Here’s my question:
Where is all of the AI coded software?
If coding is a nearly solved problem, if AI is making it so easy to make anything, why aren't we seeing the market flood with cheap alternatives to software that we all use daily and hate?
Microsoft office, anything from Adobe, Jira
Make your own list. Every designer I’ve ever met hates Adobe and their subscription model. My company spends a hojillion dollars a year on office365 and everyone hates it. Why aren’t we seeing alternatives popping up?
Should we be expecting this as a part of the whole “democratization” promise of AI?
As an addendum: I picked office and adobe just because they are two things we always whine about in tech, I realize they are platforms/ecosystems now, maybe it would be better to say Word and Photoshop, etc. but substitute in any commercial software you want. If AI coding can take over the internet application space can’t it also take over the software application space?
Because I see AI coded software everywhere: HN has been complaining about the recent influx of Show HNs due to AI. Take a specific product category: there are about two new dictation apps on Show HN every week.
Here are my theories:
- adobe/office are simply too complex for AI to replicate (as of now). Theo Browne has been testing all the frontier models with a relatively simple fish game. The results have been subpar, at best. I don't think anyone would prefer the AI-generated games over the original. Even if the AI-generated version was free.
- Most AI-coded software won't be distributed like classic software. Instead of a bag of hundreds of features (adobe/office), AI will be used to create the specific 5% of the features a single user really needs (home-cooked software). The software may even be per project or even per task. Why spend 20X to 100X the tokens/time when you only need a small subset of the features?
- AI may not be able to replace human "taste" or design. There is a lot of design that went into adobe and office. I don't think AI can fully understand all the "why's" of these designs. We can already kind of see this with open source: Gimp and LibreOffice are alternatives to adobe/office. Yet most people prefer the paid options.
- Even if AI solved coding, there are other barriers like marketing.
Some examples:
- I built my perfect window management system with the help of AI. It suits my very specific needs very well, but it's nowhere near ready for distribution. So this is AI-generated software no one else will ever probably see.
- Going with the home-cooked meal analogy: there is a dish served at all Chinese restaurants in Korea: jjajangmyeon. However you can't find it in China because that dish is only found at home, not at restaurants in China.
If you are open to the idea that those premises are incorrect, you have your answer. Coding is helped, not solved. AI makes it easy to prototype, but not easy to launch a product with production-grade code, support, sales, marketing, ops, scalability, etc.
With that change in perspective, you'd see exactly what we are seeing - scads of prototype-level product attempts that don't succeed when launched into a market.
AI is very optimized to producing code, but not there (yet? ever?) in making products. Show me a serious software made entirely with AI. There's no AI SolidWorks, AutoCAD, MacOS.
And there lies a contradiction: AI is making it easy to start and produce copious amounts of code that looks okay-ish. We start releasing a product and notice subtle changes. We did not internalize our understanding through it. There's a million lines of smart looking code, so we don't know where's what. We ask to AI to "make the button blue" https://opusfived.dev/ and the cycle begins.
I am working on a large software at work, trying a "I didn't even look at the code" approach with GPT 5.6 Sol. It started really promising but by now (and about $10k in tokens) it's a complete clusterfuck. Yes I use GSD and code graph etc.
It's full of the results of millions of meetings, decisions, and feedback from users over the course of many years. A.I. can't just one shot all of that in an afternoon.
I'm working on software that's highly driven by A.I. development right now, and while we're going faster than we would if we were developing it by hand, there's still tons of discussion happening between us and the client, disagreements, decisions to be made about UI and layouts and architecture and other things going on, gaps that don't necessarily be caught by A.I. in the moment, etc. It's still taking 6+ months to put together just Phase 1 for a new corporate web application, with a team of 15+ people.
And we also have to be somewhat careful how many AI tokens we are using because it's not cheap (but we're still using a lot).
1. AI produces a lot of code but very few releases. There have been a lot of articles about how AI has increased code output, but the number of releases on Github for example hasn't increased with the higher coding output. That is because the human review process and shipping is still a bottleneck, and more code doesn't always necessarily mean better software.
2. Which brings me to the next point: AI might have made (arguably) the coding skill more available, but design skills are harder to teach, harder to emulate, and impossible for the AI. Even with all the best AI out there, if you're a poor designer, you won't make a good system. Bad systems don't get hyped up, and as such you won't hear about them.
3. Finally, marketing. Microsoft is a big as it is because it spends billions every year on marketing. Smaller teams guiding AI to make software on the cheap will never be able to compete. You won't hear about them because your company's team in charge of purchasing software is flooded with Microsoft adverts. If not Microsoft, then it'll be the next largest competitor. And so on and so forth. We're not seeing alternatives popping up largely because the people involved in purchasing software are usually not too concerned with finding out the best, they'll take the path of least resistance and buy whatever is cheaper and easier to find. Also there's a question of support, Microsoft Office may be horrible, but if you have a problem with it, you can look at official microsoft documentation or call Microsoft yourself for help. Smaller teams likely won't have the customer support teams to enable large enterprises to use the software effectively.
It takes years for new software brands to be common names. This just started.
There are countless competitors to Photoshop. It's not a lack of software. It's not a lack of "cheap competitors."
The problem is usually a licensing/compliance/support one. Everyone hates Photoshop and Lightroom, but it's what all of the professional plugins support. Nobody likes paying for Microsoft Word, but it supports Active Directory and the weird document signing formats that some businesses use. There may never be a FOSS replacement for these workflows, regardless of AI coding.
> If coding is a nearly solved problem, if AI is making it so easy to make anything, why aren't we seeing the market flood with cheap alternatives to software that we all use daily and hate?
Because that is a business problem, not an LLM and software problem. Perhaps coding isn't the solution for those problems, but companies that can potentially become profitable enough to be competitors to more expensive alternatives. Is a company an alternative to Terraform and Vault? Or OpenTofu and OpenBao, for example. Microsoft Teams won because Microsoft bundles it for free, even though it's garbage. You use it because that's what a CFO or CIO picked.
Code doesn't build systems (for profits and non profits) to continually operate and improve software; people do. Everyone else has mentioned how non-AI alternatives exist. You should ask "Why aren't they used?" and "What would it take for cheaper and/or better alternatives to be used?" LLMs might, but might not be, a tool to change the outcome.
https://xkcd.com/2347/
If a tree falls in the forest and no one is around to hear it, does it make a sound? If someone makes a program for themselves, but doesn't publish it to GitHub and doesn't spend $10,000 on Google and Facebook ads to shove it in your face that it exists so you hear about it, does it even exist? If you're looking for something, go find it! Or are you incapable of doing your own research?