One of the clearest tech stories this week did not come from a model benchmark, a headset demo, or another AI assistant trying to become your operating system. It came from a much uglier metric: the App Store is filling up faster than people are finding reasons to care.
On July 20, 2026, new reporting built on Sensor Tower data showed that Apple added about 560,000 new App Store releases in the first half of 2026. For comparison, the full year of 2025 landed around 600,000. That is basically a full year's worth of software arriving in six months. Sounds like a boom until you look at the other side of the chart: downloads in the first half of the year only grew about 2 percent, to 17.6 billion.
That gap is the whole story. Supply is exploding. Demand is barely twitching. Which means what we are watching is not some magical new age of software abundance. We are watching AI turn app creation into a volume game, and volume games always mutate into ranking games.
Vibecoding did not just democratize app creation. It industrialized app submission.
As soon as I saw those numbers, the feeling was familiar. Not from mobile. From SEO. Specifically, from every era when publishing got cheaper, distribution stayed scarce, and the market flooded with pages, wrappers, clones, and low-effort variations designed less to be loved than to be surfaced.
That is why this matters beyond Apple. The App Store is starting to look like search results used to look when the fastest operators realized the real product was not the page. It was the slot.
This is not an app boom. It is a packaging boom.
There are absolutely real developers building useful things with AI. Some of them are fast, smart, and worth paying attention to. A software engineer can now go from idea to prototype in a weekend instead of a quarter. That part is real.
But the market consequence of that shift is not automatically better software. It is way more software-shaped objects entering the store at a speed the old filters were never designed for.
When the cost of building drops this hard, the submission bar changes. Weird little experiments get shipped. Thin wrappers get shipped. Niche one-feature utilities get shipped. Clone concepts get shipped. Joke apps get shipped. Disposable trend apps get shipped. Things that would have died as half-finished side projects last year now make it all the way to the marketplace because the distance from idea to icon got compressed.
That is why calling this a revival misses the point. Revivals show up as renewed user demand, stronger retention, and better products finding real audiences. This looks more like content inflation. More units. More noise. More competition for the same handful of discovery surfaces.
I wrote earlier this morning about YouTube finally treating synthetic sameness like a spam problem in YouTube Is Treating AI Slop Like Spam. Same pattern here. The platform issue is not "AI versus human." It is cheap throughput versus scarce attention.
Discovery breaks before quality does
People love arguing about whether AI-made apps are good or bad. That is not actually the first failure mode. Discovery is.
If app supply doubles while downloads barely move, the average app does not become twice as valuable. It becomes twice as invisible. That means incentives slide toward screenshots, titles, trend-hopping, keyword stuffing, cloned positioning, and whatever else might squeeze a little more visibility out of ranking systems normal users never see.
Again: this is old SEO logic wearing mobile clothes.
When distribution is scarce, every marketplace quietly teaches people the same lesson. Do not just build something good. Build something that looks legible to the system. That is how you get an explosion of derivative naming, lookalike onboarding, recycled subscription hooks, and the same product concept wearing ten slightly different outfits.
The app is not always the thing anymore. Sometimes the app store listing is the thing. The screenshots are the thing. The ranking exploit is the thing. The trend timing is the thing.
And once that starts, the marketplace stops feeling like software and starts feeling like search spam with prettier gradients.
The moderation problem is already here
This would still be mostly annoying if the downside were just clutter. It is not. The trust layer is wobbling too.
On July 17, 2026, San Francisco ordered Apple and Google to remove a batch of AI "nudify" apps from their stores. That matters because it turns the abstract fear into a concrete question: if the stores are this full and review throughput is this high, how many exploitative, deceptive, or low-grade abusive products are cruising through on the same pipeline as the harmless junk?
Apple told reporters it still reviews 90 percent of submissions within 48 hours. Fine. Maybe that is true. The problem is not whether the queue moves. The problem is whether the queue still means what people think it means when submission volume accelerates this fast.
A review process can be fast. A review process can even be technically rigorous in spots. But if the surrounding market has been transformed by AI-assisted abundance, then the real challenge is not just malware detection. It is pattern recognition at scale: what is useful, what is derivative, what is exploitative, what is misleading, what is just a wrapper around somebody else's commodity feature, and what is likely to poison trust if it keeps multiplying.
That is a much harder marketplace problem than approving binaries.
Vibecoding changed what "finished" means
The deeper shift here is cultural. Vibecoding does not just make developers faster. It changes the threshold for what counts as done enough to publish.
That is a big deal. The old software instinct was to polish, test, refine, and ship when the thing could stand on its own. The new temptation is to ship when it is good enough to enter the feed and let the market sort it out later. That sounds agile until you realize it also sounds exactly like content production logic.
Apps are increasingly being treated like posts, memes, or merch drops. Something to launch, watch, remix, abandon, relaunch, and package again if the first angle does not hit. Some of that is healthy experimentation. Some of it is creative chaos. But a lot of it is just the economics of cheap generation pushing software toward the creator economy's worst habits.
That is why this story feels bigger than Apple. It points at a broader trend I have been circling in pieces like Typing Into Apps Is Starting to Look Old: software is becoming easier to produce, easier to reshape, and easier to treat as an interface layer instead of a finished object. The good version of that trend is powerful. The bad version is a flood of interchangeable wrappers pretending to be products.
The winners will be curators, not generators
People keep framing AI distribution fights like they are about raw making power. I think that is backward. Making power is getting cheap. Selection power is getting valuable.
The real moat for a marketplace in this environment is not "can more people publish?" They obviously can. The real question is whether the platform can help users find durable, credible, differentiated software without turning the front door into a landfill.
That means stronger curation signals. Better provenance. Better reputation systems. Better editorial judgment. Better bundling around trust. Maybe even harsher friction in categories that are obviously getting churned by low-effort cloning. In other words: the part Silicon Valley usually hates, because it looks less like infinite scale and more like gatekeeping with taste.
But some form of that gatekeeping is exactly what abundance markets eventually rediscover. When creation gets too cheap, filtration becomes the premium product.
- Users do not need twice as many apps. They need faster ways to know which ones are real, safe, and worth keeping.
- Developers do not just need better tools. They need a marketplace where quality can still separate from noise.
- Platforms do not mainly need more submissions. They need credibility in how submissions are ranked, reviewed, and surfaced.
- AI lowers the cost of making software. It does not lower the cost of trust.
Phones are getting their own slop layer
That last sentence is the one I keep coming back to. AI collapsed the cost of generating code, UI, copy, and packaging. What it did not collapse is the cost of being worth somebody's home screen.
That is where the current wave starts looking less like progress and more like oversupply with nice branding. The store gets fuller. The average listing gets thinner. Discovery gets gamier. Review gets harder. Trust gets shakier. Meanwhile everybody keeps pretending the sheer number of new releases is proof of vitality.
It is not. Sometimes more output means more life. Sometimes it just means the filters are losing.
I do not think the answer is to sneer at every vibecoded app. That would be lazy. Some genuinely good software is going to come out of this era. Probably a lot of it. But the current metrics should kill the fantasy that more AI-assisted creation automatically means a healthier ecosystem.
What it means, at least right now, is that phones are inheriting the same problem the web already knows too well: when publishing gets frictionless, the first thing to scale is not excellence. It is clutter.
And once clutter becomes a business model, somebody has to decide whether the marketplace exists to surface software or merely to warehouse uploads.
Apple is getting very close to having to answer that question in public.