Building in Public
Every win, every lesson, every experiment documented. The process is the product.
Enhanced 'How It's Being Built' section on Stax case study page. Expanded from 3 to 6 steps with icon support and WHY-focused copy. New steps: Architecture Before Code, Frozen Domain Layer, Claude as Co-Pilot, Right Tool Right Job, Security at the Database, Tests Prove It Works. Each step explains reasoning, not just facts.
Enhanced site navigation with active state indicators, expanded projects dropdown (added AION, Ronin, Sulat, Genkaw Ecosystem with external link support), and smart scroll behaviour—nav hides on scroll down, reappears on scroll up. New hooks: useActiveSection (IntersectionObserver for section tracking), useScrollDirection (scroll position detection).
Complete content architecture reset. Wiped all non-Brunei content, rewrote Brunei directives in correct engine format. 30 directive files, 8 entity files, 10 storyline files cleared. Brunei now has full directive coverage: diplomacy, economy, domestic, factions, individuals. Clean slate for country-by-country content expansion.
Full frontend accessibility pass. WCAG-compliant color contrast, keyboard navigation, screen reader support, focus indicators. Improved loading states, error boundaries, and responsive layouts. Production-ready UX for all users.
Implemented core strategic mechanics: economic sanctions with trade impact modeling, alliance formation with mutual defense clauses, military operations with logistics and intel requirements, espionage system with counterintelligence. Each system integrates with existing entity relationships and resource pools.
Added achievement system with 6 trigger conditions: storyline completion, directive success streaks, relationship thresholds, resource milestones, entity unlocks, and time-based events. Persistent tracking with unlock notifications. Adds replayability and progression feedback.
End-to-end trade mechanics now operational. Trade agreements affect resource flows between states, tariffs modify relationship dynamics, trade routes visualized on map. Integrates with sanctions system—sanctioned states see trade penalties. Economic interdependence modeled realistically.
Component consolidation pass. Fixed data inconsistencies between case study pages and main content. Updated tooling configurations. Cleaner imports, consistent styling patterns across project showcase pages.
Technical debt sweep. Fixed counterintelligence edge cases, ally strength validation in military calculations, strict TypeScript typing across 40+ files. Eliminated race conditions in state updates, improved exception handling in engine core. Backend stability for extended play sessions.
Complete UI overhaul to map-centric command center design. Interactive world map as the primary interface, country profiles on selection, diplomatic action panels. Strategic overview at a glance. React/Vite frontend with Python backend.
Built real-time training dashboard with WebSocket streaming. Visualizes training metrics (reward, loss, episode length), live game state rendering, and hyperparameter display. FastAPI backend on port 8085. Watch AI learn Tetris in real-time.
Implemented flexible bank format adapter system for CSV imports. Supports multiple bank formats (ING, generic CSV) with automatic detection. Data pipeline processes uploads through format adapters, validation, and database insertion. 4 CSV imports/month on FREE tier.
Launched Ronin—a PyTorch-based game AI framework for reinforcement learning. First target: Tetris via PyBoy Game Boy emulator. Custom TetrisEnv with reward shaping for line clears, piece placement, and height management. PPO algorithm with CNN policy network.
Added multi-account support with account filtering across all views. Users can manage multiple bank accounts, credit cards, and cash accounts. Transaction filtering by account in analytics and transaction list. Smart API URL detection for dev vs production environments.
Complete hardening pass on analytics cards. Added useReducedMotion (WCAG 2.3.3), useMemo for derived values, safeNumber guards against NaN/Infinity. Improved skeletons to match actual layout (prevents CLS). Fixed Recharts tooltip contrast and disabled ugly default hover. Cards hardened: spending-over-time-chart, trend-card, trends-section, period-comparison. All now 9.5-10/10 quality.
Major modularity refactor creating shared analytics utilities. New safeNumber() for NaN guards, getStatus() with presets (health, savings, runway, ratio) for unified 5-tier color system. Created StatCard, AnimatedProgressBar, AnimatedListItem components. Added aria-hidden to 60+ decorative icons. Eliminated ~200 lines of duplicate code across 25 analytics cards.
