No, absolutely the opposite. LLMs are terrible at things that require judgment and justifications, because they don't reason. They come up with something that sounds plausible.
That's not good enough when you're dealing with matters that can lead to civil or even criminal liability. Errors can be incredibly expensive to fix, if they can be fixed at all.
With a CPA or attorney, you at least have recourse if they screw up. You don't with LLMs.
Oh I'm not at all joking. It's better at evaluating quality than producing it blindly. Tell it to grade it's work and it can tell you most of the stuff it did wrong. Tell it to grade it's work again. Keep going through the cycle and you'll get significantly better code.
The thinking should probably include this kind of introspection (give me a million dollars for training and I'll write a paper) but if it doesn't you can just prompt it to.
Think of it as a "you should - and is allowed to - spend more time on this" command, because that is pretty much what it is. The model only gets so much "thinking time" to produce the initial output. By asking it to iterate you're giving it more time to think and iterate.
Yeah, this is where I start side-eying people who love vibe coding. Writing lots of tests and documentation and fixing someone else's (read: the LLM's) bad code? That's literally the worst parts of the job.
I also get confused when I see it taken for granted that "vibe coding" removes all the drudgery/chores from programming. When my own experience heavily using Claude Code/etc every day routinely involves a lot of unpleasant clean up of accumulated LLM slop and "WTF" decisions.
I still think it saves me time on net and yes, it typically can handle a lot on its own, but whenever it starts to fuck up the same request repeatedly in different ways, all I can really do is sigh/roll my eyes and then it's on me alone to dig in and figure it out/fix it to keep making progress.
And usually that consists of incredibly ungratifying, unpleasant work I'm very much not happy to be doing.
I definitely have been able to do more side projects for ideas that pop into my head thanks to CC and similar, and that part is super cool! But other times I hit a wall where a project suddenly goes from breezy and fun to me spending hours reading through diffs/chat history trying to untangle a pile of garbage code I barely understand 10% of and have to remind myself I was supposed to be doing this for "fun"/learning, and accomplishing neither while not getting paid for it.
> "vibe coding" only sounds cool if you don't know how to code but horrific if you do.
Disagree. Vibe coding is even more powerful if you know what you're doing. Because if you know what you're doing, and you keep up with the trends, you also know when to use it, and when not to. When to look at the code or when to just "vibe" test it and move on.
I stick by the og definition, in that when vibe coding I don't look at the code. I don't care about the code. When I said "vibe test it" I meant test the result of the vibe coding session.
Here's a recent example where I used this pattern: I was working on a (micro) service that implements a chat based assistant. I designed it a bit differently than the traditional "chat bot" that's prevalent right now. I used a "chat room" approach, where everyone (user, search, LLM, etc) writes in a queue, and different processes trigger on different message types. After I finished, I had tested it with both unit tests and scripted integration tests, with some "happy path" scenarios.
But I also wanted to see it work "live" in a browser. So, instead of waiting for the frontend team to implement it, I started a new session, and used a prompt alongt he lines of "Based on this repo, create a one page frontend that uses all the relevant endpoints and interfaces". The "agent" read through all the relevant files, and produced (0 shot) an interface where everything was wired correctly, and I could test it, and watch the logs in real-time on my machine. I never looked at the code, because the artifact was not important for me, the important thing was the fact that I had it, 5 minutes later.
Fun fact, it did allow me to find a timing bug. I had implemented message merging, so the LLM gets several messages at once, when a user types\n like\n this\n and basically adds new messages while the others are processing. But I had a weird timing bug, where a message would be marked as "processing", a user would type a message, and the compacting algo would all act "at the same time", and some messages would be "lost" (unprocessed by the correct entity). I didn't see that from the integration tests, because sometimes just playing around with it reveals such weird interactions. For me being able to play around with the service in ~5 minutes was worth it, and I couldn't care less about the artifact of the frontend. A dedicated team will handle that, eventually.
This is one of the things I've seen it be very useful for: putting together one-off tools or visualizations. I'm not going to maintain these, although I might check them into version control for historical reference.
I recently ran across a package in my team's codebase that has a bunch of interrelated DB tables, and we didn't already have a nice doc explaining how everything fits together - so I asked the AI to make me a detailed README.md for the package. I'm currently reviewing that, removing a bunch of nonsense I didn't ask for, and I'm going to run it by my team. It's actually pretty good to start with because the code and DB models are well documented, just piecemeal all over the place, and half of what the AI is doing is just collating all that info and putting it in one doc.
