Game Concept: Driftlands

Created: 2026-08-25 Status: Draft

Creative Director Review (CD-PILLARS): CONCERNS — all three sharpening edits applied and locked 2026-08-25
Art Director Review (AD-CONCEPT-VISUAL): CONCEPTS — "Tideline" direction selected 2026-08-25
Technical Director Review (TD-FEASIBILITY): FEASIBLE WITH CONSTRAINTS 2026-08-25
Producer Review (PR-SCOPE): REALISTIC WITH ADJUSTMENTS — all three adjustments adopted 2026-08-25

Elevator Pitch

It's a top-down co-op crafting-survival game where your party's island home is
permanent — but the lands around it are not. Drifting landmasses dock against
your shores on a real-time cycle, each bringing resources, ruins, and threats;
when they drift away, whatever you didn't harvest goes with them.

Core Identity

Core Fantasy

**"We don't travel to the world. The world travels to us — and some weeks, what washes up wants us dead."**

The private-server fantasy: our world, our base, our characters. Each friend drifts toward what they enjoy — no assigned roles — while the shared home grows from camp to keep. Every item anyone wears was made by one of the group, from materials someone hauled home before the land that bore them drifted away. The world breathes in real time whether anyone is playing or not; checking "what's docking tomorrow?" between sessions is part of the game.

The group's core motivation (stated by the developer): **progression on something that is your own and lasts** — building a thing that persists and gets better throughout the game. The permanent home island is that thing; the tide exists to feed it. Any design that makes the lasting home feel disposable or resettable fights the core motivation.

Unique Hook

Like Valheim, AND ALSO the map is a tide. Instead of the party traveling to distant zones, the zones come to the party: landmasses dock against the home island's shores on a real-time drift cycle and depart again, taking unharvested riches with them. This inverts the MMO structure, makes the base emotionally central by construction, generates "one more cycle" anticipation between sessions, and provides built-in base-defense pressure (hostile lands dock too) that never needs contriving. Missed lands recur eventually — possibly changed — so absence creates stories and appetite, never permanent loss.

Player Experience Analysis (MDA Framework)

Target Aesthetics (What the player FEELS)

Key Dynamics (Emergent player behaviors)

• Players check the tide state between sessions and plan in group chat ("obsidian coast docks Thursday")
• Division of labor emerges without enforcement — one friend gravitates to smithing, another to deep expeditions
• "One more trip inland before bed" — departure timers create self-imposed session goals
• Missed lands become group lore ("the tin highlands came while we all slept")
• Prosperity invites danger: as the base grows, hostile dockings raise collective stakes

Core Mechanics (Systems we build)

• The Tide — deterministic real-time drift schedule; lands dock at fixed shore sockets and depart, recurring eventually
• Fast top-down action combat — dash, strike, kite; friction in the world, never in the hands
• Provenance-free crafting chain — gather → haul → smelt → craft; cloth → leather → bronze → iron tiers; no loot drops, ever. Tier rollout (canonical): MVP ships cloth only; bronze arrives at the start of Tier 1; leather and iron complete the set within Tier 1.
• Shared base building — walls, workshops, storage; the fixed point everything orbits
• Flat-stats progression through gear — no classes, no skill grind; character power and identity come entirely from crafted equipment (decided 2026-08-25 at the Systems Design gate: skill-drift is cut — flat stats preserve Pillar 5's floor, remove the power-divergence social risk between friends with uneven playtime, and let Pillar 3 carry character identity)

Player Motivation Profile

Primary Psychological Needs Served

Player Type Appeal (Bartle Taxonomy)

• [x] Explorers (primary) — every docking is new ground; the tide system itself rewards understanding
• [x] Achievers (secondary) — tier progression, base growth; but no 100% checklists — the tide makes completionism impossible by design
• [x] Socializers (secondary) — the retention engine for a friend group: shared home, shared plans, shared stories
• [ ] Killers/Competitors — explicitly not served; no PvP, no rankings

Who this is NOT for: competitors, checklist completionists, and players who need pause-anywhere solo play. The tide waits for no one.