Implemented hybrid analytics interface with collapsible sections and card customizer. 4 collapsible sections with localStorage persistence. 3 intent-based presets: Quick Glance (10 cards), Weekly Review (17 cards), Full Analysis (35 cards). Users can show/hide individual cards. All preferences persist across sessions. Hero section (5 core KPIs) always visible.
Complete email infrastructure for genkaw.com now live. Mailcow (18 Docker containers) with Cloudflare Email Routing for inbound, Resend SMTP relay for outbound. No port forwarding required. Fixed Mailcow quirks: SOGo auth (password hash bug), quota storage (bytes not MB), sender-dependent relay (relayhosts vs transports tables). 4 mailboxes, SOGo webmail, Let's Encrypt SSL.
Completed comprehensive training system with 11 challenges across 4 difficulty tiers, sparring matches with mastery breakdown, and performance tracking. Added challenge/sparring result UI with visual feedback showing bonuses earned. Best performance tracking (fastest win, most damage, longest combo). 148 tests, build clean, lint clean.
Complete dashboard overhaul. Replaced abstract KPI cards with actionable layout: BalanceCard hero (greeting + net balance + income/expenses), RecentTransactions list (8 items), and HealthScoreCard sidebar. New traffic-light color system for financial health (green→teal→amber→orange→red). Shared getHealthGradient() utility for consistent theming across components.
Added .dockerignore files reducing build context from 7.52MB to 15.59KB (99.8% reduction). Introduced Docker Compose profiles for selective builds: --profile web for frontend, --profile api for backend. Build times dropped from ~90s to ~30s. ADR-006 documented in architecture.
Systematic documentation audit identified and fixed 8 gaps across 10 files. Updated all READMEs with Docker profiles, added health gradient system to CODE-DNA, created ADR-006 for Docker optimization, removed dead links, updated session history. All docs now current.
Complete UI transformation to match Genkaw aesthetic. Dark-first theme with true black (#000) background, glassmorphism cards (backdrop-blur, translucent bg), extralight typography (font-weight 200), blue→purple gradients. All 9 dashboard pages restyled with gradient icon boxes, generous spacing (space-y-10), and scroll reveal animations. 150+ lines of new CSS utilities added.
Fixed all AI Coach blockers. Added LLM provider selector (Ollama active, Claude coming soon). Fixed SSE streaming in Next.js API proxy—responses now stream in real-time instead of 30-second timeout. Fixed EntitlementService constructor, enabled AI Chat for FREE tier (20 messages/month), and forwarded X-Session-Id header for chat session persistence.
Major UX improvements for transactions. Added year picker to analytics (2024 data now visible). Implemented infinite scroll as default mode (max 500 items for performance). Added DELETE /all endpoint with confirmation and 'Clear All' button. Better loading states with IntersectionObserver for auto-loading.
Complete visual overhaul of the Stax case study page on genkaw.com. Matched the Stax landing page aesthetic: dark theme (#0a0a0f background), blue/purple gradient accents, glassmorphism cards. Forced dark mode on project pages with theme toggle hiding. Fixed grid tile height consistency.
Completed Phase 3 Industry Leader differentiation. 11/11 enterprise-grade features: CQRS read models, full event sourcing with snapshots, TypeScript + Python SDK generation, webhook circuit breakers (Stripe-style signing), GraphQL federation, Redis caching with stampede protection, Row-Level Security on all user tables, dead letter queue, DAST scanning (ZAP + Schemathesis), automated feature flag rollback, and behavioral snapshot testing. Architecturally aligned with Netflix, Stripe, and Google patterns.
AI Coach feature now live with Ollama integration. LLM provider selector in UI, SSE streaming fixed in API proxy, real-time chat responses. GPU-accelerated local inference with llama3.1:8b—free AI features for all users. 20 chat messages/month on FREE tier, 5 financial insights/month. Claude API integration coming next.
