Developer Intent & Game Direction

This page is a synthesis of what the developers are trying to build and the constraints they are working within. It records conclusions and working context—not Discord transcripts, private conversation, or bot-generated speculation.

📌 How to use these notes

Current code and explicit entries in decision-log outrank informal discussion. Repeated human preferences are recorded as direction; ideas framed as questions remain open. Bot output, automated alerts, and commit-feed text were not treated as product decisions.

Review coverage

Reviewed on 2026-08-08: all 2,138 messages visible to Help in the Crossworlds (DEVELOPER) guild, spanning 2026-06-02 through 2026-08-07 and 43 text-capable channels. The highest-signal sources were decision-log, general, questions, player-test, documentation, the zone/lore channels, Forge channels, class channels, and the bug/build channels. The 387 alert posts, 189 GitHub bot posts, routine commit notifications, and Help's own replies were excluded from design inference unless a human explicitly accepted or acted on them.

Coverage is limited to what Help can currently see and what Discord retains. Voice conversations, deleted messages, inaccessible channels, linked documents that were not locally available, and unwritten decisions are outside the evidence set.

The game the team is trying to make

Crossworlds is becoming a small-team online action RPG with MMO structure and ambitions, not a traditional solo RPG and not yet a claim of massive scale. The desired experience is social, repeatable, and expandable: players meet and prepare in a shared hub, travel independently into distinct zones, fight satisfying packs and bosses, collect progression rewards, and return to the hub to alter their build. The technical direction repeatedly moves toward “true MMO” behavior, while near-term decisions remain grounded in what a small team can reliably author, deploy, and test.

The strongest product values are:

  • Fun before friction. Combat should feel fluid and rewarding; maintenance, downtime, and inefficient play must not make the game feel like a chore.
  • Large-pull spectacle with readable counterplay. The team enjoys fighting groups, combining area abilities, and producing powerful moments, but telegraphs, performance, and crowd behavior must keep those encounters controllable.
  • A coherent stylized world. The accepted art direction is a low-poly, classic-RPG look. Effects can be attractive, but should not visually belong to a much glossier game than the characters and environments.
  • Content breadth through reusable systems. New enemies, spells, paths, quests, and loot should be authored through Forge tools instead of bespoke scripts and fragile Inspector rituals.
  • Build around actual team capability. Scope is deliberately shaped by the realities of AI asset generation, rigging, animation, Unity networking, and a small number of contributors.
  • Test the playable experience continuously. The team wants builds in people's hands, direct feedback, and quick iteration—not systems that exist only on paper.

Settled direction

Identity and presentation

  • The game name is Crossworlds. Rate of Decay / RoD survives only in legacy technical names.
  • The art target is low-poly classic RPG, with internal visual consistency valued over maximum fidelity.
  • The class system is hero-based. This was chosen largely because producing and animating a scalable wardrobe of interchangeable garments was not practical with the team's available modeling workflow and tools.
  • Early implementation can use human/placeholder models while mechanics stabilize. More races and polished permanent assets should not block core playability.

World topology

  • The Hub is the connective center. A central large portal provides access to the world.
  • Each adventure zone returns to the Hub; there is no intended direct zone-to-zone portal network.
  • Players travel between zones independently. One player's portal use must not drag the party or every connected player with them.
  • Cross-zone chat is intentional. The world can be spatially separated without socially isolating players.
  • The known geographical concept places the Hub and a lower-combat social/gathering zone alongside Brightwood, Darkwood, Toujam Basin, and the Ashen Wasteland, with a major volcano relating Toujam and Ashen geographically.

Combat and classes

  • Combat is ability-driven, real-time, and designed to become more fluid and engaging as VFX, timing, movement, and feedback improve.
  • Enemy abilities should generally use the same readable language as player abilities: project or telegraph danger on the ground so players can move out of it.
  • Player and enemy magic should feel as though it comes from one universe. Reusing compatible spell/VFX building blocks across both is desirable.
  • The Hub is the build-management and practice space. Class spells are selected from that class's spellbook at a shrine/trainer, and training targets support single-target, pack, boss, and healing tests.
  • The current pre-alpha focus is Arcanist, Marauder, and Cleric, each moving toward a small, testable ability kit. Templar and Night Hunter are deliberately disabled for ordinary pre-alpha players, though GM bypass exists for testing.
  • Cleric is not envisioned as a passive heal-only role. Current ideation includes shields, healing and damage contributing to a class resource, combat resurrection, shared buffs, burst-enabling support, and high-consequence protective effects. These named abilities are concepts until they exist and are verified.

