CodeWisp Lets You Build Real Games by Typing What You Want

The Macro: Game Development Is Still Way Too Hard

Making a video game is one of the most complex creative endeavors that exists. You need programming skills, art skills, sound design, level design, game balance knowledge, and the patience to debug physics engines at 2am. Tools like Unity and Unreal have made professional game development more accessible, but “more accessible” still means months of learning before you can ship anything playable.

There is a huge gap between people who want to make games and people who actually can. Roblox partially solved this with Luau scripting and its creator ecosystem. Scratch taught kids basic programming through game creation. But both have real limitations. Roblox games live inside Roblox. Scratch projects look like Scratch projects. Neither produces something you would proudly share as a “real” game.

The AI wave is hitting game development from multiple angles. Scenario.gg generates game art. Inworld builds NPCs. Ludo.ai helps with game concept brainstorming. But nobody has really nailed the “describe what you want and get a playable game” experience. That is the gap CodeWisp is targeting.

The Micro: A YouTuber Turned Game Engine Builder

Elvin Fu founded CodeWisp, and his background is unusually well-suited for this. He has been making games since he was 10. He taught game development to over 22 million YouTube subscribers. He built two game engines by himself. This is not someone who learned about games last year because AI made it trendy. He has been in the weeds for a long time.

The product lets you describe a game in plain English. CodeWisp generates the code, structure, and assets. You can refine with follow-up prompts. The output is a real, playable web game that you can share and publish. The platform currently hosts over 1,600 games, and 2,000 people are actively creating new ones every week. Those are real traction numbers for a product this early.

What I find compelling is the community layer. CodeWisp has categories like Featured, Top Liked, Recently Updated, Multiplayer, and 3D. There are leaderboards. There are learning courses like “Introduction to Coding” and “AI in Coding.” This is not just a generation tool. It is a platform with social features, discovery, and education built in. That matters because game creation tools live or die based on whether creators can find an audience.

Fu runs the company out of San Francisco with a team of three, part of YC Winter 2026 with partner Harshita Arora.

The Verdict

CodeWisp is smart because it is not trying to compete with Unity or Unreal. Those tools serve professional developers. CodeWisp serves everyone else. The 99% of people who have a game idea but will never learn C# or Blueprint visual scripting. That market is enormous if the AI generation quality is good enough.

The competitive risk comes from Replit, which has AI game generation capabilities, and from tools like Bolt.new that can generate web applications including simple games. But neither of those has the community, the courses, or the gaming-specific focus that CodeWisp is building. Platform effects in game creation are powerful. Once creators build an audience on CodeWisp, they do not want to leave.

In 30 days, I want to see the average session length for creators. Are people spending 10 minutes generating one game and leaving, or are they spending an hour iterating on a project? That tells you whether the refinement loop works. In 60 days, the question is whether any CodeWisp game has gone viral. Distribution proof matters enormously in consumer products. In 90 days, I want to see multiplayer adoption rates. If people are building multiplayer games that other people actually play, CodeWisp has built something much bigger than a generation toy.