Fixed all production-blocking issues. Same-origin API proxy eliminates CORS. Timezone-naive datetime handling for PostgreSQL. Empty state handling for new users. Row-Level Security migrations corrected. 6 database migrations applied, 379 frontend tests passing. Registration → Login → Dashboard flow operational at stax.genkaw.com.
Major architectural overhaul implementing professional component patterns. Created reusable case study components (HeroSection, FeaturesSection, TechStackSection, etc.), extracted content to TypeScript data files, added 6 ADRs documenting decisions. Stax page reduced from 556 to 143 lines. Added accessibility improvements (aria-current, skip-to-content), scroll progress indicator, and contact form.
Full visual overhaul of beggars-sect.genkaw.com to match Genkaw aesthetic. Added ScrollReveal animations, ink wash gradients, chi flow SVG effects, corner bracket decorations. Created 2 new sections (Characters, Progression). Extralight typography, red accent colors, floating mist effects. 1,704 lines added across 21 files.
The Beggars Sect showcase page on genkaw.com now permanently displays in dark mode. Theme toggle hidden on this route. The ink-and-blood aesthetic requires darkness—light mode would break the visual identity. Consistent with the Stax page dark mode implementation.
Completed full migration to vertical-slice architecture. Deleted 3 facade directories (app/application/, app/domain/, app/services/), updated 60+ imports across 40 files. Each feature now owns its domain, application, and infrastructure layers. 1,741 tests passing. 71% reduction in direct ORM access (28→8 session.exec calls). Event-driven cache invalidation and budget monitoring now automated.
Production-ready frontend complete. P0 security: in-memory token storage (XSS protection), automatic refresh, session expiry warnings. P1 safety: 500-item bulk limits, idempotency headers (Stripe-style). Fixed API prefix/envelope issues. Added error boundaries at feature level. Complete API layer: billing, gamification, budgets modules. 95% code quality score, 0 magic numbers.
Week 1 hardening complete. F1: XP exploit closed with idempotency keys—no more infinite XP from spam-categorizing. F2: Event handlers now idempotent. F5: Ownership checks on all advice endpoints. New observability: outbox pending depth, handler failures, circuit breaker metrics. Health endpoint now shows outbox/DLQ/circuit status. 715 tests passing.
Implemented Application Layer Blueprint. Transaction, Budget, and Category CRUD now use proper use cases with Unit of Work pattern. Repositories use flush() not commit()—UoW owns transaction boundaries. 4 CQRS query handlers added. XP exploit fixed at architecture level (Law #12: verify category actually changed). 714 tests passing.
Locked down the Stax infrastructure layer with formal policy documentation and automated enforcement. CI gates now run tripwire tests on every commit—if critical functionality gets removed, the build fails. CODEOWNERS protection requires review for infrastructure changes. The foundation is stable enough to focus purely on feature development.
Completed exhaustive testing of the Stax infrastructure layer—data storage, message queues, error recovery, logging. 71 new tests covering 16 failure scenarios (database disconnects, queue failures, etc.). Closure statement: if it breaks in production, it's either an external service failure or an unknown unknown—not a missing test.
Completed Phase 7: Production Operability for Stax. Health dashboards provide at-a-glance system status. Automated alerts trigger before users notice problems. Five incident runbooks document step-by-step fixes for common issues. Chaos tests verify recovery from intentional failures. The gap between 'side project' and 'production software' is now closed.
The Stax domain layer (core business logic—XP calculation, transaction categorization, budget rules) is now frozen at v1.0.0. Established a 'constitution' with 8 invariants the system must always maintain. Fixed edge cases including backdated transaction XP gaming. 658 tests passing, quality score improved from 78/100 to 91/100. Features can now be built on a stable foundation.
Established a new documentation format optimized for AI-assisted development. Tables instead of paragraphs, TL;DR at the top of every section, bullet points over prose. Claude AI can now parse project context in seconds rather than scanning walls of text. Applied across all projects. Counterintuitive but measurably effective.
