I've posited for a while now that the people who find spicy autocomplete to be exciting are the people who can't really do what it does.
I played with Image Playground last year some time. It was really fun. You know why? I can't draw, and I can't paint, to save my life. It's letting me do something I can't do well/at all on my own.
Using an LLM to do something I can do, with the caveat that it's pretty mediocre at the task, and needs to be constantly monitored to check it isn't doing stupid things? If I wanted that I'd just get an intern and watch them copy crappy examples from StackOverflow all day.
The same logic explains the use of LLM's to write emails/other long form text.
It makes accessible something that people otherwise cannot do well. Go look at submissions on community writing sites. The people who write because they're good at it, are adamant they don't use an LLM.
People use LLM's to do things they're otherwise not able to do. I will die on this hill.
"I've posited for a while now" and you post the most lukewarm and outdated take like it's an enlightenment. I've been coding for 20 years and can very well do everything the AI does, and so can all devs I know. We use it because it amplifies us, not because we couldn't otherwise. You've chosen a very ridiculous hill to die on.
Is your argument that there is no imaginable situation where someone who was competent at software development could find use for a semi-automated tool for writing software?
That would imply that either the person in question has infinite time, or has access to all software that could ever be of utility to them, which seems unlikely.
.... that an IDE providing a suggestion about what comes next as you type is not new, and the entire basis of how an LLM works is "what word probably comes next".
I'd have thought someone who's so enamoured with the tech would have at least a basic understanding of how it works.
Indeed. To be honest, I think everyone on HN is aware of how LLMs work at this point, it’s not actually adding a great deal to the discussion to keep going on about autocomplete or ‘stochastic parrots’.
At this point if someone calls it auto complete they can be written off as a Luddite with nothing valuable to say. The irony being they themselves are being a stochastic parrot by parroting the jargon other people say about llms.
Initially I wanted to write more but I can boil it down to taste and context mismatch. By that I mean some people see LLM output as tasteless or kitsch (which I ascribe to generally) and another set of people (though sometimes overlapping more often than not) hold disdain or at the very least look funny at heavy LLM users like gym-goers would look at someone in the middle of the gym loudly suggesting using a dolly or forklift instead of barbell training.
So yeah, I guess the value of doodles has shot up simply because of optics.
Somewhere else in this comment section someone tried to broaden the definition of nerd so much so that pretty much anybody who is a consummate professional is also a nerd. The hill I will die on is that people don't actually dislike all this new AI stuff but more so the attitude of people heavily invested in it.
And to add another data point regarding your hill my drawing/painting moment was NLP stuff. Now if I want to do (rudimentary) sentiment analysis or keyword extraction I can lean on a local LLM. Yet I don't go around yelling Snowball (I think?) is obsolete.
While you are dying on a hill, with the help of LLMs, I'm shipping quality software and features to my customers at a pace I haven't been able to before. And no, not some nextjs slop. If you are letting your LLM look at StackOverflow, you are doing it wrong - it needs to be grounding in your stacks official docs and any other style/rules you prefer wired with other tooling like linting/formatting, duplication checking, etc. And yes, you have to constantly monitor the output and review every line of code - but it's still faster and if managed correctly, produces better code and (this is the hill I will die on) better test suites and documentation than I would have written.
I'd be more general and say it needs verification to guide it, and narrowed scope so it doesn't wander off. How those get provided can vary. While I can do what I'm asking it to do, and have so many times that I don't want to anymore, I can't do it as fast as it can. But as someone said, it is stupid really fast. The bottleneck is now me slowing down this intern who thinks fast by stopping it to redirect it when it does bad things. The more pre prompting and context and verification tools I give it the less I have to do that, so the faster it goes. Then I get to solve the parts of the problem I haven't done until its boring.
More LOC committed per day is probably the only one that's guaranteed when you let spicy autocomplete take the wheel.
I don't think it's at all possible to reason about the other more meaningful metrics in software development, because we simply don't have the context of what each human is working on, and as with the WYSIWYG fad of 3 decades ago, "success" is generally self-reported, by people who don't know what they don't know, and thus they don't know what spicy autocomplete is getting woefully wrong.
"But it {compiles,runs,etc}" isn't a meaningful metric when a large portion of the code in question is dynamic/loosely typed in a non-compiled language (JavaScript, Python, Ruby, PHP, etc).
If you are on the right team with the right professionals you can measure. when we first started using LLMs we decided to run the same process as if they did not exist, same sprint planning meetings, same estimation. we did this for 6 months and saw roughly 55% increase in output compared to pre-LLM usage. there are biases in what were tried to achieve, it is not easy to estimate something will take XX hours when you know some portion (for example writing documentation or portions of the test coverage) you won’t have to write but we did our best. after we convinced ourselves of productivity gains we stopped doing this. saying you can’t measure something is typical SWE BS like “we can’t estimate” and the other lies we were able to convince everyone off successfully
Your views might carry more weight if the crux of your rebuttal wasn't manufactured outrage that I used a laughably accurate nickname for a type of software.
Hey, good question. The cli tool itself is more or less just modifying files within the repo so you should be good to go with Mercurial or any other vcs. To smooth things out for AI agents the skill files and descriptions would need to be migrated for the respective vcs of course.
Actually yes and no. The current implementation is more like a monitor of things going on. I used it to visualize the current state of a project that I let AI agents work on. So the board is read-only, and I think that's mostly fine for now. Porting this read-only view to HTML and put a little http/websocket service in there would be easy on the one hand, but break the TUI setting somewhat. Feel free to create an issue for it, though.
It's wild that these companies have convinced you to pay to be a beta (at best; arguably much of this is pre alpha quality shit) tester and you're perfectly happy with that scenario.
The lengths people go to not to solve the actual issue never cease to amaze me.
Currently "I don't know how to configure xdebug so I wrote a convoluted alternative" is tied for top place with "I don't know how MySQL sockets work so I rewrote the entire thing to use Mongodb".
I played with Image Playground last year some time. It was really fun. You know why? I can't draw, and I can't paint, to save my life. It's letting me do something I can't do well/at all on my own.
Using an LLM to do something I can do, with the caveat that it's pretty mediocre at the task, and needs to be constantly monitored to check it isn't doing stupid things? If I wanted that I'd just get an intern and watch them copy crappy examples from StackOverflow all day.
The same logic explains the use of LLM's to write emails/other long form text.
It makes accessible something that people otherwise cannot do well. Go look at submissions on community writing sites. The people who write because they're good at it, are adamant they don't use an LLM.
People use LLM's to do things they're otherwise not able to do. I will die on this hill.
reply