- One crowd is using to research algorithms, libraries, write boilerplate code, write test harnesses, introspect and integrate with APIs, do hands-off refactoring, and automating what would otherwise be boring tasks. They still think about architecture, best practices, understanding things in detail and the general shape of the solution is in their hands.
- Another crowd is curating prompts, setting up autonomous agents, creating tooling and guardrails around it, anything else but getting actually involved in how the sausage is made. They are working on meta tasks around the problem, in the hope the solution will write itself.
These two crowds are currently living in very different worlds, and getting very different results. We'll see what survives soon.
Are we including just technical people in these crowds?
Because there's a third crowd: everyone else/the general public that are standing up vibe coded websites and don't give a hoot how things work in the background or know as long as money is coming in. There are people that are using AI and thinking less and less causing their brains over the long term to become more inelastic.
We're in for a very, very painful future that will have mixed results. On one hand, you can boostrap things a lot quicker with less mental effort and it helps get up to speed without having to know some complex things (e.g. deep knowledge in coding). This can help us innovate on basic things faster, probably.
On the other ... people aren't going to learn. If something breaks in that state where they don't know how something works, what, we're just going to ask another AI to fix it? I don't know how I feel or think about that. On a long enough timeline, there are people that won't know how any of this was designed in the first place.
That's the world we actually live in. And that's what will survive despite crowd 1 and 2 that you mentioned above.
I’m working with an ideas guy, no skills and no capital but for the first time can deliver code
He wants my tech expertise, his code is spaghetti, he is making all the mistakes, he is experiencing AI psychosis, his AI makes md files warning him that its all going to burn him which he forwards to me lackadaisically without reading
But can he sale? Yes
Its tempting for me to proselytize that he isnt using feature branches or project tickets or even deploying with committed code
But I bite my tongue and tell him to focus on the MVP since he wants to prompt Claude Code for 48 hour sessions without there being any indication of how other devs could contribute
Because he has clients that wants what he described, and because he has no capital I get a huge cut of that
I’m fine with that, I’ll clean up the project very quickly
It could be okay if it wasn't for the fact that they have a high esteem of themselves and don't learn shit.
The first POC they vibecoded and managed to sell was a pile of trash that we couldn't realistically deploy as it.
So we spent two weeks to make it something decent enough to have the minimum confidence of the thing being reliable and safe enough to reach staging.
Two weeks during they told us daily how we were slowing everything down.
After that we spent an hour explaining them how making something works on their computer, without tests, without thinking about edge cases... was not the same thing as deploying it and releasing it to actual users.
We agreed on, next time, asking us for an estimate on how much it would take to move the POC to something that could be released, before signing on any engagement with a customer.
Well, last week they came with a new deal they just signed on a completely new POC. Customer was expecting it for yesterday.
To make it work, we have to setup a VPN between our infra and customer infra.
Their internal process make it impossible to have this under a month.
Now, my guy and the customer, are mad at me because I can't deliver.
And the first POC? Customer wants new features. My guy don't want to deal with it because it's not their job.
For good reasons, they don't have access to our infra.
I wasn't involved, but I heard last year they tried to vibedeploy, provisioning an AWS account in their own name with their own credit card.
I let you imagine what happened and why they haven't retried this since.
You probably saw this, but future readers won't: just last week, anyone could steal any Instagram account by telling the AI support chatbox they'd lost access to their email address.
I know LLMs are getting the majority of the attention of late, but there are others of us training and using AI to…
…reads MRIs and video to detect cancer
…analyze genomics for early target discovery
…assisting surgeons
…folding proteins
And the list goes on in other fields as well. Just hoping the recent AI counterculture doesn’t stigmatize other uses of AI.
Am I a fan of Claude code? Not particularly, but I have used it on occasion. And I’ll never understand someone using an LLM to write anything (especially a comment on a site like HN) intended for consumption by other people. Not because I think it’s subpar, but because the point - IMO - is to make human connections, learn, teach, and debate. That’s hard (impossible?) to do if you’re just typing a 30 second prompt and then copy/pasting the output.
This problem only exists because of the marketing move to call anything even slightly ML related "AI".
I see it as a double edged sword. People that want the category of AI to succeed can claim a victory when someone uses it to approximate protien folding and invent new drugs. But that also means the entire field is constantly being dragged down by low quality vibe coded sites, slop videos on social media, whatever horrific thing Grok is doing this week, etc.
>This problem only exists because of the marketing move to call anything even slightly ML related "AI".
We need to remember what "Artificial Intelligence" actually means. It refers to the field of research starting in the 1950's developing algorithms related to combinatorial search, planning, and reasoning. Machine Learning isn't AI in the sci-fi movie sense, but it's among the topics you'll find in a textbook like Russell and Norvig.
A problem like protein folding isn't tangentially related to AI, it's at the heart of the kinds of problems the field has been trying to tackle for decades. Yet when there are legitimate breakthroughs, people deride it as "not real AI."
That isn't how we break up the field of AI. ML are the algorithms that are based on statistics and other numeric methods. The other algorithms are based on logic and perhaps some philosophical methodologies. We don't really have a name for that second group, its just part of AI. Then there is Reinforcement Learning which is a sub-field of ML and incorporates a Pavlovian methodology.
AI is the sum of all of these groups. Also, the "not real AI" thing is more about not real AGI. That's a very different target.
