I have to think about this. Back in the days when rapid prototyping meant checking things out with (SHUDDER) Basic, or Tcl, or later, Hypercard, I actually learned something along the way that came in very handy later in a different context.

So, a dumb problem came up --- Really dumb --- how to make the world's simplest online calendar so that someone could just enter event dates (there are less than 10 a month) and not wait until the very last one came in, for an email to go out. Key dates are known and it's the 13th and no January calendar yet.

So, it's like convert date formats, which Numbers refused to do, and then pair that column with days of the week. Before I went back to the days of "Unix Text Processing" and the insanely useful "paste" command. That came from the days when I studied that book and put a ton of it into practice. It was a guide top shell and built-ins via example.

Doing this, I thought that perl would solve it all, and despite not having used it in countless years (I retired long ago) it started coming back to me.

But the shell stuff worked better than a spreadsheet (software from hell) and perl, which can have some incredibly poetic solutions. I promised myself to re-read what I found. Because it's "neat".

I learned a lot before even usenet and Google, and that has stuck for the most part, because "I had nobody to lean on".

I agree with this wholeheartedly! Even something a small as fixing up some horrific diverged orphaned git history can help me on a hobby project. And the fact that I see the commands that, or the way a containerfile/caddyfile is written, helps me learn along the way. Not as good as a textbook, but like the author I'm reading a textbook for <this thing over here> not <that boring infra stuff>.

Kinda summarises my experience: I get a lot more done but learn a lot less

I can now claim I have experience with Docker, AWS, Rasberry PIs, … but don’t ask me to do it myself manually

Another thing that doesn’t get mentioned enough is that AI can help you quickly breadboard concepts to see if they’re viable at all.

One of the worst experiences is having an idea for a game, spending a bunch of time building a prototype, and then discovering in playtesting that it’s just not fun.

It still stings but at least with AI you’ve invested far less time getting to that point.

*Case in point*

A few days ago I built a tiny web app that generates a random sentence, speaks it using TTS, records the result with Chrome’s built‑in speech recognition, and then loops the process to see when it converges.

Turns out the Chromium speech recognition is really good, so at best it would unroll contractions (don't -> do not) and stabilized almost instantly whereas I was hoping for a rather crazy TTS version of the old game of phone tag.

Total time spent: about 20 minutes and most of that was it quietly building in the background while I noodled on my piano.