Longest rant of devplace so far. We do not have AI yet and especially not the blablabla intelligence people were so hyped for.

Time for a divorce. A divorce with Grok. But sadly, since it does a Web search superfast for every query, it's by far my favourite and in my opinion the most truthy one.

But we've been talking and it was about security. It does exactly what most humans do, panic, panic, panic. I say "if I do it x and y way" he be like, wel, that does not have to be secure because your code can have a memory leek and blablabla. And I tell it, so what if it has that, it's not directly in anyone's favor and it basically.. Is a very theoretical thread. A memory leek that will be actually usefully be exploited delivered by the host app... It just does apply to my case since the memory leak will be in a contect that the script that is executed (own language, mind you, this own wm).

Stop thinking, we're gaslighted by now. So first, I have to make a memory bug, on a level that is completely different than the script lives, and then, on top of that successfully be exploitable by a script that does literally provide no syscalls functionallity, a real file system, memory management etc. You could consider it comparable with js in browser.

So, I was like, I gonna write now something in python that is guaranteed safe, like, well, it does a lot for ya. But then, it said, it's not safe. And I was like oooeeeeh.. So what's wrong? It said: the python functions could have a bug (non syscall functions people) and was like, now I'm done.

Conclusion, a lot of stuff is just safe in the world, puss and dick pics of all your exes are actually somewhere stored right?

But this style of conversation, you will gain nothing with it. It's just not helpful.

Is a car safe? Sure, it has breaks that are checked every year and so far I've never heard of someone wherefrom the brakes completely failed. "But what if they do fail?" > well, actually, you have a certain zone / cage protecting you and > no! This conversation should already have ended. Do whatever you want AI, go walk or whatever llms do but I take the car.

The AI that we have is just not intelligent. It is not. It even hardened mainstream opinions that were always not very based.

Example, I write servers since I was 15. Networking and stuff was a bit late because I lived in middle of nowhere. I kept it all as a fetisj, did actually read books dedicated to it, some common books like hacking and violent python and whatnot. Well, we all know, never build your own database, and also not a mailserver. But I was like, my emailsetup is broken again, it was mailcow. Great stuff, easy to setup, but when it broke I was like hmm. This stupid product is just glue between a lot of shit and how it really works is not very common knowledge I guess. But ofc, I know the protocols etcetera, networking, and with this knowledge PLUS ai, we are in a way different situation than what it was when people said "Don't build your own mailserver". So, I started building the fucker. Can write a book about it. But, I made a few very good choices: one, I make it in python, because it is hard enough not getting fucked by memory issues. The second good choice was a monolith, what is actually the core feature - no config / dependencies anymore. It would be solid as fuck. So, I removed complexity with monolith approach and was not so stupid to write it in a language costing much concentration. Finally, a personal preference, no encryption at all and a smart way to store all email files. Not even an sqlite database. No, also not for searching. Read your whole sda disk to /dev/null > zoef. For email, we do with the hardware of today not even a database. I did keep the possibility to fail at some point so it would be very nice to have all data in file format. In my opinion very reasonable so far.

New line 😁

So, I start deep researching (I have many different cli tools for that, it's an serious addiction) all the stuff I wasn't sure or may not be wrong. Critical information. It ended up with protocol specification documents that it should minimally support to communicate with mobile devices and whatnot. That was nice. When everything was deep researching and building documents I wad prompting for many hours. I knew exactly what I wanted and how i wanted it, what should be isolated and what to process async and so on. These are the days that I reach 100k keypressrs on a day.

Ah fuck, I'm getting tired. Tldr; I have now the first mail server ever that I understand and that feels very good.

What is my point: 75% of the tie first part the AI kept demotivating me in the process with even refusals. AI is motivating you not to develop yourself and tells you the same shit everyone does: do not build a mailserver. Now finally when it's finished, it literally declared it art. Well, fucker. How many cool stuff that requires some effort did you cause not to be created.

In the future: do not configure your own hosting, do aws. Actually, that is current situation already.