Why AI writes software but doesn't build a good product
If you believe Death by Claude, Trello, Notion, and other tools are "already dead." When I read that and realized that many people are declaring SaaS dead for reasons like these, I thought: Why should I prefer an ugly but functional solution that Claude Code or Claude Design can build for me in minutes to a tool that's been developed over years and optimized in terms of UI and UX?
Whether we'll still be using tools like Trello or Notion in 10 years depends heavily on how good AI becomes in the area that's currently still in its infancy: user interface and user experience. Sure, Claude can already write clean code and generate real software within a few days. But would anyone actually want to use it? Frello is the best example: I've invested most of my time in precisely the areas where AI is practically useless. User interface and security.
Colors, typography, and the combination of both are the least of it, and simultaneously the first area where AI is naturally very weak. The heart of Frello, the task workflows, team communication, the detailed view of a task, AI could never have mastered. Never. Perhaps that will change someday, but currently: no chance. In this respect, Claude is like a friend or colleague whose opinion you ask, but on whose basis you would never make a fundamental decision.
There are certainly plenty of examples when it comes to security. Claude almost replaced my entire policy at Supabase once, making sensitive data available to all users—data that is normally restricted to certain groups. But that's another story.
You surely understand the point I'm making. Will AI ever be able to write software that looks or functions really well? I'm not so sure. But we should probably ask ourselves: Do we want innovation in the future, or mass-produced products that all feel and function the same? Wouldn't stagnation be something that just doesn't suit us humans at all?
Hmmm.
I've been experimenting with something a bit weird lately where two LLMs basically debate the strategy decisions for a small web project and the human mostly executes.
What surprised me is that the models are actually decent at challenging each other when you force them into disagreement mode. One will push for scaling content faster, the other will argue the indexing isn't validated yet. One optimises for distribution, the other for monetisation alignment, etc. Sometimes the debate is genuinely useful.
But the failure mode is exactly what this thread is talking about: they can generate internally coherent reasoning that still drifts away from reality if nobody grounds it. Especially around product judgment or user psychology.
The interesting part is that for certain kinds of products, especially simpler content/utilities stuff, the gap matters less than I expected. Bad UX decisions on a banking app are fatal. Bad decisions on a small calculator site are usually recoverable.
So I'm starting to think the "AI can't build products" argument is very category dependent. There are probably whole classes of internet businesses where strategic taste matters more than engineering depth, and LLMs are already weirdly capable there.
Yes, I see it the same way: If I want to build a simple to-do app, AI is obviously great at that. I would say that when it comes to user experience, it's very difficult to train LLMs. This is simply because, while there are many books and other content that LLMs can read and analyze, when it comes to thinking and weighing options in a specific situation, AI becomes very ambivalent. At some point, I get the feeling that the AI just wants you to be happy and makes less sensible decisions. If you do the preliminary work and only give the AI a blueprint, AI is incredibly good at quickly delivering good MVPs. If I remember correctly, Google recognized this a long time ago and, for example, always emphasized that AI is a tool to become more efficient and to take over tedious tasks, but not to replace specific roles in product development.
That matches what I've been seeing too. AI is surprisingly good once the direction is constrained, but the actual "which tradeoff matters here?" part still feels very human.
Especially with UX where two reasonable choices can both work technically, but one just feels wrong once a real person touches it.
Yes, that's a good summary. How something feels is something that AI simply cannot currently do (and perhaps never will have to).
Exactly. I think AI gets underestimated on execution and overestimated on judgment at the same time.
Once the direction is clear it can move insanely fast, but defining what good actually means in a messy real-world context still feels much harder to outsource.
Really interesting perspective and largely agree.
I built S-BIZ (inava.app) using Claude Code and the experience taught me exactly what you're describing. Claude is an extraordinary builder but it has no opinion on what to build or why. The architecture, the logic, the rules engine that watches every task continuously — that all came from years of running an engineering company and feeling the pain of zero visibility firsthand.
And here's something people don't talk about enough — you have to be in the driver's seat every single step of the way. The moment you let Claude make a fundamental decision you don't fully understand, you've lost control of your own product. It will confidently go in the wrong direction and take you with it. I reviewed, challenged and redirected constantly. The speed is remarkable but the judgment has to be yours.
The UX point is real too. Getting S-BIZ to feel right — the colour coding, the way alerts surface, the psychology of making people feel accountable rather than surveilled — that required constant human judgment. Claude would build exactly what you asked and nothing more. The craft is still human.
On security — completely agree. You have to know what you're doing. Claude will confidently build you something that works perfectly and exposes everything if you're not watching carefully. I caught several moments where a wrong turn could have been catastrophic.
But here's where I'd push back slightly. The question isn't whether AI can replace product thinking — it can't. The question is whether a founder with deep domain expertise and Claude Code can now build something in months that previously required a team and years.
I think the answer is yes. And that changes everything about who gets to build software.
S-BIZ exists because I had a problem nobody was solving well enough. Claude Code let me build the solution. The insight, the architecture, the judgment — that was mine every step of the way.
That's a new kind of founder story — and I don't think it leads to mass produced sameness. It leads to more niche, more specific, more painfully accurate solutions to real problems.
I couldn't agree more. Honestly, the question has always been whether to prioritize quality over quantity. I, for example, value consistency in every button and view – and I don't neglect that. AI can deliver good results if I know exactly what the AI is doing. I fundamentally agree with you, but we inevitably have to think about how we want to interact with software in the future and what software should be capable of. For example, we're now seeing a trend where text written based on certain parameters can be identified as AI-generated. It's becoming noticeably more impersonal and increasingly overly precise. For me, it's much more about the essence of good software, which has always been characterized by prioritizing character over arbitrariness.
I thought about the precise nature of AI generated text. After working with AI for sometime now you kind of get used to it. I think if ai is conveying information you want it is alright as you can get more of your ideas out there. Only caveat is you got to proof read and approve what is being sent out and you got to own it. The question is does this make us more dumb as we don't have to think every word or smarter as we have to read more and digest what is being written. What do you think? This is an interesting topic.
Good point. I think it makes discussions in our society even more difficult because AI filters out so much, and we might even forget how to talk to each other. It's already difficult enough without AI. Unfortunately, I'm not at all optimistic about this and would actually advocate for bans in certain sectors. AI is always about efficiency, but very little about what's much more important and is more of a philosophical question: How do we want to consume media in the future? What should our working world look like? What specific problems will AI solve? And how much technological progress can humanity even handle?