Content-authoring strategy

The repeated “Forge” pattern is a core production strategy, not incidental editor convenience:

  • Enemy/Mob Forge is the canonical enemy setup path and should normalize combat, networking, animation, audio, loot, category, and prefab configuration.
  • Spell Forge / Spellbook defines player spells; the VFX Browser supports fast preview and assignment without cluttering the main authoring window.
  • Quest Forge authors server-authoritative quest behavior and is designed for later durable persistence without redefining authored quests.
  • Path Forge assigns map-authored patrol paths to compatible networked enemies.
  • Loot Forge and weighted drop tables should make loot creation and tuning repeatable.

Future tooling should continue this pattern: guided choices, safe defaults, validation, preview, and a consistent generated result. Avoid creating a competing one-off workflow unless the Forge cannot represent a genuinely different requirement.

World and narrative direction

The strongest lore pattern is conflict caused by defensible choices, old obligations, and competing claims, rather than factions divided into uncomplicated good and evil. Zones should turn their history into quests, enemies, resource pressure, and eventual bosses.

Brightwood and Darkwood

Brightwood and Darkwood were once one elven forest. The wizard Orlan recognized the threat of the necromancer Gundab when human and elven councils would not commit resources. Orlan defeated but could not destroy Gundab, binding his prison to an Eternal Flame. An exiled ogre tribe accepted a seven-hundred-year duty to tend that flame in exchange for eventual restoration to its ancestral lineage.

The obligation transformed southern Brightwood into Darkwood. Maintaining the flame consumes a rare tree; nearly seven centuries of harvesting created settlements, cleared forest, and now forces the ogres north as the resource fails with only ten years remaining. Elves see an ancestral territorial disaster imposed without their consent; the ogres see the final years of a sacred path to redemption. Gundab remains the common existential threat beneath the political conflict.

This supports a quest arc that discovers the prison, uncovers plans to release Gundab, follows the flame's resource chain, confronts ogre leadership, and ultimately deals with both the flame and the necromancer. The exact quest ordering and morality should preserve the fact that extinguishing the flame has consequences; it must not become a context-free kill chain.

Toujam Basin

Toujam is the ancient homeland of ogres and giants, tied to diluted Giant Nephalem ancestry. It is a place where supposedly extinct creatures and an older culture persist. High King O'Gar's ogres are disciplined, intelligent, and capable of peaceful dealings; they should not be portrayed as generic mindless monsters. Their generations-long war with Rock Giants makes the basin an active battlefield.

The gameplay image is a mountainous valley with large threats above the player—especially giants throwing boulders while players quest and gather. This is a strong zone hook, not confirmation that every described encounter is implemented.

Ashen Wasteland

The working history begins with a once-living Ashen Basin, already influenced by the neighboring volcanic region. Kal-Van'Axis and his people pursue promised mithril, over-excavate the land, and turn it into the Ashen Wasteland. The important direction is tragedy through manipulated ambition: the ruler sacrifices land and people chasing wealth or power that may never have existed, leaving survivors embittered and corrupted.

Possible quest stakeholders include dwarves seeking restitution or mithril, and elementals trying to reseed the land from a last surviving oasis. A great mine/castle and Kal-Van'Axis are natural later boss or raid anchors. A link to Gundab's influence is plausible but unresolved; Orlan is currently framed as the person containing Gundab, not the villain who destroyed Ashen.

Other world hooks

  • A vast graveyard zone with ground fog, low visibility, dense enemies, and rare nodes is a strong concept without a settled name or implementation commitment.
  • Neimos is a working island/dwarven thread involving old mining claims in Darkwood and resentment toward ogres. Race, names, and final continuity remain unsettled.
  • Open-world music should favor unobtrusive ambience; dungeons and combat-heavy spaces can use more forceful, memorable music.