Flow State Design

• Onboarding curve: the new-friend test (Pillar 5) — anyone can fight, gather, build, and craft within their first session; the home island is the safe tutorial space
• Difficulty scaling: spatial, not statistical — richer nodes and nastier inhabitants live deeper inland on each docked land; players choose their exposure
• Feedback clarity: chunky harvest bursts, telegraphed enemy attacks, the color-temperature tide clock — the world states its rules visually
• Recovery from failure: death is a setback, not a spiral — no full-loot corpse runs; tension comes from opportunity cost (what you didn't harvest), not loss of what you had

Core Loop

Moment-to-Moment (30 seconds)

The sortie beat: dash-and-strike combat with light kiting; mining and harvesting are fast, chunky interactions (three swings, satisfying burst of ore), never hold-to-channel. Enemies telegraph, you dodge, positioning beats gear-checks. Every fight and every node feeds the bag on your back — and the bag going home is the point.

Short-Term (5-15 minutes)

The run inland. A docked landmass is a wild zone bolted to your coast: richer nodes, rarer flora, and nastier inhabitants toward the interior. Cycle: gear up at base → cross the shore threshold → push, harvest, fight → haul back → smelt, craft, store. "One more trip inland before bed" — the land will leave, so every trip counts.

Session-Level (30-120 minutes)

A session is shaped by what's currently docked — the world sets tonight's menu, the group chooses from it. One friend smelts and fortifies, two push deep into the docked marsh, one hunts the coast. Natural stopping point: the haul is banked, the walls held, the forge is lit.

Long-Term Progression

Three braided arcs: character (identity expressed through crafted gear — flat base stats by design), material tiers (cloth → leather → bronze → iron, each era unlocked by the right lands docking), and the base (camp to keep — and prosperity attracts worse tides, so defense pressure scales naturally with growth).

Retention Hooks

• Curiosity: what docks next? What's deeper inland on the current land? What does a recurring land look like changed?
• Investment: the base, the gear you personally forged, the tide literacy you've earned (was "your drifted skills" — skill-drift cut 2026-08-25; corrected 2026-08-27)
• Social: the group's shared plans and shared near-misses; the tide moves whether you're there or not
• Mastery: combat skill, tide-schedule knowledge, efficient expedition routes

Game Pillars

Pillar 1: The Tide Waits for No One

The world moves in real time whether anyone is playing or not — the game never pauses the tide to spare you.

Design test: When choosing between "compensate players for absence" and "make returning exciting," we choose the living world. Absence creates stories and anticipation, never debt. Banned outright: grace-period timers, partial-credit harvesting after departure, catch-up resource boosts, and "the tide waited for you" mechanics of any kind. The only legal mitigation is recurrence — a missed land comes back eventually.

Pillar 2: Home Is the Fixed Point

The shared base is the emotional and mechanical center of the game; all content orbits it.

Design test: If we're debating where something should happen, the answer is "within sight of home." No far-flung quest hubs, no teleport networks, no second bases.

Pillar 3: Made by Our Hands

Progression comes from materials you hauled and gear you crafted — never from drops, quest rewards, or vendors.

Design test: When designing any reward, we give materials and knowledge, not finished items. If a sword exists, one of the players made it. **Ruins and defeated threats yield materials, blueprints, and knowledge — never finished gear.** An ancient ruin gives you the knowledge to forge the sword, not the sword.

Pillar 4: Scrappy Hands, Serious World

The controls are fast and joyful; the consequences are real. Friction lives in the world — tides, logistics, threats — never in your hands.

Design test: If a mechanic adds weight, ask where the weight sits. Slow menus, stamina-gated swings, encumbrance micromanagement → rejected. A land leaving with unharvested tin → embraced.

Pillar 5: Four Freedoms

Any friend can pursue any playstyle at any time — no classes, no assigned roles, no system that demands a specific player exist.

Design test: The new-friend test — if someone who's never played logs in tonight, can they immediately fight, gather, build, and craft without needing a role someone else filled? Emergent specialization by choice is fine ("Alex became our best weaponsmith"); mechanical gatekeeping is not (only "builders" can place advanced structures).

