If you value: security; loose coupling; explicitness; disk space; separation of concerns and being able to focus on the business domain of your software: stay the f*** away from Nuxt and Next

Test rant

https://justice.gov/epstein/files/...

@12bitfloat, we're having a new compatitor in the game:
***benchmarking***
Time C: 0.41911959648132324
Time Rust: 0.4686765670776367
Time CPP: 1.7605431079864502
Time Borded CPP: 1.140819787979126
Time Jest Rust: 1.7274057865142822
Time Swift: 0.4625704288482666

It's swift: https://retoor.molodetz.nl/retoor/...

How does it run locally at your PC?

What happened to ostream? I miss his stupid shit rants.

~2020
Us: "We should switch from user-defined table types to JSON for stored procedure parameters. Its too much hassle to make database changes when we need to add or remove fields."
DBAs: "JSON is just a varchar and not structured. We can just keep adding number suffixes to the type name. See, Invoice15 works just fine."

~2022
Us: "Specifications changed, here are the fields we'll need to update .."
<4 hours later>
DBAs: "You developers need to make up your mind on what fields you need updated. My team doesn't have time to keep creating new user-defined table types and updating all the corresponding stored procedures."
Us: "Why not use JSON? That way the schema binding really falls on us to keep everything in sync. New fields or removing them would be easier, faster, and safer on a case-by-case basis."
DBAs: "NO! JSON is just a varchar and too slow for mass updates!"
Us: "We haven't needed mass updating since moving the data notification pushes to Service Bus almost 10 years ago."
DBAs: "JSON in SQL Server is still too slow, unsecure, and hard-coding field names in string is not best practice!"

~2024
DBAs: "We're only using JSON for this one stored procedure. Its going to be up to you guys to keep the your schemas up to date and be prepared to constantly make changes. We may not support JSON if this doesn't work."

2026: <more than 80% of the stored procs use JSON to pass in parameters>
Us: "Specifications changed, here are the fields we'll need to update .."
<30 seconds later>
DBAs: "Done. Fields ready to go in the JSON schema. If you need to add or remove fields, just let us know. Dealing with JSON is sooo much easier now."

Just doing some math about AI financials and its pretty crazy:

A one-gigawatt data center allegedly costs $50 billion to build.
Lets say its written off totally over 5 years. The building, cooling and power infra won't be,
but the GPUs and CPUs -- which are really expensive -- will be much quicker than 5 years,
so I think this is a fair assumption, not to mention personell and power costs

$50 billion over 5 years is $833 million per month, divided by $20 is 41 million

OpenAI needs 41 MILLION (!!) customers that pay $20 a month just to pay for
this ONE data center, and obviously they require a lot more than that

How do you cope with genuinely hating your employer? I’ve been fully remote since 2023, and at this point I can’t ignore how badly it’s wrecking my mental health.

What makes it even worse is the company itself. It’s a 5,000+ employee mess built around ancient, proprietary, closed-source “high-security” appliances. Every single day, working in this environment drains whatever motivation I have left. It’s a constant barrage of bureaucracy, cargo-cult security policies, and security-by-obscurity nonsense held together by duct tape and ego.

They love preaching about “security” while running a brittle, opaque infrastructure that actively fights anyone who tries to do things properly. Every meeting, every ticket, every decision just reinforces how little they understand modern software or trust their own engineers.

At this point, even logging in feels exhausting. The disconnect between what they claim to value and what they actually build is demoralizing, and the longer I stay, the harder it gets to care at all.

:big sigh: The industry recommends writing clean code... and what do we see pervasively in real projects at work? : )

I think you'd have to be lucky to join a company that doesn't write messy code... not just in one project, but in all. Legacy. I hate legacy. Arrghhh.

lol.

Hr is where we keep anyone who shouldn't be allowed to get near anything useful. Now if we just could make some chatbot that fills their useless surveys and performance reviews... we could put this entire occupation in a matrix and be rid of them forever!

There was a time when Java 8 and React 16 were all the hype as if it were yesterday... and now we have Java 25...

The world of IT just moves way too fast. lol

if AI replaces all the humans what will the humans do do you think?

Liquid glass, my arse.

devRant is the ultimate backend to test my proxy server with since it's random responses. When your proxy server can handle devRant scenario's, it's robust. Making a proxy server that is actually robust and don't let it flaws itself with slow connections / servers is quite a challenge. Depends also what you're writing it with. In python was way simpler than in C. In Python some things magically just worked. If you think that python sockets are native, you're wrong kiddo, they did al sort of shits under the hood. Had to do a whole rewrite of my original proxy server to allow pipelining. Pipelining is req>req>req>resp>resp>resp instead of the traditional req>resp>req>resp>req>resp. It's keep-alive on steroids. Most server don't support it out of self preservation but I just want to. Caddy is a huge pipeliner. What most servers do is, they close the connection the moment the second request is sent / marking the second request invalid. Clients support that behavior and will do the second call under the hood.

My new proxy server has exactly the same dashboard as the previous one but the many statistics that it has have zero performance impact. It persists every second a quick snapshot to database, every 60 seconds a big one. The statistics are persistent for when the server reboots.

What an adventure.

I do apologize for the radio silence regarding youRant, I have had my hands VERY full recently, and every time I think I'm going to get some time to work on projects like youRant I remember yet another thing that I urgently need to take care of. Plus, University.

Do hope I'll get some time to work on it, add profile pages and whatnot

Why I still don't fully believe in vibe coding: Even the frontier models are constantly wrong!

Claude says c# structs with LayoutKind.Auto can have differing layout for each usage or JIT compilation. ChatGPT says they can't (and that's what I assume is true)

Claude says to "Multiply by a large prime", 0x9e3779b97f4a7c15
...That's NOT a prime!

AI generated code looks really convincing but they always make these little mistakes. And even if you inspect the AI code yourself (which let's be honest, we're all too lazy for that :P), will you be able to spot these? I really doubt it