Software needs that Steve Jobs touch because new Software feels average now

Agentic coding models and harnesses have become so effective that a large share of new software is now written by agents. For a good reason, it’s faster, code is at times higher quality than if human written, and there are seemingly no downsides. We thought.

Frontend

But the majority of new Software feels average, in a sense that it feels a like AI slop, instead of software that was created with craft and care. For all that follows here, it’s important to understand, that AI coding produces impeccable results, if done right. But it all comes down to the process. Years ago we used to start at an idea, sketch it out by hand or on a whiteboard, then refine the user experience and exact optics in Figma, then start development. We used to style every button to perfection, manually set border radii to be concentric and look correctly. Things took time and while something taking longer doesn’t mean that it is better, it does usually mean that more thought has had to be put into.

Today, many teams jump directly from idea to their coding agent (Claude Code, Codex). This quickly gets them an MVP, but not something that excites users. There’s this well known saying that getting the last 20% right takes 80% of the work, and here I would say that getting the last 20% right takes 95% of the work. And it’s these 20% that make the difference between AI slop, and a thought out solution.

Describing what “AI slop” in product development is, is so hard if you try to nail this down on certain things in the outcome. It’s not single specific things (although there are giveaways), you simply … feel it. You can feel when something doesn’t have craft and thought in it. This exists in backend code, frontend code, image generation, companies and also text. Describing AI slop in frontend is hard, with text it is easy, because everyone can imagine a text that seems to hold no value, simply be a word salad without soul. And the same applies to UI, although many people don’t “see” good/bad UI (there have been many web-app redesigns that I have been hyped about and others asked me, what’s different), but people do feel good UI subconsciously. So when I say “slop” in product development, what I mean is creating something that ticks the boxes, but doesn’t tick them with care. And AI make it really easy to create exactly that, which we then call “AI slop”. And idea, craft and effort will always shine out.

Here’s a video recording of an animation in Apple’s music app. I know for a fact that this has been implemented with AI. It’s hard to describe, but again, you feel it.

Backend

The same thing repeats itself here. It’s easy to create a new SAAS platform. But to create something where the user asks himself wtf how is this possible, that’s hard.

New Software feels average now. People ask ChatGPT what feature to add, they tell Claude to implement it, and we get software that simply doesn’t spark delight, or produce Steve Jobs’ wow effect. This is so hard to describe, because often even when on paper everything is working great, Software looses a certain touch. Because implementation can be done so quickly, that developers simply don’t test or feel what they build and it takes a very strong vision to be able to use Coding Agents to implement things.

It takes people who understand the craft and push the limits of what is possible with these coding agents. It takes people who have a vision so strong that they are willing to go through extra iterations until they are really exactly there, at their vision. If you’ve watched The Playlist, this reminds me so much of that scene where Daniel Ek repeatedly asks Spotify’s engineers to re-iterate on their playback delay until it’s near instant.

Apple

When all of Software becomes AI slop, we end up with nothing being pleasant to use. And I don’t want to task in a negative way at all about one company in particular, especially because the issues that I am describing here equally (if not even more) apply to competitors like Microsoft. In a lot of areas though, Apple’s software was known for their polish. For their engineers to care about the implementation and for the result to feel magical. And this no longer is the case.

When people say “AI is the downfall of Apple” they mean that that’s because Apple can’t, or chooses not to, come up with a model of their own. But the bigger problem is that (at least from outside it seems as if) all their engineers seem to use AI without proper training, and without caring (and I don’t know what happened to QA).

Every single software update they bring doubles down on my feeling that there is no heart, no effort put into. It feels like AI slop. And I have lost trust their process.

I want to feel that spark again, I want to see them launch something and think “wow, that’s how it’s done”. They need this Steve Jobs effect here, that bullet proof piece of magic that just works.

And ironically, the company where every single product release – I don’t mean models here, I mean products (is the model the product?) – produces exactly that spark, is OpenAI. OpenAI is one of the very few companies who’s products simply are not AI slop.

Siri needs that Steve Jobs touch

The future of computing looks fundamentally different from how we compute right now. And companies, including Apple, are racing to build that future. To play a role in the future of software, you need to own either

1. the model
2. the product

There is a case to be made that the model is the product, but, apart from the fact that I disagree with this stance, for the sake of simplification, I am going to refer to the “model” as the foundational LLM, and the “product” the thing that wraps the model, be it a chat website, coding agent, classic OS, etc. Notice how product doesn’t mean hardware in this context.

As of now, Apple either couldn’t manage, or decided not to, participate in model wars in the sense that they are racing to create the most powerful foundational Large Language Model. They might be creating Small Language Models (specialized models for specific, generally lightweight tasks), but seem to have decided to integrate other companies’ models in their software instead. In its current state this isn’t necessarily a problem, but in case they decide not to own the model, they need to own the product, aka, the integration.

And they don’t.

Apple doesn’t work the integration side either. When they launched Apple Intelligence in 2024, we thought they finally did. Well, that’s almost 2 years ago, and all they shipped by now is the worst ChatGPT wrapper I’ve seen. And I am serious when I say that. Nothing works, and some of the mistakes are fundamental things that you think off when creating an LLM wrapper. And nobody at Apple can possibly have tested these, because if so, they wouldn’t have launched it. And nobody at Apple can possible be using these, because if so, they have fixed it.

These are some of the violated principles.

1. Give the model the context it needs to know what it can and can’t do. ChatGPT in Siri can’t to tool calls (“do things”), it can only answer. ChatGPT in Siri regularly asks me if I want something being put in my calendar, but it can’t actually do that.
2. You re-try on error, don’t just output an error message.
3. You create meaningful features, don’t ad AI for the sake of it. Take Image Playground for example. No one is using that because 1. it doesn’t do anything useful and 2. doesn’t work well enough for useful things. Genmoji doesn’t work once.
4. You add Markdown and Latex support for the outputs. For long we simply got Latex not rendered in ChatGPT.
5. You create smart experiences instead of sprinkling AI gradients on top of the old system.

I don’t care if they use their own model or Google’s as a foundation, but the integration needs to feel magical, like a personal assistant that understands what you want to do. It needs to be frictionless and easy in the sense that the user doesn’t even know they are using an LLM/AI, they just use their software and it feels smart. Apple does the opposite. They added ugly orange gradients that scream AI all over the place.

Apple was the one company that I thought would get it right and they had me when they said they wouldn’t ship a chatbot. They were the one company that I thought wouldn’t ship slop. But I was so wrong, sadly.

And yes, AI is hard to get right. Gemini in Android currently (november 2025) doesn’t get it right either, but they own the model, the product and the ecosystem, and I have confidence in their structure and them doing exceptional engineering work.

Look at that ad, this is what is great, and actually possible now with our current tech. Back in 2024 this would have been noble, in 2025 personal context and tool access is normal. They have to be able to do that after WWDC26.