Anti-Pillars (What This Game Is NOT)

• NOT an RTS: You never command units from a detached view; you are always your hero, boots on the ground. Protects Pillars 2 & 5.
• NOT a content-treadmill MMO: No quest logs, no dailies, no XP-gated zones; the tide is the content generator. Protects Pillars 1 & 3.
• NOT a punishing survival sim: No hunger death-spirals, no full-loot corpse runs; tension comes from opportunity cost, not loss spirals. Protects Pillar 4. Corollary (load-bearing, per TD review): nothing happens to the base while offline — only the tide schedule advances.
• NOT a public live-service: Private world for 2–5 friends; no matchmaking, no monetization, no anti-cheat. Protects the actual goal and the timeline.

Visual Identity Anchor

Direction: "Tideline" (graphic silhouette, tide-as-protagonist)

One-line visual rule: *If it doesn't read at a glance while the world is moving, it doesn't belong.*

Supporting visual principles:

• Silhouette-first — bold, poster-like shapes with minimal interior detail noise (Don't Starve × Hyper Light Drifter clarity). Design test: any new asset must be identifiable from silhouette alone at gameplay zoom.
• The shape mismatch — home is soft, rounded, familiar shapes; drifting lands arrive with jagged coastlines that visibly don't match your shore. The mismatch itself says "this will leave." Design test: a screenshot must show at a glance which land is home and which is transient.
• The color-temperature tide clock — lands warm toward amber-gold as they approach docking and cool toward blue-violet as departure nears; the drift cycle is readable from the world itself, no UI timer required. Design test: if a tide-state needs a HUD element to communicate, first ask why the world isn't saying it.

Color philosophy: a restrained core palette (2–3 desaturated, stable "home" hues) so the tide can speak through light. Threat lands break the restraint with a hard saturated warning hue (crimson/acid) — on purpose and only then.

Why this direction: it makes Pillar 1 visible rather than merely mechanical, serves fast combat legibility (Pillar 4), and is the most solo-feasible art pipeline of the candidates — detail lives in silhouette and palette state, not painted texture, so biome count doesn't multiply art labor.

Inspiration and References

Non-game inspirations: tidal coastlines and harbor towns — the rhythm of things arriving and leaving on schedules older than you; Bronze/Iron-Age material culture (bloomeries, hammered metal, woven cloth).

Target Player Profile

Technical Considerations

The no-server architecture (TD-confirmed): the tide is a pure function of (world seed, wall-clock time) — the world is a save file plus math. The only mutable persistence is per-land harvest diffs and base state. No backend to build, host, or pay for. Constraint this rests on: nothing happens to the base while offline.

Risks and Open Questions

Design Risks

• Drift pacing is everything: too slow feels static, too fast punishes the group's real-life schedules. Needs tuning against actual friend availability.
• ~~Skill-drift progression may create visible power divergence between friends who play uneven hours~~ — RESOLVED 2026-08-25: skill-drift cut in favor of flat stats + gear-carried identity (Systems Design gate decision).
• "Emergent quality" of the tide must stay legible — players must be able to learn and anticipate the schedule, or anticipation collapses into randomness.

Technical Risks

• Combat netcode feel — RESOLVED (prototype-confirmed 2026-08-25): the gating spike confirmed host-authoritative NGO 2.x + local-only prediction delivers instant-feeling dash-and-strike combat at simulated 70–140ms RTT (see prototypes/combat-netcode-feel-concept/REPORT.md).
• Residual risk: real-network jitter, packet loss, and NAT behavior are untested — simulation covers delay only. CD-mandated condition: one real 2–3 person friend-group playtest over real internet connections before the netcode ADR is locked.
• Save forking if multiple friends host from stale copies — mitigate with timestamp checks on load; never attempt save merging.
• Partially-harvested land state must round-trip through the harvest-diff save contract (bounded, but must compose with docking).

Market Risks

• None in the commercial sense — the audience is the friend group. The real "market risk" is social: if the group goes quiet for a month, the tide keeps moving. Pillar 1 says that's a story, not a bug — but it's worth knowing going in.

Scope Risks

• Tier 1 as scoped is a 6–9 month outcome at part-time hours, not 3 months. At the 3-month mark, expect "solid MVP, thin Tier 1."
• Base-defense events are a second combat subsystem in disguise (base-aware AI, wave director, mechanically meaningful fortification) — scoped as its own line item, pre-designated first-to-slip.
• Save-anyone-can-host is a stealth-hard distributed-state problem — pre-designated first-to-slip; fallback is a fixed host.
• UI/UX for co-op (shared storage, who-carries-what) is chronically underestimated — watch it.

