What's your "I wish I knew this before starting" tech stack decision?

Don't jump on hyped technologies (like Kubernetes, or Kafka) from day 1. You would be surprised how much you can do with $5 VPS running PHP + Postgres.

Don't fancy your tech stack, shipp the bloody project!

- With Next.js: That it works best on Vercel and has issues on other hosts, especially around caching and invalidation (which are big reasons to use Next), e.g. on Cloudflare https://github.com/cloudflare/next-on-pages/issues/292 or Netlify https://answers.netlify.com/t/netlify-support-for-next-js-ca.... I don't think this is some big conspiracy or that they deliberately brick other hosts, but there's an undeniably an inherent conflict-of-interest there. Next itself is open-source, but Vercel only makes money from hosting, so if they make it too easy to deploy elsewhere with all the features working, they don't get paid at all. I kinda wish Next followed an open-core model instead, like MUI, and had more vendor agnosticism. As a frontend dev I love Vercel, but their business model makes it a harder sell to higher-ups who are afraid of vendor lock-in.

- With web dev in general: How quickly everything gets obsoleted. I started back in the 90s/early 2000s, and over time sooooo many fashion trends have come and gone (ColdFusion, Perl, Java, ActiveX, RealVideo, framesets, Shockwave, Flash, Silverlight, Air, Bootstrap, jQuery, XHTML, XPATH/XSLT, Angular, tables, CoffeeScript...). And some things go through cyclical periods of popularity, like clientside vs serverside rendering vs static pages, various VR/AR implementations, etc. By and large I don't know that the web is really that much better today than it was in the 2000s. Is all this additional complexity truly justified? XMLHttpRequest (what we used to call AJAX) and media queries/responsive pages were pretty big leaps forward, but it seems like we've just piled on unnecessary complexity ever since. Will we look back in five years and scoff at Tailwind and Next? Moreover, we've gone from the thriving small web (GeoCities, personal blogs, academic sites, etc.) of those days to what, like 2-3 big social media walled gardens and endless seas of SEO garbage?

- With .NET: It seemed to hold such early promise and I was really excited to start learning it around version 1, but Microsoft did what they typically do. Everyone was still deathly afraid of their EEE strategies back then, and Microsoft never really made it truly cross-platform, especially at the GUI layer, until decades later. By then it was much too late. Now I think it's still used in the E.U. but I haven't met a U.S. .NET dev in decades.

- With Drupal: Oh wow, so THIS is what childbirth feels like. Yeesh. Never, ever again.