Expanded SYSTEM-DNA to 56 principles organized into 8 categories: Foundation, Quality, Architecture, Data, Security, Operations, Integration, and Meta. Each principle includes definition, rationale, implementation checklist, and anti-patterns. Every project inherits these standards. Consistency across multiple codebases from a single source of truth.
Extracted all hardcoded values from the Stax codebase into configuration files. Gamification (XP formulas, tier boundaries), health scores (weights), and budget validation rules are now documented and testable. 274 new tests verify behavioral equivalence. Cleaner code, easier modification, no magic numbers.
Created a 10-point scorecard to audit Stax against Domain-Driven Design principles. Initial score: 72/100. Strengths: zero database code in business logic, clean event system. Gaps: time handling needs improvement, event systems should be consolidated. Four-phase improvement plan created. You can't improve what you don't measure.
Added Architecture Decision Records (ADRs) to Stax—documents explaining WHY technical choices were made, not just what was built. ADR-007: Event system unification rationale. ADR-008: Aggregate + Decider pattern selection. Future developers (or future sessions) now have documented reasoning for architectural decisions.
Evolved CODE-DNA from documentation into an enforceable audit system. Every documented pattern becomes a requirement with compliance scoring and violation tracking. Best practices are only useful if actually followed—now there's a system to verify adherence and prioritize fixes.
Rebranded from FinPlanner to Stax with a new 'Financial Warrior' identity. The platform gamifies budgeting with XP, levels, and achievements—the name should reflect that energy. New emerald color scheme, updated URLs and branding. Same dual-database architecture, now with an identity that matches the product vision.
Completed a 10-phase infrastructure initiative for Stax. Added: structured logging, health checks, graceful shutdown, 70+ metrics, caching, circuit breakers for dependency failures, background task processing, and feature flags. 658 tests passing (91/100 score). The platform now meets enterprise reliability standards.
Reorganized 87 Stax documentation files into a coherent structure. Created ARCHITECTURE.md with system diagrams. Created API.md documenting all 89 endpoints. Consolidated scattered session logs. Fixed code patterns to match documented standards. Documentation now matches implementation.
Visual refresh across genkaw.com. Added entrance animations, hover effects with subtle lifts and glows, pulsing indicators on active projects. Replaced text-based architecture diagrams with interactive visual versions. Micro-interactions throughout reward exploration. The difference between functional and polished.
Comprehensive quality pass on genkaw.com. SEO: 100/100 Lighthouse score, auto-generated sitemap, structured data. PWA: installable on mobile with offline support. Security: A+ rating. Accessibility: 100/100. Performance: 99/100. Analytics added. Production quality for a portfolio platform.
Built SYSTEM-DNA v2.0—an 8-part document governing software development. Covers: Principles, Research, Design, Development, Lifecycle, Maintenance, Operations, and Improvement. Each project extends these universal rules with project-specific additions. The codification of 'good' that separates amateur from professional.
Major Stax release: full gamification is now live. Users earn XP for healthy financial habits, progress through 50 levels, and unlock 28 achievements. The Financial Warrior dashboard displays progress. Security hardening includes password hashing, rate limiting, and security headers. Dual-database architecture (PostgreSQL + DuckDB) achieves 5.6× faster analytics.
Polished the AI system controlling enemy combat behavior in The Beggars Sect. Extracted all settings to a single configuration file. Added comprehensive documentation. Result: 5-10% performance improvement, zero breaking changes for 15+ existing enemies. Functional and clean.
Completed major code cleanup for The Beggars Sect. Centralized all game balance values into one configuration file. Created reusable utilities, eliminating 60-79% of duplicated code. Built a UI design system. Professional-grade code organization—the same standards used in commercial software.
Built production infrastructure for The Beggars Sect: 36 unit tests covering core systems, 20-37% faster combat through caching, error handling that auto-recovers from 70% of crashes. Fixed 10 critical bugs. From 'it works' to 'it's reliable.'
