[scottLobster]

Maybe the internet has made me too cynical, and I'm glad people seem to be having a good time, but at time of posting I can't help but notice that almost every comment here is suspiciously vague as to what, exactly, is being coded. Still better than the breathless announcements of the death of software engineering, but quite similar in tone.

[0xbadcafebee]

The other week I used Copilot to write a program that scans all our Amazon accounts and regions, collects services and versions, and finds the ones going EOL within a year. The data on EOL dates is scraped from several sources and kept in JSON. There's about 16 different AWS API calls used. It generates reports in markdown, json, and csv, so humans can read the markdown (flags major things, explains stuff), and the csv can be used to triage, prioritize, track work over time. The result is deduplicated, sorted, consolidated (similar entries), and classified. I can automatically send reports to teams based on a regex of names or tags. This is more data than I get from AWS Health Dashboard, and can put it into any format I want, across any number of accounts/regions.

Afaik there are no open source projects that do this. AWS has a behemoth of a distributed system you can deploy in order to do something similar. But I made a Python script that does it in an afternoon with a couple of prompts.

[probably_wrong]

> almost every comment here is suspiciously vague as to what, exactly, is being coded

Why? You don't trust a newly-created account that has not engaged with any of the comments to be anything but truthful?

[slopinthebag]

In my experience, I have "vibe coded" various tools and stuff that, while nice to have, isn't really something I need or brings a ton of value to me. Just nice-to-haves.

I think people enjoy writing code for various reasons. Some people really enjoy the craft of programming and thus dislike AI-centric coding. Some people don't really enjoy programming but enjoy making money or affecting some change on the world with it, and they use them as a tool. And then some people just like tinkering and building things for the sake of making stuff, and they get a kick out of vibe coding because it lets them add more things to their things-i-built collection.

[tqwhite]

I will say that I grieve the passing of 'coding', per se. I used to love getting the flow, envisioning the data flows and object structures and cool mechanisms, refactoring to perfection. I truly miss it.

But the payoff for letting that go is huge.

[slopinthebag]

> I used to love getting the flow, envisioning the data flows and object structures and cool mechanisms, refactoring to perfection.

You still have to do this, LLM's are still quite bad at choosing the right data structures and dealing with their interdependent relationships.

[kdheiwns]

Yes. I never really see people say wtf they're making. It's always "AI bot wrote 200k lines of code for me!" Alright, cool. Is the project something completely new? Useful? A rushed remake of a project that already exists in GitHub with actual human support behind it? I never see an answer.

[tqwhite]

I wrote SuperSecretCrypt.com, ScoreRummy.com. Other stuff, too.

I have integrated Claude Code with a graph database to support an assistant with structured memory and many helpful capabilities.

I have clients. I automated a complicated data ingestion pipeline into a desktop app with a bulletproof process queue, localhost control panel and many features.

For another, I am writing an AI-specific app that is so cool. I wish I could tell you about it but it's definitely not a rushed remake of anything.

I hope that helps.

[sensanaty]

> SuperSecretCrypt.com

Is down. And the scoring one, no offense, seems like a project a junior would make to pad out their resume/portfolio. Nothing wrong with that of course, but I fail to see how this translates to all the hype being thrown around.

[agenticbanana]

SuperSecretCrypt.com doesn't work.

[idopmstuff]

I am currently using a Claude skill that I have been building out over the last few days that runs through my Amazon PPC campaigns and does a full audit. Suggestions of bid adjustments, new search terms and products to advertise against and adjustment to campaign structures. It goes through all of the analytics Amazon provides, which are surprisingly extensive, to find every search term where my product shows up, gets added to cart and purchased.

It's the kind of thing that would be hours of tedious work, then even more time to actually make all the changes to the account. Instead I just say "yeah do all of that" and it is done. Magic stuff. Thousands of lines of Python to hit the Amazon APIs that I've never even looked at.

