The Game Discovery Crisis Nobody's Talking About

Hey everyone — MrPhil here.

Over the past week, I've been writing a series of posts that I didn't plan as a series. They started as separate observations — about AI replacing programmers, about $2/hour robot labor, about mass layoffs in gaming — and somewhere around the third one I realized they were all pointing at the same thing.

Not the thing everyone's talking about. Not "AI is going to take our jobs" or "AI is going to make amazing games." Those conversations are happening plenty. The thing I can't stop thinking about is what happens after the creation boom. The part nobody seems to be planning for.

Discovery.

The Thread

Let me pull this together.

In The Last Techno-Scribes, I wrote about how AI is doing to programming what the printing press did to hand-copying manuscripts. The barrier to making software — and games specifically — is collapsing. People who couldn't build games last year are building them this year. The scribes are being replaced by the printing press, and what comes next is a creative explosion we can barely imagine.

In $2 Robot Labor and Jevons' Paradox, I followed the economics. When the cost of doing something drops by 90%, people don't just do the same amount for less money. They do ten times more. That's Jevons' paradox — the counterintuitive truth that efficiency gains increase total consumption, not decrease it. Applied to games: cheaper development doesn't mean fewer games. It means an explosion of games the market has never seen.

In What Block's Layoffs Signal, I looked at the industry restructuring. One-third of US game workers laid off in two years. Playtika cutting 15% of its workforce explicitly to replace headcount with AI. Block cutting nearly half its employees and watching the stock jump 24%. This isn't temporary belt-tightening. It's structural. Small teams with AI are the new unit of production.

Each of those posts stands on its own. But the conclusion they all point to is the same, and it's the one nobody in the industry seems to be seriously grappling with.

When everyone can make a game, how does anyone find one?

The Math Nobody Wants to Do

Let's start with where we are. Steam released just under 20,000 games in 2025. That's already a record. The year before was a record too. And the year before that.

The trend line is clear, and it was clear before AI tools started seriously accelerating development. Now factor in what's happening:

• AI coding assistants can compress months of programming into days. I've watched it happen in my own work.
• Generative art tools are eliminating the asset creation bottleneck that killed more indie projects than anything else.
• AI audio and music tools are making professional-quality sound design accessible to solo developers.
• New AI-focused game dev tools emerge every week, lowering the floor further.

The GDC 2026 State of the Industry report found that a third of developers are already using generative AI in their work. That number will be higher next year. And the year after that, the question won't be "do you use AI?" but "how much of your game did AI build?"

I don't think it's unreasonable to project 50,000 games released on Steam within two to three years. 100,000 within five. Not because I'm being dramatic — because the math of cheaper production plus Jevons' paradox makes it almost inevitable.

Here's the thing: Steam's entire discovery infrastructure — the queues, the tags, the curator system, the recommendation algorithm — was designed for a world of 10,000 releases. It's already straining at 20,000. What happens at 100,000?

It doesn't strain. It breaks.

Why Storefronts Will Never Fix This

I want to address the optimistic response here, because I've heard it: "Steam will just improve their algorithm." Or "Epic will figure it out." Or "The platforms have every incentive to solve this."

They don't. They really don't. And the reason is structural, not technical.

Every storefront has a conflict of interest. Steam can only recommend Steam games. Epic can only recommend Epic games. The PlayStation Store can only recommend PlayStation games. No storefront will ever link you to a competitor's product, even if that competitor's product is a better match for your taste.

This seems obvious when you say it out loud. But think about what it means in practice. If you love Factorio and the best recommendation for you is a game that's only on GOG, Steam's algorithm will never tell you that. It'll recommend the closest thing it sells. Which might be great. Or it might be mediocre. You'll never know what you missed.

This isn't a bug in these systems. It's the business model. Storefronts make money when you buy games from them. They have zero incentive to build a platform-agnostic discovery layer, because that layer would sometimes point customers away from their store.

A former Valve developer said publicly last year — covered by Polygon — that game discovery on Steam is "broken." And this is from someone who built Steam's discovery features. If the people who designed the system think it's failing, and the economics guarantee it'll get worse, waiting for storefronts to fix it is not a strategy.

The game discovery problem is a platform-agnostic problem being served exclusively by platform-locked solutions. That's the structural failure. And it's about to collide with the biggest supply increase in gaming history.

The Genre Lie

There's a second structural failure underneath the first one, and it's so deeply embedded in how we think about games that most people don't even notice it.