Open Questions

• ~~Does host-authoritative NGO + Steam transport combat feel Hades-adjacent at realistic latency?~~ → CONFIRMED by the netcode spike prototype (2026-08-25) at simulated 70–140ms RTT — see prototypes/combat-netcode-feel-concept/REPORT.md. Remaining CD-mandated condition: one real friend-group playtest over real internet before the netcode ADR is locked.
• Does the deterministic tide scheduler + harvest-diff contract survive contact with docking/undocking? → Answered by a 0.5–1 day standalone scheduler test plus a dock/undock round-trip prototype.
• What drift cadence fits the group's actual play schedule? → Answered by MVP playtests with the real friend group.
• How do sessions get organized — fixed host or rotating? → A 30-minute group decision, to be made before Tier 1 persistence work.

MVP Definition

Core hypothesis: *The venture → fight → harvest → haul → craft → build loop, under a real-time tide that takes unharvested resources away, is fun for 2 players across their first 3–4 sessions — base-up and the first recurrences.*

*(Window stated explicitly 2026-08-27 at the Technical Setup gate, cross-GDD review D3/D4: MVP validates the mechanical skeleton and the **tide fantasy** — a missed land reads as "damn, we missed it → it'll come back", the haul beat carries bank-or-push tension through base-up. It does not test sustained-session appetite: every MVP material sink is one-time and the economy flattens by session 3–4 by design (base-building OQ11), and the maker's-mark, walls-as-memory, and weeks-long combat-skill fantasies are Tier 1 (crafting, base-building, combat honesty clauses). A playtest read past session 4 is a Tier 1 question, not an MVP verdict.)*

**Gate 0 (before all other MVP implementation): the netcode spike.** 2–3 weeks. NGO + Steam transport, host-authoritative, local-player prediction only. Two real machines at throttled ~40–80ms latency. Go/no-go on combat feel.

RE-AFFIRMED 2026-09-02 at the Technical Setup → Pre-Production gate (CD + PR
concern; production/gate-checks/gate-technical-setup-to-pre-production-2026-09-02.md).
Gate 0 still blocks, and this re-affirmation sharpens what it blocks rather than
softening it.

What Gate 0 blocks — unchanged and binding: no art production, no crafting
implementation, no world-system implementation until the spike passes. Concretely:
no epics or stories may be written for any requirement routing through ADR-0008–0012
(netcode wire, player transform authority, inventory replication, structure registry,
enemy simulation), and no asset production may begin against them. The spike is
not the 2026-08-25 localhost prototype: that ran UnityTransport on one machine
with simulated latency and its own report names the gap — "the Steam transport +
real-network session is still pending." Gate 0 requires **two real machines
talking over the actual internet at real (not simulated) latency**.

Amended 2026-09-02 — the gate does NOT require Steam. QQ-04 research established
that NGO's NetworkTransport abstract surface is stable across 1.x→2.x (10 abstract
members, verified against NGO source), so the transport is swappable after the
playtest without invalidating anything it measures — the netcode model sits entirely
above a pure byte-pipe. The playtest therefore runs on UTP + Unity Relay (~1 hour
to first packet, free, no App ID, no router configuration), and Steam becomes a
ship-time concern. What Gate 0 measures is combat feel at real RTT and jitter,
which is a property of the network, not of who carries the bytes.

**What Gate 0 was never meant to block, stated explicitly because the ambiguity has
already cost a gate cycle:** *design and architecture work that is independent of the
wire.* The seven approved UX specs, the art bible, and ADR-0013/0014 were authored
after this gate was written and are correctly sequenced — every one of them is
netcode-independent by construction (ADR-0013 §7 "networking: none"; ADR-0014's
Ordering Note "does not depend on ADR-0008"; title-screen.md's deliberate
mechanism/payload split). Foundation and Presentation ADRs are likewise gap-free and
unblocked. Reading Gate 0 as "stop all work" would have idled the only work that
could safely proceed while friend-group availability — the one variable the project
does not control — sets the schedule.

