BLOG: The Mediocrity Epidemic

Hello Everyone!

AI did not make software worse. It made bad software faster to make.

Before this era, building an application had a natural filter built into it. You needed weeks to write a working backend, weeks more to make the frontend not look broken, and somewhere in that grind, most people who did not actually understand the problem they were solving simply gave up. The friction was doing quality control for free. Vibe coding removed the friction. It did not remove the requirement for understanding, it just stopped checking for it. So now anyone with an idea and a Sunday afternoon can prompt their way to a login screen, a dashboard, a payment button, and call it a product. It runs. And it is almost always hollow, not because the code is bad, but because a single paragraph describing the problem to an AI became the entire depth of the founder's thinking on it.

I watch this happen from both sides. Building Polipics, I am one of the people using these tools to move at a speed that was not possible two years ago. And running StudioD, I watch client after client come to us frustrated after trying five different mediocre tools that all promised to solve their marketing problem and none of which actually understood marketing. Same industry, same pattern, everywhere.

There is a version of ambition that looks like ambition but is actually avoidance. Launching a new app every week feels like output. What it actually is, most of the time, is a refusal to sit inside one hard problem long enough for it to embarrass you.

Going vertically deep means eventually hitting the part of the problem with no easy answer, the edge case that breaks your clean model, the thing your competitors also missed, which is exactly why it is still available to solve. That is where real products get built, and it is slow, uncomfortable, and produces nothing you can post on a Tuesday. Horizontal expansion never touches any of that. Vibe coding did not create the instinct to avoid depth. It just removed the last obstacle that used to force people into it.

Then there is YouTube. "I built an entire app in one prompt" is not a claim about software, it is a claim about attention, and it works because it flatters the viewer with the idea that building something valuable is now nearly free. What actually became free is producing something that merely resembles a product. I do not think most of these creators are lying. What they have gotten good at is not development, it is distribution: the algorithm, the thumbnail, the fifteen seconds before someone scrolls away. That is a real skill. But distribution and development are not the same discipline, and no one has found a way to hack the second one. You can compress the time it takes to look like you built something. You cannot compress the time it takes to actually understand the problem you built it for.

The founders building shallow apps lose time and money, but they can absorb that and try again. The people who cannot absorb it are the end users.

Somewhere in this flood, a handful of genuinely good applications exist, built by someone who went deep enough to find the ugly edge cases and solve them anyway. That application is now competing for attention against fifty superficial clones with louder launch videos, and the user cannot tell the difference from the landing page. So they try one. It half works. They try another, equally frustrating in a different way. By the third attempt they have stopped evaluating products and started forming a belief: that this entire category overpromises and underdelivers. That belief does not stay contained to the apps that earned it. It spreads to the good ones too, because trust, once it breaks in a category, does not break selectively.

Going deep looks unglamorous from the outside. Three weeks on the part of the problem with no visible output, because getting it wrong quietly breaks everything built on top of it later. Saying no to five adjacent features because none of them are the actual problem, and yes to fixing the one thing that is broken but boring. None of that shows up in a one-minute video. All of it shows up, eventually, in whether the thing still works six months later.

AI did not lower the ceiling on what is possible to build. It raised it, for the people willing to do the hard part themselves. What it lowered is the floor, and right now almost everyone is building at the floor.

The opportunity, if you actually want to take it, is exactly where everyone else is refusing to go. Down, into the one problem. Not sideways, into the next ten.

Thank you!