Genre is a terrible way to describe a game.

"Strategy" contains both Europa Universalis IV and Into the Breach. One is a thousand-hour grand campaign across centuries of real history with pausable real-time mechanics, a diplomacy layer deeper than most games' entire design, and a learning curve measured in weeks. The other is a tight, puzzle-like tactics game you can play in thirty minutes with a clean undo button and pixel art.

Both are "strategy games." This label tells you almost nothing useful.

"Action RPG" contains both Dark Souls and Diablo IV. One is a methodical, punishing game about patience and spatial awareness where every enemy can kill you. The other is a fast-paced loot fountain where you mow down hundreds of demons per hour. Recommending one to a fan of the other based on a shared genre label is actively unhelpful.

Genre was designed for shelf organization. Physical stores needed categories so employees could sort boxes. "Sports" goes here, "Racing" goes there. It was a logistics tool masquerading as a description. And we've been stuck with it ever since.

Steam improved on this with tags — games can have dozens of community-applied labels. That's better than genre, but it's still a flat list. "Open World, Survival, Crafting, Base Building, Multiplayer, Early Access" tells you more than "Action" but it still can't capture how a game feels. Is it cozy or tense? Meditative or frantic? Does it respect your time or demand 200 hours? These are the things that determine whether you love a game or bounce off it in 20 minutes, and no tagging system can express them.

The taxonomy we use to describe games is fundamentally inadequate for recommendation. And we're about to need recommendations to be much better than they've ever been.

What If Games Had DNA?

This is the question that won't leave me alone. Not "how do we fix genre tags" but "what would it look like if we described games the way they actually feel to play?"

Not a single label. Not a flat list of tags. A multidimensional fingerprint — a DNA — that captures the essence of what makes a game the specific experience that it is.

I've been building this at GameLegend.com. Here's the model: every game gets mapped across nine dimensions, each with multiple attributes.

Core Mechanics — what you actually do. Twenty-one categories, because games are complicated. Combat, building, exploration, puzzle-solving, automation design, territory control, deck building, diplomacy.

Feel/Pacing — the moment-to-moment experience. Is it meditative? Frantic? Methodical? Explosive? Cozy? This is the dimension that genre labels completely fail to capture, and it's arguably the most important factor in whether you'll love a game.

Progression — how you advance. Roguelike loops, open world exploration, linear story, skill mastery, collection. A Dark Souls fan and a Hades fan both like hard games with combat, but they want completely different progression models.

Social Mode — how you play with others. Solo contemplative, co-op chaos, competitive ranked, async community. A game's social mode shapes the entire experience and has almost nothing to do with its genre.

Aesthetic — visual identity. Pixel art, photorealistic, hand-drawn, voxel, minimalist. Not just a surface detail — aesthetic shapes mood and expectation.

Themes — what the game is about. Existential dread, wholesome friendship, political intrigue, cosmic horror. Two games with identical mechanics can feel completely different based on thematic framing.

Complexity — from pick-up-and-play to spreadsheet territory. This isn't a value judgment — some people want the spreadsheet. But recommending Dwarf Fortress to someone who asked for something "like Stardew Valley" because they're both "simulation games" is a failure.

Session Length — five-minute runs to multi-hour deep dives. A parent with 30 minutes after the kids go to bed has different needs than a college student with an empty weekend.

Strategic Scope — for strategy games specifically: map scale, economic depth, warfare emphasis, emergent narrative.

When you describe a game this way, you get something powerful: a real basis for comparison. Not "these are both strategy games" but "these both have methodical pacing, high complexity, deep economic systems, and multi-hour sessions — but one emphasizes political intrigue while the other focuses on warfare." That's a recommendation you can actually use.

More importantly, it enables a kind of search that doesn't exist yet: "Show me something that feels like Factorio but with a narrative." Or "I loved Hades — what else has fast-paced combat with roguelike progression but a hand-drawn aesthetic?" Try expressing that with genre tags. You can't.

The Crowd Makes the Map

Here's where I have to be honest about the limitation. Right now, the DNA profiles on GameLegend are mostly my own assessments. That's fine for a proof of concept. It's also the weakest part of the system.

My perspective is one perspective. I think Factorio is "meditative." You might call it "methodical." A speedrunner might say "frantic." They're all right. A game's DNA isn't a fact — it's a consensus.