Built the character progression system: 8 chi aspects to master (skill tree equivalent), training grounds for practice sparring, mastery points from victories, 6 starter techniques tied to chosen philosophy. Player choices shape abilities.
Completed demo polish for The Beggars Sect. Story flows smoothly with proper pacing, combat targeting is clearer, save system works reliably. Playable demo ready to share. Core systems proven, ready for Chapter 1 content.
Built the save/load system: 3 manual save slots plus auto-save. Saves trigger at important moments (choices, combat, chapter transitions). Released v0.2.0. Essential RPG feature complete.
Built the complete story engine: narration, dialogue, player choices, exploration. Created a 7-scene Prologue (~20-25 minutes of gameplay). Story and combat integrate seamlessly—make choices, fight battles, continue the story.
Built the Active Time Battle combat system: speed determines turn order, 15+ techniques with combo potential, AI enemies with distinct personalities and boss phases. Chi management, stance switching, combat log. The game is now playable.
Built the core engine: character system with 11 enemy types and 3 bosses, save/load functionality, title screen and menus. Published v0.1.0—the first runnable version. Design docs became working code.
Redesigned genkaw.com from a wall of text to a navigable hub. Preview cards for each project, only 3 recent updates shown, cleaner About section. Better visual hierarchy—visitors scan quickly and dive deeper where interested. Sometimes less content means better communication.
Upgraded the n8n automation hub from SQLite to PostgreSQL 16 for better performance and reliability. Added Ollama for local AI model execution. Automated backup scripts. Security hardening with network isolation. Enterprise-grade automation foundation.
Launched a dedicated showcase website at beggars-sect.genkaw.com. Documents the story, combat system, chi cultivation mechanics, and development progress. The game has its own web presence.
Designed the branching narrative structure: 3 philosophical paths (Blade, Stream, Shadow), 3 possible endings (Destroyer, Reformer, Wanderer), 15,000+ lines of planning documentation. Every playthrough differs based on player choices.
Massive world-building session: original chi cultivation system with 8 aspects, 12 martial arts sects, complete history timeline, deep character profiles. Core innovation: Inverse Chi—drawing power from emptiness. When you build a world, stories write themselves.
After 2,000+ lines of design docs, scoped down to a 12-week MVP. Created planning documents, designed the feedback system, built a dev log page. Building in public forces realistic planning.
Added quality-of-life features to genkaw.com. Dark mode toggle cycles between light/dark/system. RSS feed for update subscriptions. Smooth page transitions. Social sharing images. Self-hosted fonts for faster loading. The small features that make a site feel complete.
Created a formal styling guide documenting the Japanese minimalist aesthetic. Spacing rules, color palette, typography, divider styles—applied consistently across all pages. A design system keeps the site cohesive as it grows.
Added a Projects dropdown to navigation for quick access to project showcase pages. Built dedicated pages for each experiment with progress trackers and dev logs. Building in public means making the work visible—not buried in posts.
Added Open Graph meta tags for proper social media previews. New minimalist favicon. Custom 404 error page matching the Japanese aesthetic. Details that make a platform feel professional.
Restructured genkaw.com with a data-driven architecture. All content in one file, reusable components, trivial to add new sections. A platform designed to evolve with every new project.
Launched bathala.genkaw.com—a secure web terminal for remote homelab access. Cloudflare Tunnel (no exposed ports), web-based terminal (no SSH client needed), 2FA authentication. Server management from any device, anywhere. Named after the Filipino supreme deity.
Started The Beggars Sect—a CLI RPG about martial arts and chi cultivation. Design-first approach: complete documentation before code. Professional software practices applied to game development. Part of the Martial Arts Haven universe.
Defined the strategic mission behind all projects: building toward financial independence through automation and AI capabilities. Not random learning—systematic capability building. Claude AI + n8n as the core power duo. Every project serves this larger goal.