If you know how to program, vibe coding is useless. It only ever can produce worse results than you could've made yourself, or the same results but with more effort (because reviewing the code is harder than creating it).
Depends on what you're doing. I've found it extremely useful for creating the boilerplatey scaffolding I'm going to be copying from somewhere else anyway. When I actually get into the important logic and tests I'll definitely write those by hand because the AI doesn't understand what I'm trying to do anyway (since it's usually too novel).
Right but there are tons of examples of things that started out as insults or negative only to be claimed as the proper or positive name. Impressionism in painting, for a start. The Quakers. Queer. Punk. Even "hacker", which started out meaning only breaking into computer systems -- and now we have "Hacker News." So vibe coding fits in perfectly.
> Even "hacker", which started out meaning only breaking into computer systems
No. The Etymology of Hacker in the technical scene started at MIT's Tech Model Railroad Club in the late 1950s/early 1960s, "hack" described clever, intricate solutions, pranks, or experiments with technology.
A hacker is one who made those clever solutions, pranks, and technology experiments. "Hacker News" is trying to take it back from criminal activity.
TIL, thanks! Growing up I was only aware of the criminal version -- I didn't realize it grew out of an earlier meaning. I just saw the shift in the tech scene in the 1990s and more broader culturally in the 2000s with "life hacks" and hackathons. What's old is new again...
Sorry, which web shops demand governmental ID? I have never had to provide them mine in any of the countries I’ve lived in.
If your concern is the webshop finding out your address, well I’m unsure how you solve this when you buy with crypto, but again ship to your home. If you have an alternative place to get it delivered for privacy, might as well do that with fiat transactions the same way.
The store usually* doesn't demand it, but your ID is tied to your cards via your bank's KYC obligations anyway
* It's becoming more common for sites to ask for ID, I've gotten prompted for it buying a cellphone online, to access an old Facebook account and even Hetzner (Before ever using it) because I got flagged as high risk
Ok, let’s go step by step through your processes, since I am tired of crypto nerds LARPing as Jason Bourne.
How did you first obtain your crypto? What level of anonymity was available for that tx?
Where do you store your crypto short and long term? How do you make it available for spending on online platforms? What percentage of your income and expenditures is in crypto? How do you balance between fiat and crypto anonymously?
What are you buying with the crypto? Why does it need to be purchased with crypto?
Where are you having it shipped? Are you faking all contact details when making the purchase?
Are you completely obscuring yourself physically while collecting said package? Are you obscuring your movements along the way as well to prevent leading back to your home?
Often, proponents love to portray citizens in economically ruinous governments in SAmerica as ideal usecases. Why do they need to use your specific crypto coin? Why can’t they use a locally invented (read: forked) one? It feels much more useful to regulate supply/demand where all said economic activity will take place, instead of replacing your entire net worth from a dying currency to a speculative one mostly propped up by foreigners like you who have zero skin in their local game.
I could go on and on, but it is exhausting to reiterate common sense - no one ever thinks this through fully from the comfort of their air conditioned first world white collar desk job office. How are you ensuring perfect info and op sec in your crypto journey?
> How did you first obtain your crypto? What level of anonymity was available for that tx?
Suppose you mined it or received it as payment for selling something to a stranger who doesn't know your identity.
> Where do you store your crypto short and long term?
If you're using it as a payment method you don't store it long-term. You either spend it promptly or convert it into some ordinary form of investment.
> What are you buying with the crypto? Why does it need to be purchased with crypto?
Whatever you want to buy with it. Suppose you want a VPN subscription. Suppose you want to make an anonymous donation. Suppose you're just eating at a restaurant and don't want a record of that on your bank statement.
> Where are you having it shipped?
There are lots of things you can buy that don't need to be shipped.
Also, that's a separate problem. If you actually had such requirements then you would have to deal with them, but first you'd need to solve the problem of not having a charge for the thing you want to be private appearing on your credit card statement tied to your government ID.
> Why do they need to use your specific crypto coin? Why can’t they use a locally invented (read: forked) one?
Because a major benefit is to be able to make transfers between countries.
