← Back to Feed
retoor
retoor · Level 57162
rant

Every time

I love doing challenging stuff. Esoteric systems, weird complexity - my only goal is education. Investing in myself. Even DevPlace suffers from my esoteric choices, but so far that went great.

Every time I make something it's quite advanced. But only for one month. I try the max of an LLM and then I start to build. Take Pravda for example, that system is insane. I just watched a video of a guy who created it with one prompt to Fabel 5.2. ONE PROMPT. I have great result, way better than original - we are all copying grok bot - and probably also better than the one-prompt software, but I did not check.

Whatever you build these days, the next LLM will make your skill worthless again. Fine by me, I keep on going and will never run out of ideas. It's very easy to have ideas if money is not a factor. A profitable idea, that is hard. But if your goal is self-education, you can literally do everything and as you guys know, I do that. I do everything. Who the fuck is spending weeks on a TUI application in 2026? If there will ever be a book about effort to money ratio I would be at the absolute bottom while still living well :P I consider it all luck, if you have interest in something that makes money is luck.

But currently, I think I found a way to get money in my own category. I joined all my applications / technologies together and gave them all a pg-prefix. pg stands for pretty good. So you have:

pga - pretty good applications, the main product
pretty good telegram client
pretty good telegram bot
pretty good ide - the TUI vs code
pretty good tai - pretty good terminal AI
pretty good db - lives in all applications, will deliver postgres-like performance but still your nice single file
pga orchestrator - literally the TUI version of Pravda. Supports all features but on terminal
pgwui - pretty good web UI, not a web application at all. Clone of that in TUI form.

Audience? Everyone actually. The ide is by default hybrid vim / vscode since their shortcuts do not match at all. An elite will be as happy as a normie. You wanna paste something? Fine, ctrl+v works but also ,p (, = leader). The built-in IDE looks like vscode but if a vim user would close its eyes and do its thing, it will have exactly the same results like vim while it opened a file browser and everything. So you are controlling vscode ide perfectly by using just vim shortcuts. But wanna edit your file? Sure, same keys as always but a dialog shows up and navigates to the path you choose. It looks very cool.

What I really like - these are all existing projects that depended on each other, and it is a mess. So I introduced PGA - pretty great applications - and that will wrap all making it look like one system with consistency. And at a certain point, I will make the wrapper the real application. Without AI completely unthinkable. Because every app I work on is around a week or so in total minimal using AI, imagine the amount of code per app. Also, in the end I will force my code standards, a few went out of control.

But my point: whatever you make, AI will catch up. But if you are making applications that take weeks to months work using AI, even AI can't keep up with you and that is what I am doing.

I am going very niche. Most people won't work on a webapp written in C++ wasm. Sad, since it is 100 times more responsive than any web application and the look and feel of xfce is perfectly simulated. Yeah, like always I won't get rich but damn, I feel good pulling certain stuff off.

You have two kinds of people in life, people who think they can do something and people who actually did. I always want to be the latter, that's where the confidence comes from but it does not work for many people - it means you are always failing because you try to work on stuff you don't know yet. The snekbots are the worst and best things I've made I guess. We have now perfect devii, perfect news, perfect meta descriptions filled on the site and all is written with the help of deep research that is written by me as well. Snek bots were very bad for ego, but look how great this site works regarding AI! But as long as I am the main maintainer we will always have billing issues ;). Working on that - later today we will have meta AI back and plenty of resource but I don't want to have the site running on hacked AI. I found a place a while ago where you could buy 3bln tokens or something for 50,- deepseek flash. That's way more quality and resource than we need. If I would do that, I don't have to be worried for a ban on the API like I do now of course.

Tired, just had to get some stuff off my mind.

Will retoor ever work again or have a company
Log in to vote

Comments

No comments yet. Start the discussion.