4 ms·
LLMs have burned Billions but couldn't build another Tailwind
- JKFSOM 10mo agoLLMs could absolutely replace Tailwind, and if there was the need, aid creating something better.
- sevenseacat 10mo agoso why haven't they?
- george_atom 10mo agois there a need?
- eadwu 10mo agoI think you are expecting some kind of collaborative OSS type of library. For LLMs to have replaced Tailwind - in part by using it themselves - this does not have mean that there _will_ be another library to reuse. In the context of LLMs, it becomes so "cheap" to customize the webpage that a library is no longer needed. Tailwind in and of itself can be considered a "highly structured LLM" - if they so took it that far.
- jaapz 10mo agoLooking forward to frontpage/dreamweaver levels of css soup already
- burnt-resistor 10mo agoThe future will gradually become proprietary and non-textual programming languages, toolchains, and runtimes inaccessible to humans.
- jrm4 10mo agoCould? I'll go a bit further. I am 100% certain they will.
- etothet 10mo agoIt feels like the author is conflating Tailwind the open source project vs. Tailwind Labs the business, which Adam says has run into a bad situation with their revenue because Tailwind Labs’ paid projects aren’t getting as much traction as fewer and fewer users visit the official Tailwind docs, which is currently the primary source people find out about those commercial products. Regardless of what happens to the company (my personal opinion is that they’ll come out of this strongee than before) Tailwind as OSS probably isn’t going anywhere for the foreseeable future.
- durch 10mo agoLLMs find the center of the distribution: the typical pattern, the median opinion. Tailwind was an edge bet. It required metis, the tacit competence to know the consensus (semantic classes, separation of concerns, the cascade) was a local maximum worth escaping. That judgment, knowing what the center is wrong about, doesn't emerge from interpolation. It emerges from the recognition loop where you try something, feel "that's not quite it," and refine. The bottleneck was never typing. It was judgment. Tailwind is crystallized judgment. AI can consume it endlessly. Producing the next version requires the loop that creates metis, and that loop isn't in the training data.
- hoppp 10mo agoYeah, I was talking about this a year ago when I noticed claude pretty much forces tailwind on people by defaulting to it, that its gonna impact their template selling business. Tailwind needs a new license that requires payment from AI providers if they generate it!!
- MattDaEskimo 10mo agoTailwind itself is not useless, but the plus package is. It's a simple convenience utility belt that LLMs can already automate. Both open-source and open-core need to be re-evaluated as labor value plummets. I also disagree with the "why". Tailwind is extremely useful with LLMs as it can set styling inside of HTML, rather than maintain an external, typically massive, convoluted CSS file. It's for the same reason that typing in programming will become a standard with LLMs: eliminate implicit/semantic density with explicit/semantic precision. In both examples an LLM can have a strong understanding of a single file/module without needing to search for its meaning externally
- gavmor 10mo agoOur concept of "Tailwind itself" should encompass, really, the economic foundations in which it's ensconced and from which it grew, even if it grew "on spec." Ie, without the promise of some path to profitability, eg the plus package, the mere library cannot justify itself. So are we to stop building FOSS on spec?
- jrm4 10mo agoSo, duh LLM's are extremely adept at turning "what Tailwind does" into something "you don't have to pay for." Yes, generative AI destroys some code-related business models. Absolutely fine by me, I'd rather work towards more people being able to be more creative with code than some company putting code tools behind paywalls or whatnot.
- etyhhgfff 10mo agoAnd even more important, they couldn't build another Taiwan.
- burnt-resistor 10mo agoNot yet. The problem is China is much more likely to attack Taiwan before there is a geographically redundant supply chain or nearly autonomous AI chip design and the complex part of fabrication.