About the project
ScubaWord is an iOS-first word-search diving game. Players drag horizontally, vertically or diagonally to find target words, remove letter tiles and reveal the next underwater grid beneath them.
The playable SwiftUI and SpriteKit campaign combines deterministic puzzle generation with tactile selection, animated level transitions, score feedback, accessible motion alternatives and an asset-backed treasure sequence.
Developed in partnership with Codex
Codex has supported the project from requirements and architecture through implementation, gameplay tuning, test coverage, visual development and delivery records. The linked summaries below document that work as a sequence of reviewable product decisions rather than presenting AI involvement as a single undifferentiated claim.
- PlatformNative iPhone app
- TechnologySwift, SwiftUI, SpriteKit and XCTest
- CampaignTen procedurally arranged levels
- WorkflowProduct-led, Codex-assisted engineering
Development records
Codex session summaries
Each record captures a stage of the project: the request, the decisions made, improvements delivered, verification performed and any work still open.
· Session progress record
The equipment room now guides the whole journey.
This phase turned the equipment room into a clearer mobile starting point. Tappable objects gained an optional shimmer, the new Help Wanted guide explains the complete play loop, Settings and Help became consistent full-screen destinations, and the safe and globe now share a practical travel shop. Screenshot-led corrections also aligned dive-log and reef imagery, repositioned the visa call to action, and simplified navigation throughout.
Read the session summary· Session progress record
Reef travel became a clear, visual journey.
This phase moved ScubaWord from deployment readiness into a complete exploration experience: upload blockers were cleared, 36 Field atlas destinations and seven reef maps were designed and implemented, travel credentials were introduced, and the interface was simplified through repeated screenshot-led refinements. Commerce and rewarded-ad foundations were also prepared, while remaining disabled until their secure back-end connections are ready.
Read the session summary· Session progress record
Collection worlds and the equipment room came together.
This phase expanded the collection experience across every reef, gave regional habitats clearer identities, and turned the equipment room into a more useful, object-led hub. Prompt-led refinements also improved the dive-complete flow, word-grid behaviour, navigation, progress feedback, and development-tool performance.
Read the session summary· Multi-session progress record
The dive now feels like a dive
Across the latest design and implementation work, ScubaWord moved from a functional paper-toned interface toward a coherent underwater expedition: habitat-led collections, an air-limited dive mode, a calmer reef wordgrid, a single-view equipment room, and a treasure reveal that completes the journey.
Read the session summary· Multi-session progress record
The journey, economy and collection experience converged into one coherent reef adventure
Three parallel workstreams moved ScubaWord from economy questions and interface exploration into tested native progress. Reef access now has a fair, explicit gem path; postcards, approval stamps, briefing and dive-log screens share an accessible visual language; and the collection redesign gives wildlife believable habitats with a clear living-reef reward for completion.
Read the session summary· Session progress record
Core stability landed, then six reefs became complete collectible worlds
This interval moved from engineering foundations into a major content expansion. Campaign starts, background preparation and save recovery became safer; high-cost visual work moved away from time-sensitive rendering; six more reef poster collections added 60 production stickers; and the full 70-sticker set now flows from in-app ownership into the Messages keyboard.
Read the session summary· Session progress record
A production treasure reveal, refined to stop at its strongest frame
This interval first clarified the remaining route to release, then turned a new treasure-chest concept into a six-stage in-game reveal. A final prompt sharpened the experience further: the animation now ends on the open, gem-filled chest instead of continuing into a separate scattered-gem reward sequence.
Read the session summary· Session progress record
Grounded maps, a polished treasure finale and stronger grid variety
This interval converted visual feedback into measurable map constraints, corrected reef markers against real geography, and rebuilt the treasure finale around a settled chest and varied gems. Follow-up prompts then improved the reward animation, reduced repeated grid patterns, preserved the explicit orientation balance, and commissioned a whole-product performance and stability review that surfaced the next important engineering risks.
Read the session summary· Session progress record
Clearer navigation, richer feedback and genuinely balanced word grids
This interval polished the app-wide presentation and navigation, promoted the approved equipment-room artwork, and made grid quality measurable. When the first catalogue audit revealed that more than half of placements were vertical, the follow-up prompt turned an earlier design preference into an exact three-way production rule. The final audit confirmed balanced horizontal, vertical and diagonal placement across every site and both play profiles.
Read the session summary· Session progress record
From repetitive word pools to a faster, more varied dive campaign
The session redesigned and integrated the full gameplay vocabulary, strengthened map and grid behaviour, and added a more polished launch and return experience. The result is a broader 36-site campaign with sharply lower repetition, faster catalogue access, clearer visual feedback, and a documented path for the remaining visual checks.
Read the session summary· Session progress record
Production maps, deeper dive progression and practical development tools
This interval strengthened the full dive loop: word grids became more deliberate, regional and local maps were rebuilt around real geography, the equipment room became brighter and more welcoming, treasure and gem presentation improved, and new development tools made progression, balances and word-pool auditing directly controllable.
Read the session summary· Session progress record
From equipment-room concept to a verified poster collection
This interval turned a broad home-screen rethink into a functional dive-base room, established a fair gem economy, replaced fragile room navigation with a native interaction model, and delivered the first production poster collection with persistent sticker ownership and gem purchasing.
Read the session summary· Session progress record
From stable word grids to a mesh-driven home-screen clownfish
This interval strengthened both gameplay reliability and development visibility, then moved into a layered animated home-screen character. All locations gained validated word pools and audit tools, unused shipping resources were removed safely, and the clownfish was rebuilt from registered mesh-driven layers before its fin assets were cleaned for production use.
Read the session summary· Session progress record
Responsive level preparation and 100-word replay variety
This interval removed a serious launch-time stall and completed the next step in replay variety. Grid generation now prepares only the current and immediately needed layers without blocking the interface, while each prototype vocabulary profile has grown to 100 curated words so a normal ten-level dive can use 75 distinct targets and an immediate replay can deliberately surface all 25 previously unused words.
Read the session summary· Session progress record
Progression, performance and replay variety became a complete game loop
This interval moved beyond the first playable journey into a persistent progression system with multiple modes, rewards, cosmetics and researched dive content. Device feedback then sharpened regional-map presentation, Hard-mode layout and animation performance, before a replay-aware word-pool prototype made repeat campaigns more varied without ever risking a missing grid.
Read the session summary· Session progress record
A true textured globe with surface-attached reef markers
This interval corrected the globe’s interaction model and then refined every layer that made it feel artificial. The approved illustration now wraps a continuously rotatable sphere, swipes carry natural momentum, and reef markers have moved from hovering screen overlays into the globe material itself.
Read the session summary· Session progress record
From a centred word grid to an illustrated world-selection journey
This interval refined the core play layout and then built a complete route from the home screen to selectable reef locations. Prompt feedback progressively raised the bar from functional navigation, to recognisable geography, to direct use of the approved illustrated artwork when procedural maps still felt visually out of place.
Read the session summary· Three-session combined progress record
From world structure to a playable, polished home experience
Reef destinations, an accessible collectible-chest sequence and the development of the game’s initial home experience into a verified visual baseline.
Read the session summary· Combined progress record
From prototype friction to an asset-ready treasure loop
Gameplay composition, score-flight tuning, rapid finale testing and the move from placeholder shapes to catalog-backed collectible rewards.
Read the session summary