The distinction, in one line: Gate 0 blocks building the game, not *deciding what
the game is*. Anything whose correctness depends on how bytes cross the wire waits for
the spike; anything that would be true on a single machine does not.

Consequence accepted at the gate: the spike carries an unpriced 1–2 day Steam
transport integration prerequisite that appears on no task list, and the playtest is
undated as of 2026-09-02 — six days past its stated week-1 deadline. Dating it is the
project's #1 action precisely because Gate 0 is being held rather than relaxed.

PROVISIONALLY PASSED 2026-09-11 (user decision; record:
production/gate-checks/gate-0-provisional-pass-2026-09-11.md). The Relay path was run
end to end on 2026-09-11 (two instances, real Relay hop, europe-west4) and the 2026-08-25
prototype had already confirmed the model at simulated latency; the friend session is
now a confirmation, not a precondition. ADR-0008–0012 may be written as Accepted with
a revisit clause, and the epics behind them may be storied. A NO-GO from the friend
playtest supersedes the affected ADRs and reopens their stories — that is the price of
not idling while a second person's calendar sets the schedule.

CONFIRMED GO 2026-09-11 — the friend playtest ran the same day (Jeroen host, Eric client, two networks, Unity Relay, ≈ 150–180 ms RTT): own actions instant on both sides, hits landed where the client saw the enemy; occasional remote stutter noted → register. Gate 0 is closed; evidence production/playtests/gate-0-2026-09-11/.

Required for MVP:

• 2-player co-op (host + 1) via NGO over UTP + Unity Relay (Steam at ship — ADR-0008)
• Home island, 1 dock socket, 2–3 authored land templates cycling on a real-time deterministic schedule
• Combat: 1 weapon type + dash; 2–3 enemy types with telegraphed attacks
• Harvesting: 2–3 resource types; cloth-tier crafting only (producer adjustment — bronze moves to the start of Tier 1)
• Basic base building: walls, forge/workbench, storage
• Host-owned save file with harvest-diff persistence

Explicitly NOT in MVP (defer):

• Bronze/leather/iron tiers, base-defense events
• 3rd–5th player (additive risk, not qualitatively new)
• Save-anyone-can-host, tide-clock visual polish (atmosphere, mist, drift-in/out cosmetics, multi-stop curves — the minimal single-channel amber/violet tint itself IS MVP, ruled 2026-08-27 at the Technical Setup gate so the anchor's no-UI-timer test can be run at the playtest), ruins/blueprints

Scope Tiers

\* Pre-designated first-to-slip items (producer adjustment): if the timeline compresses, these cut first, in this order: save-anyone-can-host (fallback: fixed host) → base-defense events (fallback: hostile lands are "go to them" content). This is a pre-agreed decision, not a negotiation under pressure. Skill-drift was fully cut 2026-08-25 (flat stats + gear-carried identity) and is no longer on the slip list — it is out of scope entirely.

Next Steps

• [x] Concept reviewed by creative-director (CD-PILLARS — edits applied), art-director (AD-CONCEPT-VISUAL — Tideline selected), technical-director (TD-FEASIBILITY — feasible with constraints), producer (PR-SCOPE — adjustments adopted)
• [ ] Fill in CLAUDE.md technology stack based on engine choice (/setup-engine — Unity)
• [ ] Create the art bible from the Tideline anchor (/art-bible)
• [x] Prototype the netcode spike first (/prototype combat-netcode-feel) — PROCEED, hypothesis CONFIRMED 2026-08-25 (CD-modified verdict)
• [ ] Real friend-group playtest (2–3 people, real internet, UTP + Relay) — CD-mandated gate, provisionally passed 2026-09-11; now the confirmation of /architecture-decision locks the netcode ADR
• [ ] Decompose concept into systems (/map-systems)
• [ ] Design each system (/design-system [system-name]) — use prototype learnings in Tuning Knobs and Formulas
• [ ] Write the netcode architecture ADR (/architecture-decision) — the one expensive-to-reverse choice; only after the friend-group playtest passes
• [ ] Build the vertical slice in Pre-Production (/vertical-slice)
• [ ] Validate core loop with the real friend group (/playtest-report)
• [ ] Plan first milestone (/sprint-plan new)