[scottLobster]

And it doesn't freak you out that you're relying on thousands of lines of code that you've never looked at? How do you verify the end result?

I wouldn't trust thousands of lines of code from one of my co-workers without testing

[idopmstuff]

> And it doesn't freak you out that you're relying on thousands of lines of code that you've never looked at?

I was a product manager for 15 years. I helped sell products to customers who paid thousands or millions of dollars for them. I never looked at the code. Customers never looked at the code. The overwhelming majority of people in the world are constantly relying on code they've never looked at. It's mostly fine.

> How do you verify the end result?

That's the better question, and the answer is a few things. First, when it makes changes to my ad accounts, I spot check them in the UI. Second, I look at ad reporting pretty often, since it's a core part of running my business. If there were suddenly some enormous spike in spend, it wouldn't take me long to catch it.

[gopher_space]

It's thousands of lines of variation on my own hand-tooling, run through tests I designed, automated by the sort of onboarding docs I should have been writing years ago.

[yieldcrv]

I've been doing agentic work for companies for the past year and first of all, error rates have dropped to 1-2% with the leading Q3 and Q4 models... 2026's Q1 models blowing those out the water and being cheaper in some way

but second of all, even when error rates were 20%, the time savings still meant A Viable Business. a much more viable business actually, a scarily crazy viable business with many annoyed customers getting slop of some sort, with a human in the loop correcting things from the LLM before it went out to consumers

agentic LLM coders are better than your co-workers. they can also write tests. they can do stress testing, load testing, end to end testing, and in my experience that's not even what course corrects LLMs that well, so we shouldn't even be trying to replicate processes made for humans with them. like a human, the LLM is prone to just correct the test as the test uses a deprecated assumption as opposed to product changes breaking a test to reveal a regression.

in my experience, type errors, compiler errors, logs on deployment and database entries have made the LLM correct its approach more than tests. Devops and Data science, more than QA.

[tqwhite]

Why wouldn't you test? That sounds like a bad thing.

Me? I use AI to write tests just as I use it to write everything else. I pay a lot of attention to what's being done including code quality but I am no more insecure about trusting those thousands of tested lines than I am about trusting the byte code generated from the 'strings of code'.

We have just moved up another level of abstraction, as we have done many times before. It will take time to perfect but it's already amazing.

[Capricorn2481]

So people don't look at the code, or the tests.

So they don't know if it has the right behavior to begin with, or even if the tests are testing the right behavior.

This is what people are talking about. This is why nobody responsible wants to uberscale a serious app this way. It's ridiculous to see so much hype in this thread, people claiming they've built entire businesses without looking at any code. Keep your business away from me, then.

[notAnAIBot768]

Do you trust the assembly your compiler puts out? The machine code your assembler puts out? The virtual machine it runs on? Thousands of lines of code you've never looked at...

[scottLobster]

None of that is generated by an LLM prone to hallucination and is perfectly deterministic unless there's a hardware problem.

And yes, I have occasionally run into compiler bugs in my career. That's one reason we test.

[notAnAIBot768]

> None of that is generated by an LLM

How did you verify that?

> prone to hallucination

You know humans can hallucinate?

> is perfectly deterministic

We agree then that you can verify, test, and trust the deterministic code an LLM produces without ever looking at it.

> That's one reason we test

That's one way we can trust and verify code produced by an LLM. You can't stop doing all the other things that aren't coding.

I get there's a difference. Shitty code can be produced by LLMs or humans. LLMs really can pump out the shitty code. I just think the argument that you cant trust code you haven't viewed is not a good argument. I very much trust a lot of code I've never seen, and yes I've been bitten by it too.

Not trying to be an ass, more trying to figure out how im going to deal for the next decade before retirement age. Uts going to be a lot of testing and verification I guess

[xigoi]

> How did you verify that?

The compiler works without an internet connection and requires too little resources to be secretly running a local model. (Also, you can’t inspect the source code.)

> You know humans can hallucinate?

