The darkness is always where the light is again found. The people who caused this situation are burning all their bridges on this. There will be a turnaround.

I like how you think

"Adversarial umbrella" is my band's new name

Seems plausible. I used to refer to StackOverflow before LLMs and a good amount of the examples there were flawed code presented as working. If the LLM had less junk in its training then it might benefit even though the volume of training on that language is lower.

What is a coder? Someone who is handed the full specs and sits down and just types code? I have never met such a person. The most annoying part of SWE is everyone who isn't an SWE has inane ideas about what we do.

Never worked on offshoring projects? That is exactly what the sweatshop coders do.

No we don't.

For one, I never saw a "full spec" (if such a thing even exists) back in my days of making 8k. Annually.

I think that the current AI tooling is a much bigger threat to offshore sweatshops than to domestic programmers.

Why deal with language barriers, time shifts, etc. when a small team of good developers can be so much more productive, allegedly?

It certainly is,

https://www.theregister.com/2026/01/19/hcl_infosys_tcs_wipro...

This is true. But there's also an overwhelming instinct for people to think the problems they face today are completely new and have no historical precedents

Fascinating. I don't understand the technical terms, but running a big coding agent locally is a dream of mine, so I thank you for your efforts!

Yes, trust the CIA, Iranians. They've never led you wrong before.

Articles like this seem to be just a way to have something to write about that mentions LLMs. You could just as well say "Hacker used internet to attack government agencies in Mexico" and make the article about how their use of the internet greatly enhanced their ability to do so. Of course they used LLM tooling in 2025 to carry out an attack.

Does this really work? Seems likely to be more hype.

Codex makes all kind of terrible blunders that it presents as "correct". What's to stop it from just doing that in the loop? The LLM is still driving, same as when a human is in the loop.

ralphs is good at letting things complete, but is far from "making commercial software for $10"

just the initial coding first requires you to actually define what the output is

if somebody can make a cleanroom agent that can explore and document specifications for commercial software, you could maybe throw ralph at building it, but then you still have to work out the parts that dont have documentation/training details, like how you are going to maintain it

the loop is pretty perfect for something like "my dependency updated. decide whether to update to match, and then execute"

itll do the mediocre job and then keep trying till it gets something working, at probably the most expensive token cost possible

I've been having success using Doppler for secret storage. Takes it off the filesystem.

My question is not about on or off storage, is more about when you give agent access, it assume the environment agent runs is safe