I've just done my first almost fully vibe coded hobby project from start to near completion, a village history website with a taxonomy, and it's taken so much poking and prodding and cajoling to get the software to do exactly what I want it to do. Having built plenty of production stuff, I know what I want it to look like and the data model was really clear, yet even trying every trick in the book to constrain them, I just found the LLMs went off and did totally random things, particularly as the project got further from the start.
Maybe there'll be an enormous leap again but I just don't quite see the jump to how this gets you to 'industrial' software. It made it a lot faster, don't get me wrong, but you still needed the captain driving the ship.
But were there ever 100 "rowers"? In this case, the commenter would have developed the website him- or herself instead of using AI. And it would have taken a little longer but probably been higher quality. In my experience, most developers are already capable captains and most of their job is "captaining." One of their main complaints is managers who treat them like rowers. AI just shifts what it means to captain?
Honestly I just wonder if everyone is burnt out so they still want to do their side projects but don't have much energy left over for the passion they had before. So the bar for "good enough" just lowers
Right but in this case, what might have taken me 2-3 days initially took me a day still with many many frustrating “no, I said I don’t want you to edit that file in that way” moments. So it’s faster but annoying.
To me it feels like when doing it’s like when you’re pair programming (which is quite intense anyway) with a frustrating newby who can program but misses the big picture no matter how many times you try. It works better for certain types of software - web it’s great, I have had much worse results when doing data pipeline stuff at work.
>They’re not all going to get their own boat and captain hat
Why not? Anyone can load up Claude code and start trial and erroring until they get something that works and has similar reliability to accepted software … what is the stat about 1 bug per 10 lines of code on average?
I am meeting a lot of non coders telling me about their projects they are getting AI to do for them, stuff to help land title something or other, stuff to work on avalanche forecast, whatever their area of expertise they are unchained and writing programs using AI that they couldn’t before.
Anyone can load up Claude code and start trial and erroring until they get something that works and has similar reliability to accepted software..
You still need to understand the code that AI is generating to fix the problems that you can't vibe a solution to. You still need to understand the process of developing software to know when something isn't working even if it looks like it is. You still need other people to trust the software that you created. None of those things comes naturally to vibe coders. They're essentially teaching themselves software engineering in a very back-to-front way.
Everybody learned somehow. I wonder what % of programmers actually have relevant training and education in programming vs just taught themselves with online resources.
Maybe the amateurs aren’t going to be writing a new distributed database but CRUD apps must be easier than ever
The studies I have seen show that AI written software is 70% more buggy than human written code [1]. I am curious where you get your data on AI code having “a similar reliability to accepted software”?
Everyone wants to be but I don’t think there will be enough seats. There are people doing boilerplate and simple CRUD stuff - they’re not going to switch to farming. Reckon this will lead to more competition for same number of senior seats
I work for a US based company from the UK. I'd be quite reluctant mostly for the hassle factor - international travel is a pain, now I have a family I don't want anything other than a super easy trip when travelling for work, I don't usually have to travel, so if it's going to be a nightmare I'd rather not go. I've heard of colleagues having to take burner phones to China and stuff in the past and noped out of that, and it feels like it's not far off that for the US these days, so while I quite like the US in general it does put me off enormously.
On another note, we're having our next international team get together in Canada rather than SF. Make of that what you will.
The amount of unnecessary travel these large corporations do is insane. All of it subsidized by taxpayers as well just to pollute and destroy the environment for imaginary profits that inflate away within 5 years.
Well folks, there you have it, @onebigtime is the ultimate decider on what "unnecessary travel" is :).
On a serious note, being able to go face to face is sometimes a huge win, and can really help hammer through sticky problems, to the point where companies that are downright cheapskates on expenses still see enough value to justify paying for travel. I promise you, if it was really unnecessary, they wouldn't spend a dime they didn't need to.
It's more that half my team is 9 hour time difference from me. When we meet up we can get a lot of discussions done without anyone having to work insane hours.
> It’s as if every researcher in this field is getting high on the small amount of power they have from denying others access to their results. I’ve never been as unimpressed by scientists as I have been in the past five years or so.
This is absolutely nothing new. With experimental things, it's non uncommon for a lab to develop a new technique and omit slight but important details to give them a competitive advantage. Similarly in the simulation/modelling space it's been common for years for researchers to not publish their research software. There's been a lot of lobbying on that side by groups such as the Software Sustainability Institute and Research Software Engineer organisations like RSE UK and RSE US, but there's a lot of researchers that just think that they shouldn't have to do it, even when publicly funded.
> With experimental things, it's non uncommon for a lab to develop a new technique and omit slight but important details to give them a competitive advantage.
Yes, to give them a competitive advantage. Not to LARP as morality police.
There’s a big difference between the two. I take greed over self-righteousness any day.
I’ve heard people say that they’re not going to release their software because people wouldn’t know how to use it! I’m not sure the motivation really matters more than the end result though.
There's always been this lesson with CI/CD - don't couple yourself to a specific product. If you do, you're gonna get screwed eventually. It happened with TravisCI, CircleCI, now it's happening with GitHub. The business model only makes sense if you can charge for it, those charges are only ever going to go up.
Interested in the Lost World: Jurassic park variants. I was 5 when that came out in 1997 and it was an odd release since barely anyone had a Megadrive anymore but we had one and my parents wouldn’t buy me a PlayStation, so they relented and bought me that as I loved dinosaurs. It is honestly such a great game and was really underappreciated since it was so late in the release cycle for that console.
> is that idea-guys' ideas crack when meeting reality.
This is definitely the hardest nut to crack. I worked on a product a while ago that needed to track maintenance periods on equipment quite carefully and then use that to filter data to provide insights about how it was performing. The 'user story' was light on detail. As we got into it, there were tons of questions about how to deal with source data that was often inconsistent or patchy, time zones became an issue because much of the data we received wasn't matched correctly to their local time (customers fault not ours) and our ideas guy just couldn't deal with it - "make it work" - when ultimately they were business questions that needed answering, not just pure software tasks. AI is so sycophantic it'd just go off and write something.
reply