And then

Written on: 2026-03-17. Last updated on: 2026-03-17.

Recently I have read several posts in the fediverse, Reddit, blog posts about the current era of LLMs being used for Software Development from comparing Vibe coding and AI assisted engineering where everyone can build anything to the fear of 3D printing where everyone can build anything and the same results, because you need experience and resources to know what you're building and actually use it, to the realization that the statistical prediction of completion for tools like Claude Code, ChatGPT among others is only that, predictions, but only for the coding.

The purpose of these tools was never to help us be more productive or write better code but to reduce the workforce and by this reduce costs for Company Owners and Investors but they are now realizing that their investments in AI will not return any profit in the short time. As a matter of fact instead of being cheaper, the cost of providing these tools for the companies is increasing and the benefits stalling.

So there are some things we can learn from this:

  1. Software Development is more about what are we building and why we are building it than how are we building it
  2. Modern Software Development is a distribution issue not a build or publish issue
  3. Software should always be built to last not to be changed in its roots every 6 months
  4. The coding is not enough nowadays, maybe the people outside our field thought that the coding was the issue, but it never was at least not the last 25 years.

Testing, deploying, maintaining, performance, accessibility, security, portability, and do the right thing, that many times it's not solved by writing more code but by removing lines or by thinking how to solve a problem are not solved by probability models but by a process we've been refining the last 75 years and we still need to perfect, but that's only my opinion.

There are still a lot of things that we need to build but for ourselves and for the people who need it when they need it not creating the problem and then selling the solution. Although there is a recurrent problem that we could solve without anyone asking for: the dependency of the software provided by the Big Tech companies and their subscription model.

I, for example, want to solve one problem: to help people own their websites and I think there are many solutions but I want to offer the simplest: static websites generated by static site generators or JAMStack tools and although I know that using JAMStack tools maybe it's not always the best solution, I think it's better than not owning your own website.