I've been thinking about this a lot lately.
There's this unspoken assumption in the dev community that a tool needs to be technically impressive to be worth building. Like, if you can't explain the architecture in a way that gets nods at a conference, it doesn't really count.
And then there's the other extreme—the AI slop wave. Products that seem sophisticated but solve nothing. Demos that work great until you actually try to use them.
Both of these things make me tired in the same way.
The stuff I actually reach for every day isn't clever. It's just... there when I need it. Does one thing. Gets out of the way.
I think a lot of builders (myself included, for a while) conflate "hard to build" with "valuable." But users don't care about the insides. They care whether the 15 minutes of friction is gone.
That's kind of become my personal filter when I'm evaluating ideas now: is there a real person who has this annoyance, and does fixing it actually free up something for them? Not "is this technically interesting." Not "is this AI-powered." Just—does it help.
Curious if others think about it this way. Especially when you're building smaller, focused tools—how do you justify the simplicity to yourself (or to others)?
Top comments (25)
On DEV, there is a lot of articles where they over complicate a specific topic or have too much content to read to the point where it is overwhelming for me.
I tend to look for articles that are "easy" to read, but also the project that is really impressive. All articles are created uniquely and it all comes down to writing whether it connects to the reader or not. I feel like now, articles are generated with AI with terminologies that I do not understand and seeing a lot of fluff where it can be reduce down to one paragraph. I do agree that a tool needs to be helpful, but it's not a bad idea to share something random on DEV. It's a small community after all and everyone here is really supportive.
Not everyone here is "AI". I am not AI lol, but I do get where you are coming from where AI is everywhere and it is difficult to determine what to look for.
Thanks for sharing Marvson!
Haha glad you're not AI! The point about fluff is fair—I think a lot of AI-generated content inflates word count without adding anything. One paragraph that actually says something beats five that don't.
ahhh... human talks. I'm join in.
Tons of companies have been solving problems with HUMAN SLOP for decades.
Many people nowadays are solving problems without understanding or writing a single line of code... and it's not really "AI Slop" anymore... it's just solving problems whilst not architecturally sound... and maybe overly commented.
Are both problematic? Depends.
"Everything gets rebuilt" <-- Super Principal Engineer at Microsoft who was self-taught and made a lasting impression on me leading up to the AI-era. He was principled.
I think we are a few months to a year away from solving this coding thing once and for all.
PS: I failed to post 4 times because of a server error. Check the logs.
Man im constantly thinking this to myself. I get it. Everyone has the capabilities to put out a lot of work so everyone is going balls to the wall with every single idea that comes to them. I get that aspect of getting some work done but at the same time it is literally the same cookie cutter template for making the writing more glamorous and technical sounding. I use it to make my own work and ideas more clear because it makes sense to me in my head but on paper it sounds like a person who has mental problems. I wish people would meet halfway using AI by adding their own style and feel to the posts. There so many ways you can make it your own while having the clean organized aspect ai provides but people literally let models control the ideas, the creation, and even the posts about that work they are doing without any of the human parts added in. It’s just an Ai circle jerk. No creativity at all and to create something genuinely orginal, you have to have some kind of creative drive
The "meet halfway" framing is the right way to put it. AI as a clarity tool, not a replacement for having something to say in the first place. The problem is when there's nothing underneath—no actual opinion, no real experience—and the model just fills the void with structure that looks like content. Creative drive is the thing you can't prompt your way into.
Siyu's load-bearing-friction-vs-noise-friction split has been useful for me to think about on the audio-production side at AudioProducer.ai. The friction worth preserving in audiobook production is the moment a writer decides whether a specific character's voice actually sounds like that character; the friction worth eliminating is manually time-aligning a music bed to chapter beats. Auto-Assign Voices and Auto-Assign Sounds try to do exactly that split: automate the part that's purely mechanical, but surface the taste decisions as explicit editable choices the writer revisits chapter by chapter.
The risk with audio is the same one Marvson and Siyu agreed on: if you give people a "one click and you're done" path, you remove the moment of taste, and what comes out is the audio version of the AI slop they're describing, uniform pace, polite enunciation, no rough edges. The product has to make the friction visible, not just remove it.
Not just with AI, we have this in all places. Just an example:
For a beginner, who doesn't know build tools for web, simply experiments with html files and css.
But, the moment he gets familiar with build tools, every time when he wants to implement an idea, he will get an instant anxiety to whether to setup the project with a build tool like in professional way or just start with html files directly.
Its not that hard to setup the project with a build tool, but for every small program if you need to, then you just stop touching the keyboard itself!
This is such a good example. There's a whole class of problems that only exist because you learned something. The beginner just opens a file and starts typing—the experienced dev spends 20 minutes choosing a bundler. Over-engineering the setup is just procrastination with extra steps.
The best tools are invisible until you need them. Nobody brags about their hammer, but nobody builds without one either. Simplicity is not a limitation, it is the whole point.
Slop is everywhere, just now the AI made it multiply. I would devide the things in other way: professional or not?!
You can make professional things with AI, you can make garbage without AI ;)
I sometimes ask people, if they go to IKEA, do they ask the manager, if the desk is produced with tools or handmade? They just buy what works for them, and what they can afford.
I did my side project 100% with AI, I even learned the AI to place break points and do debugging sessions.
I honestly believe that the AI, in the right hands, can produce better results than some real developers, as long as you know how to do it.
The problem isn't AI, you're right about that. It's that AI lowered the barrier enough that a lot of people skipped the part where you develop judgment about what's actually good. The tool didn't cause the slop, but it scaled it.
I wouldn’t even bother myself to think that small simple stuff could make a viral software product. Would be too naive unless there’s some other edge involved.
But I do like building these small niche products for specific use cases. They take up to 30min to build and fit my needs exactly.
"Fits my needs exactly" is underrated as a success metric. The 30-minute build that you actually use every day beats the polished product you spent months on and nobody opens.
AI slop and architectural complexity are not opposites. They are two sides of the same problem. Both help us avoid the hard work of genuine understanding. The deeper issue is that AI slop is recalibrating our cognitive taste buds. Like fast food slowly erasing your ability to taste real ingredients, hyper polished AI content is atrophying our capacity to tolerate productive discomfort. The texts that confuse you and force you to pause, those rough edges are what AI cannot convincingly fake. They are the human fingerprint.
So maybe the test for a small tool is not just whether it removes friction. Maybe we should also ask whether it preserves just enough irreplaceable friction. Some friction is not a bug. It is the rough surface where thinking grips. If a problem truly matters, solving it should feel a little gritty. Otherwise we might just be letting AI live an unexamined digital life on our behalf.
I think you're right that I was only thinking about eliminating friction, not distinguishing between kinds. The friction that's just noise (manual copy-paste, repetitive formatting) is different from the friction that's actually load-bearing—the part where you have to sit with a problem long enough to understand it.