Currently I am using the regular Claude extension with VS Code, but one of my friends is working on a new harness called NodeCode that doesn't require regular compaction, so I'll be trying it out as soon as I can get some time to do it: https://nodecode.ai/
I am always doing everything alone, so I am liking the LLMs a lot. My own code is very good, but nobody but myself cared about that. Even the one job that I got which allowed me to gather the capital needed to move into trading feels like complete luck. This is a new golden age for entrepreneurs and apocalypse for wage workers. Nobody whose status depends on his own productivity would complain about it going up sharply.
Probably the LLMs saved me because deep down I knew that my own power by itself wouldn't be enough to let me do what I really want, but right now the future does seem bright.
I think you’re heavily miscalibrated. Anyone can use LLMs, the floor has risen drastically. What supposed productivity you achieve, every one else can because there’s 0 skill to using LLMs, this is probably a much darker future ahead for you than the rest, you’ll have to keep constantly agent maxing to try and outcompete millions of other humans doing the same.
LLMs are constantly driving off the cliff. Constantly. They require some actual agency steering the ship.
This idea that there is 0 skill to using them is the opposite: it requires far more skill because the human has to manage the whole building all at once rather than focusing on putting up one beam at a time, one rafter at a time. Your responsibility is totally different. It has to scale far higher!
Many devs have had strong leads that do the ticket work, that figure out what to do. It's unclear what role there is for those folks now.
I'm not sure this is credible. Using AI very much seems to be a skill with its own craft and best-practices. I could imagine a similar argument was made by painters about photographers. Yet, we believe good photography takes skill.
Anyone can do work on their own car. It's not especially hard.
But a professional working on their own car knows tricks and has tools / ways of working that a regular person doesn't have access to.
You can tell the difference in the result if you know where to look.
Sometimes it matters: the way you change some components matters - and while you're there with half the engine bay cleared, you can fix a few other things that usually fail in this specific model.
Sometimes it doesn't: oil change is hard to fuck up, but not impossible.
The the things I've made with agents aren't hard to do for anyone, but I can claim that I can pick up any of my projects 6 months from now and confidently add features without worrying about stuff breaking or the agent doing crazy crap - because I build the limits into my projects by default.
People are afraid to do work on cars (and mains wiring, and plumbing) because even though the work is relatively simple the consequences for fucking up are so severe (severe damage, injury, death, etc). That's most of the reason they pay someone else to do it.
For the vast majority of software that isn't the case at all.
Most people don't understand computers and do not even know what is dangerous and what is not. They won't suddenly become vibe-programmers lest "AI deletes all of the personal photos and will send them somewhere".
No need, I still have vibe coded stuff from the times before monthly plans for Claude and Codex over a year ago. I can jump into most of them and just start adding stuff without any major issues - they have better test coverage and CI than many projects at my day job :)
It's surprisingly easy to do that shit properly when budgets and time aren't an issue.
If not all agents are equal and require calibration, then it seems like people can specialize, so you're only competing with other specialists in that area (and not necessarily trying to out-compete).
lol wat? there is a very clear gap in output between what a good swe can produce with ai versus a bad one, let alone a random person who doesn't even know what javascript is.
If I were an artist, I wouldn't really care that a machine took over the role of drawing lines on paper, and as a programmer, I don't care that the machines are taking the role of writing code.
The thing that actually matters is understanding the problem domain, and thinking about and visualizing the program I am trying to make, so I am happy to be doing less drudge work so I can focus on what really matters.
I've been doing quant research for the past year, and the amount of time (and tedium) LLMs have saved me is huge.
LLMs aren't even AI, if this amount of disruption is causing difficulty, having actual AI will give these people breakdowns.
"I am not , but if I were, I'd have about it." isn't a terribly convincing argument. Opinions (good ones, at least) are shaped by experience - what experience is there to shape your opinion on AI affecting art when you aren't an artist?
Are those important questions that need to be answered? Rather, you should be asking whether you could use backprop and NNs to create something other than text predictors. What have LLMs changed about your conception of intelligence?
As far as I can see, none of the hard problems of intelligence that we had a decade ago got touched. What LLMs ultimately changed is the way I see language. I've underestimated its power, and I will admit, the companies making these models went far further than I could have imagined with them.
But intelligence and AI have a high hurdle to clear, and it won't be done with LLMs.
Revealing the secrets of intelligence will involve new concepts and understanding that that we don't have now. LLMs will be helpful with the plumbing and as replacement for natural evolution, but not so much natural intelligence.
Consider that maybe AI should mean AGI. That was what its original meaning was supposed to be, and calling generative models AI is watering down the original term and accepting the frame of the LLM makers.
What's really disappointing about current LLM models isn't that they aren't useful, but that they really tell us much about intelligence at all. The techniques used to make LLMs aren't generalizable at all, and I cannot use them to solve problems of interest in the real world.
That wouldn't be the case with real AI. I want better methods and algorithms that could live up to the expectations, and I am sure we'll get them at some point even if we don't have them now.
The history of AI is people deciding that it really means AGI but then getting enamored with some method of the time and calling it AI, before going back to the original meaning. We're just living through the latest iteration of this where LLMs mean AI, and we're forced to invent the term AGI for what used to be AI.
I'll admit that to the people who don't know how LLMs work, which includes most investors, the lie seems very plausible so of course the CEOs running the companies are going to frame it as such.
AI has included everything from neural nets to video game NPCs for as long as I'm aware of. What else is the broad umbrella term? We have to have one. And for anything else, you need to be specific about its properties anyway. In particular, usage of "real AI" is nonsense without a more precise definition.
It isn't nonsense. If AI was such a broad umbrella term, why would companies be tripping over themselves to use it? If it meant video game NPC level of intelligence, nobody would be in a hurry to adopt it. It's very clear that when Nvidia or OpenAI use the term AI, they really want you to think these LLMs are actual human level intelligence. That is why they are doing it.
When I use nonsense like "real AI" I think you very well know what I mean, and don't find it in any way confusing or ambiguous.
Everyone thinks they know what "real AI" means, but they usually mean some mishmash of "general intelligence", "human-like intelligence", and "deserves moral rights", none of which firmly imply each other but which they'll often freely equivocate between. If you mean something more specific than this semantic disaster zone, say that instead.
AI company marketing usage only indicates they're trying to take advantage of this confusion. They're bullshit artists who are trying to make sales, not communicate.
Fun story, but I really wish OpenAI got its act together and started making actual AI breakthroughs instead of funneling compute into LLMs. I'd really like some new algorithms to get me excited about the field again. Kuddos to them for making LLMs really useful, but this is not the ride I wanted to get on.
This is the internet, so I cannot tell at all whether you're being sarcastic or not. In my view, what LLMs should get us to reconsider isn't whether there is more to intelligence, but whether there is more to language. It's the latter which I underestimated.
reply