I don't like the constant association between using LLMs and writing slop code. It's a narrative that follows me every time I mention AI in my work, and frankly, I'm tired of defending against it.
Every time I share something I built, the question inevitably comes up: "But does it work? Is it not just slop?" I understand the concern. We've all seen what happens when speed comes at the cost of maintainability. But I've learned to separate the legitimate questions from the noise.
Here's what I've observed: the people who wrote good code before still write good code. They just do it faster now. The quality of output correlates strongly with the quality of input and intent. If you cared about architecture before, you still care about it. If you understood why tests matter, you still write them. If you knew how to write readable code, you still do. The tools change, the standards don't have to. And if you didn't care about those things before, no tool will make you care now.
My Personal Shift
For me, it's been quite a shift. I still value well architected and tested code that is easy to read and reason about for the long term. That hasn't changed. But I've learned to treat that as a secondary concern when I'm exploring ideas. I can prototype, validate, and throw away. The safety net of clean code comes later, when I know what I'm building actually works for someone.
It's a struggle when people look at these contributions through the lens of "the code is not perfect" or "the code is not clean." They're prototypes. They're meant to be rough. The goal isn't perfection, it's learning.
This is a crucial distinction. I'm not abandoning quality. I'm changing when I apply it. There's a time for careful architecture and a time for rapid exploration, and these are different times. Pretending otherwise just slows down the work that matters.
The Endurance to Validate
I tend to have a lot of ideas. Like, a lot. And I have the endurance to push through rejections and "it won't work" to validate them myself. That's been true since before LLMs existed. What LLMs give me is a tool to shorten that path. I can try, have fun, and throw it away if it has no value or no impact for users. The cost of experimentation dropped dramatically, and that's a gift for someone like me who needs to explore many directions.
I think this is crucial in business too. When you need to explore directions you could go, you need a tool to measure the temperature, to gauge interest, and to push the boundaries of what is possible. Otherwise it is easy to invest a lot in something people will not pay for. LLMs let me build a prototype in hours instead of weeks, put it in front of real users, and get real feedback. The signal comes faster. The waste decreases.
Is It Slop?
Is it slop? I do not think so.
Initially it might feel like that. Not perfect design, missing features, rough edges everywhere. My Clawdbot breaks pretty much every day when updates come through. It is messy and imperfect and constantly evolving. But it is trending towards stability and a very good shape as people validate it and I learn what actually matters.
If Peter Steinberger had started trying to find the perfect architecture before launching, who knows if it would have ever happened. Sometimes done is better than perfect, and LLMs help me get to done faster so I can learn what perfect should actually look like.
The key difference between slop and a rough prototype is direction. Slop has no direction, no learning, no improvement. A rough prototype is a hypothesis being tested. I know which one I'm building.
How I Work
LLM is the best tool for a builder like me. It is how I work, and I have learned to ignore the dialogue about slop. Show me before publishing. Let users tell you what they actually need. Let reality correct your assumptions. That is the builder's way, and AI does not change that, it just makes the feedback loop tighter.
The slop narrative assumes that speed and quality are mutually exclusive. For me, they are not. Speed and certainty are what I traded, and that trade has paid off. I build more, I learn more, and I ship things that people actually use. That is what matters to me.