-coordinated dispatch; cloaking rules; home-defence recall; Warp Gate auto-connect; team-aware coordination; standing orders; Brutal coordinated assault). Terrain feature visual + effects pass (v4M13 — black-hole / ion-cloud / radiation / asteroid / nebula gameplay effects; commissioned bitmap sprites + per-type procedural overlays + 3-tier parallax). Galactic map enhancements (v4M6 — three star classes with yield modifiers; four strategic overlays; warp destination picker; lane removal; strict FoW for unexplored systems). Three-layer fog of war + LRSA cross-system reveal (v4M10) + in-system entity-level fog (v4M11) + Jump Prediction Sensors arrival-angle visibility (v4M12). Planet Structures UI Enhancements (v4M7 — build queue subsection with active highlight + cancel + confirmation modal; structure detail readouts; multi-line tooltip effect ladders). Research progress retention + queue (MFix42.8 — switching tech parks progress instead of resetting; click-to-queue future targets). Full keybind rebinding (MFix42.7 — 12-action registry in Settings). Polish: bombardment hit VFX + planet damage indicator, ODP ring-node sprite, Scout in-system patrol, beam render redo, burst/pulse multi-emitter offsets, missile salvo accounting, weapon VFX cohort, dev-spawn cross-race aliasing, Missile Weapons Platform removal, per-kind minSpacing on space structures. Save-format hard break from Alpha 0.6 — SAVE_FORMAT_VERSION 34, no migration.

Hyperspace travel rework (per-ship range cap + Hyperdrive Range Mk 1–6 research; distance-scaled fuel cost; continuous Jump Area replaces fixed jump points; Warp Gates instant + free; per-class charge-up tier). Teams (vision sharing, blocked friendly-fire, joint win conditions). Galaxy shapes (Even / Spiral / Bottleneck / Core and Rim). Polish pass — commissioned backdrops, autosave, .susave export/import, four new Settings tabs, music beds, in-game tooltips.

First public playtest. UI, Economy, Races, Diplomacy, Terrain Features, Tech Tree, Structures, Fleets and Ship Orders.