[Beginner] · What to build first, in what order, and the specific ways beginners lose two years.
The three failure modes
1. Building an engine instead of a game. You will learn a lot and ship nothing. If the goal is understanding, that’s fine — say so. If the goal is a game, use an engine → engines.
2. Scope. The classic first project is an open-world multiplayer RPG. It will not be finished. Every experienced developer’s advice converges on the same instruction: make something absurdly small, and finish it.
3. Endless tutorials. Following a tutorial teaches you to follow tutorials → tutorial hell. Do one, then build something different without one.
The ladder
Each step introduces exactly one new thing. Finish each — including menus, sound, and a build someone else can run. The last 20% is where the learning is.
1. Pong. Input, movement, collision, score, a win state. A weekend, and you’ll still learn something.
2. Breakout. Many objects, level layouts, particles, juice — screen shake, hit pauses, sounds. Juice is the difference between a mechanically correct game and one that feels good, and it’s a real skill.
3. A platformer, one level. Character controller feel — coyote time, jump buffering, variable jump height. Hand-authored movement, not physics → physics. Tile maps, a camera that follows well, checkpoints.
4. A top-down shooter or roguelike. Enemy AI (state machines → game AI), spawning, pathfinding, procedural levels, progression.
5. Something 3D and small. Transforms and cameras in 3D → transforms, lighting, models and animation.
6. Then pick a direction — graphics, gameplay systems, tools, engine, networking.
Ship each one. itch.io is free and takes ten minutes. A finished Pong on the internet beats an unfinished RPG on your disk, and the act of finishing is the skill you’re actually training.
Game jams
The single highest-value activity in this list. Ludum Dare, GMTK Jam, or any of the hundreds on itch.io: a theme, 48 hours, ship something.
They work because they force the thing you’ll otherwise avoid — brutal scope discipline and an actual finish line. You’ll produce more learning in a weekend than in a month of tutorials, and you get a finished artefact and feedback from strangers.
Do one before you decide whether you like game development. It is the cheapest possible test of the real experience.
What to learn, and when
Now: vectors, the game loop, state machines, one engine.
Soon: transforms, basic shaders, spatial partitioning, profiling.
When you need it: advanced rendering, physics internals, networking.
Probably never: most of the graphics roadmap, unless you go into graphics specifically.
The roadmap.sh game-developer roadmap is enormous — 150+ topics, heavily weighted toward graphics and physics research. Read it as a map of the field, not a curriculum. Working through it linearly before making a game is a way to spend two years and ship nothing.
What this vault already gives you
This is the useful part, and it’s why game-development is a thin folder rather than a thick one. A surprising amount of the roadmap is already written here for other reasons:
| Roadmap area | Already in this vault |
|---|---|
| Linear algebra, vectors, matrices, transforms | transforms · rigid body transforms |
| Quaternions | robotics 04 |
| Rendering, rasterisation, PBR, shading, ray tracing | computer graphics — 9 notes |
| GPU pipeline, shaders, parallelism | GPU & parallel computing |
| Integration, ODEs, numerical stability | numerical methods |
| Trees, BVH, spatial partitioning, A*, graphs | DSA · graph theory |
| Neural nets, decision trees, RL, naive Bayes | AI & ML |
| Memory, caches, data-oriented design | computer architecture |
| C++, Rust, Python | languages |
| Networking, UDP, latency | networking |
| Rigid body dynamics, materials | continuum mechanics |
Game development is largely an integration discipline. The maths and systems are the same ones robotics, graphics and simulation need — which is exactly why the skills transfer outward, and why the time isn’t wasted even if you never ship a commercial game.
Resources
- Godot docs — genuinely excellent, and the best free way in
- Fiedler, Fix Your Timestep! — the loop, properly
- Gambetta, Fast-Paced Multiplayer — prediction and reconciliation
- Game Programming Patterns (Nystrom) — free online, and the best book on the subject. Read it after your second project
- Real-Time Rendering (Akenine-Möller et al.) — the graphics reference
- Game Engine Architecture (Gregory) — if you want to know how engines work
Related
- what game development is — including the industry, honestly
- what to build next — the same argument, generally
- one active course — read this before starting
- project ideas — the vault’s build list
Source: [reference] — cross-referenced against roadmap.sh game-developer.
Software engineering
Infrastructure & security
Computer science
Hardware & engineering