This is why GameLegend has a reputation ladder. You don't get to shape the catalog on day one. You browse, you search, you rate games. After you've rated enough, you can vote on DNA tags — agree, disagree, suggest alternatives. If your votes consistently align with the emerging consensus, you earn more trust. Eventually you can submit new games, create curated lists, bulk-tag DNA profiles.

The key principles:

• Quality is protected by consensus. Multiple users vote on every dimension. Scores are aggregated and weighted by reputation. No single user can distort the data.
• Submission is high-trust. Only experienced curators can add new games. This prevents the catalog from being flooded with junk — ironic, given that the whole point is solving a flooding problem.
• Reputation is earned, not gamed. No leaderboards, no point chasing. Capabilities unlock quietly based on contribution quality.

The goal is a living map of the gaming landscape, shaped by the people who actually play these games. Not a critic's assessment. Not an algorithm's guess. A collective understanding of what makes each game tick.

Who Captures the Value?

Here's the economic argument that I think makes discovery the most important unsolved problem in gaming.

Throughout history, when a resource becomes abundant, the bottleneck shifts — and whoever controls the new bottleneck captures the value. This isn't theory. It's pattern.

Music. When recording technology made music reproduction nearly free, the scarce resource shifted from pressing vinyl to getting heard. Spotify, Apple Music, and YouTube — the discovery and distribution layers — now capture more value than most artists. The labels that survived did so by becoming marketing machines, not production houses.

Publishing. When self-publishing made it trivially easy to release a book, the scarce resource shifted from printing to finding readers. Amazon's recommendation algorithm and Kindle Unlimited became the kingmakers. Being published stopped being impressive. Being discovered became everything.

Apps. When mobile development tools made it cheap to build an app, the App Store and Google Play became the bottlenecks. They took 30% of every transaction — not because they built your app, but because they controlled who could find it. Discovery was the tollbooth.

Search. Google's entire $1.7 trillion market cap is essentially a discovery company. They don't create content. They help you find it. When the supply of web content exploded, the entity that helped you navigate the explosion captured almost incomprehensible value.

Games are about to follow the same pattern. AI is making creation cheap. Jevons' paradox guarantees supply will explode. The creation side of the industry — engines, tools, asset pipelines — is increasingly commoditized. The discovery side is wide open.

Whoever builds the system that helps players find the right game in a world of 100,000 annual releases isn't solving a nice-to-have problem. They're positioning themselves at the bottleneck. And bottlenecks capture value.

Why I'm Building This

I could pitch GameLegend as a pure business opportunity. The market analysis supports it. But that's not really why I'm doing it.

I've been making and playing games for nearly three decades. I've shipped games at Ludum Dares on deadline adrenaline. I'm building a 4X strategy game right now that's consumed the last year of my life. I maintain directories of indie game dev tools because I care about this ecosystem.

And I keep running into the same problem from both sides. As a developer, I know how hard it is to get your game in front of people who'd love it. The marketing treadmill — wishlists, Steam Next Fest, social media, influencer outreach — has almost nothing to do with making a good game. Some of the best games I've ever played had fewer than 100 reviews on Steam. Not because they weren't good. Because nobody could find them.

As a player, I know how frustrating it is to wade through thousands of titles looking for the one that fits. I don't want "popular strategy games." I want "a methodical empire builder with deep economic systems that I can play in 90-minute sessions." That search doesn't exist anywhere. Not on Steam, not on Epic, not on any review site.

GameLegend is live now in open beta with a handful of strategy games — enough to see the concept work. Enough to see how games that share genre labels have wildly different DNA, and games from different genres share surprising similarities. You can see it, search it, and start telling me where I'm wrong about a game's profile at gamelegend.com.

The Clock Is Ticking

I don't know exactly when the flood hits. Maybe it's already started — 20,000 games a year is a lot, and most of them vanish without a trace. Maybe it's two years out, when AI tools mature enough that non-programmers are shipping polished games. Maybe it's five years out, when $2/hour robot labor makes AAA production values available to bedroom developers.

But whenever it arrives, the infrastructure to handle it needs to exist before it hits. You don't build a levee during the flood. And right now, the discovery infrastructure for games is a series of genre tags and platform-locked algorithms designed for a world that's about to stop existing.

The creation tools are getting all the attention. Every week there's a new AI art generator, a new code assistant, a new game engine feature. Those are important. But the harder problem — the one that determines whether the creative explosion actually reaches the people it's meant for — is on the other side.

Not how to make games. How to find them.

That's the crisis nobody's talking about. And I think it's the biggest opportunity in gaming.

— MrPhil