Just yesterday OSS community (me included) found a new topic to reflect on.

TailwindLabs (creators of successful CSS framework TailwindCSS) got a pull request to add llms.txt to their documentation site. It was declined with a reason.

Reasoning (full in a screenshot): LLMs are already killing TailwindLabs business, due to lower traffic on their documentation (down by 40%) which cuts revenue by 80%. That led to job cuts: 75% of the people on TailwindLabs engineering team lost their jobs.

Here's the deeper issue: if established, popular libraries like Tailwind are struggling, what chance do new, innovative frameworks have? This brings me to a broader concern.

After long night of self-reflection I came to broader statement: LLMs are killing innovation (especially in web development) and OSS (open-source software) businesses.

AI is supposed to help us work less on stuff that are boring, so we can focus more on complex solutions. That sounds like we will be innovating more. I am sure we will in some areas.

I am mostly worried about my area of interest which is web development.

Nowadays vibe-coding is becoming a thing. It is easier than ever to create a website. You can ask general purpose LLMs to do the work or use specialized services (like Lovable.ai).

90% of those generated websites are using React and... Tailwind.

Itโ€™s not surprising that LLMs are choosing the most popular libraries. But it worries me a lot, that we might land in time where everything will be React based.

Donโ€™t get me wrong, I like React and work with it since 2015. (Anybody remember React without functions, hooks or even classes?)

Problem I have is that new frameworks donโ€™t have a place to be in AI era. Best example is that Claude Opus writes pretty decent Svelte code, but unless you define it, it will go with React.

For me React becoming the default itโ€™s a bad thing.

New frameworks always shows that there are new ideas how to solve complex problems. Perfect example is how React took over the web. React solved the issue with reactivity of webpage elements. Until React any change of elementโ€™s โ€œstateโ€ required full refresh of a page. React influenced changes in Angular. It influenced the raise of Vue and Svelte.

If we leave everything to LLM, there will be no new ideas. No new frameworks. No innovation.

But thatโ€™s not all.

Letโ€™s come back to Tailwind. As for now, it is a default CSS framework for any AI-generated website. TailwindLabs was able to maintain solution that drives gen-AI CSS. TailwindCSS is also the foundation of bigger popular abstraction, shadcn, which is also used by GenAI. (Shocking!)

What will happen if TailwindLabs will disappear and TailwindCSS become deprecated (or worse, it will have security breaches). No one will be able to fix issues with Tailwind. (Of course, there is still concept of forking the library, but that might introduce fragmentation within one library.)

Right now LLMs are good at reading any documentation, and create code based on it. But OSS (especially libraries) creators need money to fully focus on the solution. They often use their documentation as an ad for sponsorship. With llms.txt file, LLM donโ€™t open the website or even propose supporting OSS creator.

This whole situation questions the existence of independent OSS solutions in my opinion.

It doesnโ€™t feel right and creates more questions:

  • Does only BigTechs (whom currently are AI biased) are able to fund OSS?

  • Will AI โ€œkillโ€ more OSS solutions which are not funded by bigger companies?

  • Will TailwindLabs survive as independent or be bought or will disappear?

  • What will happen with TailwindCSS if TailwindLabs will disappear? What will happen with libraries that are based on Tailwind like shadcn or daisyUI?

Link to comment in PR: https://github.com/tailwindlabs/tailwindcss.com/pull/2388#issuecomment-3717222957