Three Years at Panteon: MMOs, Tower Defense and Puzzle Games

Hi! Today marks a good moment to look back: three years ago I walked into Panteon as a junior client developer, still midway through École 42. Since then I have worked on three very different games — an MMO, a live tower defense and a puzzle title — and each one taught me a different side of this craft.

2023–2025: Kingdom of Pandoria — building big

My first production codebase was Kingdom of Pandoria, a midcore MMO strategy game published under three names during its soft launch. I spent two years deep in its UI layer (built on an in-house data-binding framework), implemented the Wave Defense and RPG Battle minigames, and worked on performance infrastructure: a runtime dynamic texture-atlas system and a shader warm-up pipeline. The project was eventually cancelled — and that, too, was a lesson. You can pour solid engineering into something the market does not want; the engineering still compounds, the code does not ship, and you carry the skills to the next game.

2025: Raid Rush — respecting the live game

Next came Raid Rush: Tower Defense TD, live since 2023 with a real player base. Working on a live game is a different sport: I redesigned the time-limited Flash Events, migrated a legacy economy system to a new one — with save-data migrations that had to be flawless for existing players — and learned that the scariest line of code is the one that touches somebody’s years-old save file.

2026: Stack Sort — small game, sharp architecture

Now I work on Stack Sort: Color Puzzle (Unity 6) and new prototypes built on our in-house Puzzle Template. The codebase is the cleanest I have worked in: a command-driven ECS core where every move is a validated, recordable, undoable command. I have shipped special tile features end-to-end, account flows and editor input-simulation tooling — and AI-agent-assisted development has become a normal part of how our team works.

What three years taught me

• Scale teaches architecture: you only really learn module boundaries when sixty subsystems must not step on each other.
• Live games teach humility: every change meets real players and real save files within weeks.
• Cancelled projects still pay: the skills survive the product.
• Tools multiply you: the best thing I build is often not a feature but the thing that makes features faster to build.

Here is to the next three. If you want the deeper technical stories, each game has its own page in the Works section.