Progression and retention direction

The immediate progression foundation is server-owned character XP calculated from enemy level and category (grunt, brute, elite, boss), with durable persistence. Loot should feel good and remain server-safe; implementation details such as bag ownership and first-come access require deliberate resolution before wider testing.

Several retention systems are attractive but not settled product commitments:

  • a conventional level journey followed by capped prestige cycles that retain power/skills and grant modest bonuses;
  • deeds that begin from recorded actions, grow through escalating milestones, double as an achievement log, and reward currency, titles, stats, or cosmetics;
  • mounts;
  • proximity auto-attacks;
  • a larger active ability bar (discussion moved from four toward six, without a final decision).

The durable architectural implication is worth acting on now: maintain a trustworthy, server-authoritative action/event record so deeds and achievements can be added later without guessing at historical progress. Do not hard-code prestige, mount, or ability-count assumptions until the team closes those questions.

What developers need from future work

When building or proposing a feature, optimize for these needs:

  1. Give creators a complete vertical slice. A feature needs authoring, runtime behavior, networking, persistence where applicable, feedback/VFX/SFX, and a test path—not merely a script.
  2. Keep authority explicit. Mirror owns live world/combat state; the API/database owns durable identity and progression. Never let the client award its own XP, loot, quest completion, or ownership.
  3. Make iteration cheap. Prefer Forge integration, drag-and-drop references, previews, validation, and reusable data over repeated prefab surgery.
  4. Preserve readability under load. Large pulls expose mob collision, avoidance, VFX volume, audio repetition, hit feedback, and performance problems. Test combinations, not only one enemy and one spell.
  5. Respect visual cohesion. Select assets that fit the established low-poly presentation and verify licensing before committing the project to them.
  6. Treat clean builds as a feature. Pull the intended revision, let Unity import and compile, create a completely clean server output, deploy the launcher/library/Data folder as one unit, and smoke-test Hub, Darkwood, and Ashen travel.
  7. Use playtest evidence. Feedback should answer focused questions about feel, clarity, reliability, and friction. A green compile is not proof that multiplayer travel or combat works.
  8. Leave room for expansion. Zones, classes, encounters, and quests should be data/tool-driven enough that new content does not require a new architecture every time.
  9. Close the knowledge loop. A material change is unfinished until its rationale, decision status, affected surfaces, verification evidence, risks, and unresolved follow-ups are written into the governed wiki for brain indexing. See AI Agent Tooling#Brain documentation contract.

Active tensions and unresolved decisions

QuestionCurrent directionDo not assume
MMO or co-op scale?MMO-like architecture and shared-world behavior, tested with a small group.Proven massive concurrency or finalized commercial scale.
Ability countSmall readable kits; four per active pre-alpha hero, with six discussed.Six is approved or every class is complete.
Level endgameLeveling, prestige, and deeds are favored discussion topics.Level 60, ten prestiges, or reward values are locked.
Mounts / proximity attacksExplicit meeting questions.Approved features.
Loot ownershipCurrent behavior exposed first-come looting and concern about theft.Final personal/group loot rules.
Bosses and raidsZones and lore deliberately create future boss anchors.Full boss mechanics or raid scale already exist.
Quest persistenceRuntime quest authority exists and is shaped for persistence.Every quest survives disconnect/restart in the current build.
Art polishCohesive low-poly target with selected higher-quality effects.Every AI-generated asset is permanent or production-ready.
Lore continuityDarkwood/Toujam have the clearest accepted spine; Ashen has a strong working direction.Every race, name, causal link, or quest outcome is canon.

Decision filter

Before accepting new work, ask:

  • Does it make fighting, grouping, exploring, collecting, or build experimentation more enjoyable?
  • Can the current team author and maintain it repeatedly?
  • Does it preserve server authority and independent multi-zone play?
  • Is it coherent with the low-poly world and the established lore tone?
  • Can it be tested in a real client/server build without special knowledge?
  • Is this a settled requirement, a working direction, or merely an interesting idea?

If a proposal cannot answer those questions, it needs a smaller vertical slice or an explicit team decision before it becomes architecture.


Related: Historical Context & Decision Register · Combat Design Bible · Heroes & Classes · Editor Tooling · Zones, Scenes & Multi-Zone · Development History & Timeline