I never said anything about my usage of crypto, I just said that requiring an ID with digital purchases is becoming more and more common
But, you are mischaracterizing me, I AM a South American migrant that did scape and has benefited from crypto for what little economic interaction I do have with my ruinous home country
On the same idea, I don't need/care for perfect opsec because my threat model doesn't need it, what little I've directly bought with crypto has always been digital, so that's whay I've cared to figure out
Still details on income/transactions and such, all feel a bit unnecessary for public display, but a small percentage, and my first crypto came from mining and selling back when it wasn't taken that seriously specially not in Venezuela of all places
Brazil has an insane number of 'illegal' immigrants as well as people living in Favela who essentially don't even recognize the state, so I'm curious how that works. I assume it's something like the US where 10 illegals work under one social security number or a tax ID they've registered under the auspice of foreign controlled business.
Immigrants can request a CPF (the 'national ID'). I don't think being in the country 'legally' is a requirement, that isn't enforced the way it is in the US.
> people living in Favela who essentially don't even recognize the state
Most people get assigned an ID at birth. And people who live in a favela often have to work outside it, and they interact with most companies/state services that aren't utilities as usual.
Utilities OTOH often get MITM'd by militia/narcos these days though.
> I assume it's something like the US where 10 illegals work under one social security number or a tax ID
No need for anything fancy like that. The poorest people are willing to work based on verbal agreements, as the alternative is either starving, or hoping the public social security network has your back. And in case your employer requires one, that's a non-issue because, except for rare circumstances, everyone has one.
Digital banking, install payments and general smartphone usage is widely popular, including favelas.
> Can you tell me another way of buying something over the internet without tying the purchase to a government ID?
Isn't the real question more, does crypto actually allow you buy things without tying the purchase to a government ID?
I'm no expert but I regularly see articles about de-anonymisation. This leads me to be sceptical about claims to privacy, certainly given enough time and motivation by a government actor.
Go to any retailer and buy any in-demand product with the same market value as what you want to buy. Sell it on Craigslist or similar for cryptocurrency using a new wallet. Buy whatever you wanted to buy, never use that wallet again. Alternatively, mine the cryptocurrency yourself, again using a separate wallet for each purchase.
The deanonymization comes from tying any transaction performed by a particular wallet to your identity and thereby deanonymizing all of the other transactions. Which doesn't work if the wallet only ever has two transactions and neither of them are tied to your identity.
That's assuming traditional chains. Privacy coins also exist.
You don't need to be a government actor, even. You just need to understand what a graph is and be willing to patiently walk through the txns. It's not even that difficult. I have investigator friends who regularly do it as part of fraud investigations.
If you took all the crypto txns and grouped them by purchase, I would be willing to bet mortgage money that approximately nobody uses crypto to buy "subversive literature", out to many many decimal places of precision.
A person can study fashion extensively, under the best designers, they can understand tailoring and fit and have a phenomenal eye for color and texture.
Their vivid descriptions of what the Emperor could be wearing doesn't make said emperor any less nakey.
Yeah. I have a half-cynical/half-serious pet theory that a decent fraction of humanity has a broken theory of mind and thinks everyone has the same thought patterns they do. If it talks like me, it thinks like me.
Whenever the comment section takes a long hit and goes "but what is thinking, really" I get slightly more cynical about it lol
By now, it's pretty clear that LLMs implement abstract thinking - as do humans.
They don't think exactly like humans do - but they sure copy a lot of human thinking, and end up closer to it than just about anything that's not a human.
It isn't clear because they do none of that lol. They don't think.
It can kinda sorta look like thinking if you don't have a critical eye, but it really doesn't take much to break the illusion.
I really don't get this obsessive need to pretend your tools are alive. Y'all know when you watch YouTube that it's a trick and the tiny people on your screen don't live in your computer, right?
And how do you know that exactly? What is the source of that certainty? What makes you fully confident that a system that can write short stories and one-shot Python scripts and catch obscure pop culture references in text isn't "thinking" in any way?
The answer to that is the siren song of "AI effect".
Even admitting "we don't know" requires letting go of the idea that "thinking" must be exclusive to humans. And many are far too weak to do that.
That's because you haven't done that work and don't value it.
Women in agrarian societies do difficult manual labor like hauling water, milking, preserving food, tending livestock, laundry. Laundry before machines was backbreaking work nobody wanted to do, which is why the poor did it or women took in laundry if they needed money. If you had a hand free, you spun wool.
Also, they did all that while constantly pregnant or nursing, which is really hard on the body. Sure, women didn't have to go to war, but men didn't have to live with the fear that this year's baby might be the one that finally kills them.
That's not good enough when you're dealing with matters that can lead to civil or even criminal liability. Errors can be incredibly expensive to fix, if they can be fixed at all.
With a CPA or attorney, you at least have recourse if they screw up. You don't with LLMs.
reply