Building LumiGameLab: The Quiet Work of Curating Educational Games

The idea for LumiGameLab started with a simple observation: most "educational" games on the internet aren't actually educational.

They look educational. Bright primary colors, friendly mascots, words like "math" and "science" on the splash screen. But play them for ten minutes and you start to notice the same patterns. The math is incidental. The science is decorative. The actual learning content is brief, easy, and disconnected from the fun part of the game.

I wanted a place that filtered those out. A place that only included games where the learning was actually the point.

What I didn't fully account for was how much manual work that filtering would take.

The Vetting Problem

The core challenge with curating educational games isn't building the website. It's the games themselves. There are a lot of them on the open web — thousands, easily — and most of them aren't very good. The only honest way to find the ones that are good is to play each one for long enough to see how it actually behaves.

Long enough turned out to mean about thirty minutes. Anything less and you're only seeing the introduction levels, where every game is gentle. The thirty-minute mark is roughly when you can tell whether the difficulty actually scales as the player improves, or whether it flattens after the first few rounds. A surprising number of educational games flatten. The kid plays the same easy challenge dressed up in different colors for hours, feels like they're doing something, and isn't actually being pushed.

Multiplied across the hundreds of candidate games I've evaluated, that's a lot of hours of playing things I had no intention of including. Most of the time spent on this site is invisible to anyone who visits it.

When the "Subject" Category Stops Being Useful

Early on I assumed I'd organize the site the way every other educational games site does — by subject. Math, science, language, geography. That's how parents search, so that's how the categories should work.

I held that assumption for about a month before it broke.

The problem became obvious when I tried to tag a specific game. It involved sorting countries by various attributes — population, area, language. Was it geography? Yes, sort of. But the actual cognitive work the kid was doing was pattern recognition and logical sorting, not memorizing facts about countries. A different game on the site, also tagged geography, was almost pure memorization. Same subject, completely different mental skill being exercised. A parent looking for "geography" would get one or the other and have no way to know which.

That's when I added a second axis to the tagging system. Subject is one axis. The actual cognitive skill the game exercises is the other. A game can be in the math subject category but tagged for spatial reasoning, or in the language category but tagged for working memory. Most parents browse by subject because that's how they think. But the skill axis is the one that actually matters for which game will help with what.

This sounds simple in retrospect. Getting the skill tags consistent across a few hundred games has been the most error-prone part of the project. I've revised the tag taxonomy three times and I still find inconsistencies when I revisit older entries.

What I Keep Finding Myself Rejecting

Over time, the same patterns of rejection have repeated often enough to become recognizable.

There's a pattern I think of as the "math wrapper" — a generic puzzle, runner, or matching game with one math problem stuck on the splash screen of each level. Take the math out and it's the exact same game. The math isn't doing any pedagogical work. It's just there.

There's a pattern of relentless reward animations. Confetti explodes after every minor success. A coin animation plays for two seconds when the kid gets a single answer right. After ten minutes the celebration starts to overwhelm the actual game. The kid's brain learns to enjoy the celebrations, not the learning. I've come to think of this as the wrong dopamine circuit being trained.

There's the ad-gate pattern, where every level transition forces the kid to watch a video ad. Whatever educational value the game has is destroyed by the friction. Kids tune out. Parents quietly close the tab.

And there's a tone problem that's hard to articulate but easy to feel — a kind of cheerful condescension where the game treats the player as if they couldn't handle a real challenge. Bright voices, exclamation points, "Great job!" no matter what the kid actually did. Children develop a sense for this and start tuning it out, the same way adults learn to tune out marketing copy.

I reject games that hit any of these patterns, even when the underlying mechanic is good. There's no point including a game I wouldn't want a kid in my own life playing.

The Slow, Manual Reality

The infrastructure side of the site is intentionally simple. A flat manifest of games and their tags. A small evaluation log where I track which games I've reviewed, when, and what I decided. No CMS, no admin panel. Adding a new game is editing a file, taking a screenshot, deploying.

The reason for keeping it simple is the same reason iDogNames stayed manageable for me — anything that adds friction to the daily work of updating the site eventually causes the site to fall behind reality. A reference site is only useful if it's current, and it's only current if you actually update it. I'd rather have a slightly ugly editing workflow that I'll actually use than a beautiful CMS that I'll avoid.

Once a quarter or so I cycle through a sample of older games and re-check them. About one in twenty needs an update — a new ad behavior, a difficulty change, sometimes the game has been pulled offline entirely. The work isn't intellectually hard. It's just steady.

Where It Is Now

LumiGameLab has been running for a few months. Of the candidate games I've evaluated, roughly a third have made it onto the site. The rest are either broken, fail one of the patterns above, or are weaker duplicates of better games already included.

That's a smaller catalog than the big educational games portals. The smallness is the point. A parent or a teacher who lands on the site doesn't have to wade through hundreds of mediocre games to find the few good ones. The wading was already done.

The site is at lumigamelab.com. Free, no account, no aggressive ads. Just games that I think are actually doing something useful for the kid playing them.

Building a curated reference site is slower and less glamorous than building a single feature-rich product. Most of the work is invisible to the user. But that invisible work is exactly what makes the site useful — and the only thing that distinguishes it from the much larger sites that didn't bother.

This article was written with AI assistance.