We are talking about compilers…

> We agree then that you can verify, test, and trust the deterministic code an LLM produces without ever looking at it.

Unlike a compiler, an LLM does not produce code in a deterministic way, so it’s not guaranteed to do what the input tells it to.

[bobanrocky]

Compiler theory and implementation is based on mathematical and logic principles. And hence much more provable and trustworthy than a LLM thats stitching together pieces of text based on ‘training’

[bandrami]

"Trust"? God no. That's why I have a debugger

[bandrami]

Also you really do have to know how the underlying assembly integer operations work or you can get yourself into a world of hurt. Do they not still teach that in CS classes?

[hippich]

Some _fun_ stuff i "coded" in a day each just in last couple weeks:

https://hippich.github.io/minesweeper/ - no idea why but i had a couple weeks desire to play minesweeper. at some point i wanted to get a way to quickly estimate probability of the mine presence in each cell.. No problem - copilot coded both minesweeper and then added probabilities (hidden behind "Learn" checkbox) - Bonus, my wife now plays game "made" by me and not some random version from Play store.

another one made in a day - https://hippich.github.io/OpenCamber - I am putting together old car, so will need to align wheels on it at some point. There is Gyraline, but it is iOS only (I think because precision is not good enough on Android?). And it is not free. I have no idea how well it will work in practice, but I can try it, because the cost of trying it is so low now!

yes, both of these are not serious and fun projects. unlikely to have any impact. but it is _fun_! =)

[izacus]

It's also usually from people who stopped coding and haven't kept their skills up.

[metaltyphoon]

Or have no more skin in the game, retirement.

[incr_me]

In the past month, in my spare time, I've built:

- A "semantically enhanced" epub-to-markdown converter

- A web-based Markdown reader with integrated LLM reading guide generation (https://i.imgur.com/ledMTXw.png)

- A Zotero plugin for defining/clarifying selected words/sentences in context

- An epub-to-audiobook generator using Pocket TTS

- A Diddy Kong Racing model/texture extractor/viewer (https://i.imgur.com/jiTK8kI.png)

- A slimmed-down phpBB 2 "remake" in Bun.js/TypeScript

- An experimental SQLite extension for defining incremental materialized views

...And many more that are either too tiny, too idiosyncratic, or too day-job to name here. Some of these are one-off utilities, some are toys I'll never touch again, some are part of much bigger projects that I've been struggling to get any work done on, and so on.

I don't blame you for your cynicism, and I'm not blind to all of the criticism of LLMs and LLM code. I've had many times where I feel upset, skeptical, discouraged, and alienated because of these new developments. But also... it's a lot of fun and I can't stop coming up with ideas.

[fzeroracer]

The combination of the internet and how insanely pushed every single facet of AI bullshit is has made me incredibly cynical. I see a post like this reach the top of HN by a nobody, getting top votes and all I can think is that this is once again, another campaign to try and make people feel better about AI.

Every time I've asked people about what the hell they're actually doing with AI, they vanish into the ether. No one posts proof, they never post a link to a repo, they don't mention what they're doing at their job. The most I ever see is that someone managed to vibe code a basic website or a CRUD app that even a below-average engineer can whip up in a day or two.

Like this entire thread is just the equivalent of karma farming on Reddit or whatever nonsense people post on Facebook nowadays.

[tqwhite]

I wrote SuperSecretCrypt.com, ScoreRummy.com. Other stuff, too.

I have integrated Claude Code with a graph database to support an assistant with structured memory and many helpful capabilities.

I have a freelance gig with a startup adapting AI to their concept. I have one serious app under my belt and more on the way.

Concrete enough?

[agenticbanana]

SuperSecretCrypt.com doesn't work.

[dplgk]

Yes and they all mention Claude as if it's the only LLM that can code.

[yieldcrv]

think about why anybody would ever associate a production level product with slop when consumers are polarized towards generative AI

this site gets indexed

there are too many disincentives to cater specifically to your suspicion and cynicism