I'm not sure if your comment disagrees with mine. I think we agree that machine learning is a sub-field of artificial intelligence. Evidenced by the fact that Russell and Norvig, the most authoritative textbook on the subject, includes multiple chapters on ML.
AGI isn't a well-defined concept. So when people say something isn't "real AI" because it's not AGI, I can't take them seriously because they're implying that everything the field has worked on for the past 70 years isn't real AI.
For me personally I am vaguely indifferent to programmers using LLMs to make more shitty code. My worry is the second and third order effects
For work currently, as an SRE, I'm being asked to maintain and look after slop as if its properly built and instrumented. Our platform has clear rules and conventions, and AI isn't following those.
For the wider world, I fucking hate that image/video generation is evaporating what is "real". For memes sure its great, but for bad actors it gives a brilliant way to say "its AI wasn't me" and then the debate moves away from "did person do bad thing" to "is it wrong to say that things are AI?"
I also worry about the debasement of value of human work. Looking at history, say of the weavers, it didn't work out to well for them when the powered loom came along.
This is the opposite of a loom. With a loom you can very easily see that what it makes is exactly like cloth made by hand, except it's more uniform and faster. There also wasn't this weird drive to drive out the heretics so the AI messiah may dwell among us, this desire to put all eggs in one basket, instead of welcoming competition and a control group.
I would suggest that you are looking at the product, not the effect on the workers.
The powered loom produced more uniform, much cheaper fabric. It wasn't colourful or particularly flamboyant. It took a lot more work to get patterns (its where punch card come from)
But thats not the point. Powered looms meant that cottage industry that employed people close to sources of production (ie cotton/calico in india and wool in england) were thrown out on their arses. The majority lived on rented land/housing. Couldn't pay the rent and were kicked out into the loving arms of the poor laws. Lots of people had to re-train, the rest went begging. Combine that with agricultural reform, meant that Lots of people moved into slums in the towns, where they worked much longer, unsafer hours.
> I would suggest that you are looking at the product, not the effect on the workers.
Sure, I was in that comment. What you said is true, too, and given how much of this is driven by greed and not need, I would even say it's more important.
But the effect on the product matters, too, of course. Not just quality and bloat wise, but in a way it's like saying we no longer need to know how to read and write, we all get a butler and just ask them what a text says, or to write letters for us. Because the butler is smarter than any human, and it's so convenient. I'm ironically "with the Catholic Church" (not really, but you know) on this one because I see people/companies who want to be like the Catholic Church in the middle ages, and even without those, how sheer laziness can plunge us into some idiocratic abyss.
The industrial loom didn't make cloth the same as made by hand. All machine made cloth is worst than the best weavers could make.
The cloth was just okay, and could be made faster and at scale and then sold.
Now you can't buy cloth to the standard it used to be made.
> I also worry about the debasement of value of human work. Looking at history, say of the weavers, it didn't work out to well for them when the powered loom came along.
They eventually moved on to other things, because that was the only option. And the world is better with the power loom. It's scary but we still have to embrace that eventually pretty much all valuable labour will be automated, and by then our society and economy needs to have been restructured for supporting humans providing 0 economic value.
>They eventually moved on to other things, because that was the only option.
Who is they? The majority of British textile workers experienced destitute conditions following industrialization.
That's one of my two gripes with AI:
1) It's posed to take over knowledge work, and yet our societies have no safety nets for the millions of knowledge workers.
2) It promotes superficial understanding. It sounds so convincing and compresses complex topics into a few messages, leaving users thinking they know more than they actually do.
1) Yes, that's a serious issues that needs to be addressed, but many are too high on the status quo, because freedom and meritocracy and self-determination and all that. Rather watch the world burn than give up the ability to be able to earn and own more than the next person.
2) That's up to users. Those who only want superficial understanding will get that, and those who want deep understanding will question more and ask for citations so they can verify.
> They eventually moved on to other things, because that was the only option.
Yes, but then whole swathes of the English countryside (and then the Indian countryside) was plunged into destitution for generations and it took rebellions, massacres and revolutions to get something like comfortable living.
Yeah there was no system in place then to ensure those who were left without means to survive would be able to survive. Shouldn't be the case with the knowledge of that history and all this time we have to actually change things. We quite literally know that it's a matter of only a few years now before the vast amount of knowledge work (at least) is fully automated away; governments should be making changes to make the economic transition more smooth. It'd be highly irresponsible to not do so, and I dread that most will be irresponsible.
>It'd be highly irresponsible to not do so, and I dread that most will be irresponsible.
This is why I can't stand people who talk about embracing AI. You suggest that society needs to adjust to AI, but then turn around and admit you doubt it'll happen.
You can stand on a beach and shout at a tsunami, and maybe if you have a stick and time it right, get a hit in when it arrives. But I don't see how that helps anything. The only thing I see is to do as much as possible to prepare for when it hits, to help as many as possible to survive the onslaught.
Like a tsunami, preventing the arrival of AI is essentially impossible, unless you nuke all the data centers in the US and China and kill all the scientists and engineers with even a modicum of interest in working on it. Unlike a tsunami, AI is actually pretty useful: there are many already getting value from it in it's infancy, and its value will only keep on increasing with every release. But it's going to turn the world upside-down in the process and there are many too afraid of change. That's just how people are.
Forget about "money", and the economy in general as you know it. All that's going out the window because it's rooted in wage labour aka jobs, which will ultimately disappear. Those who don't like the actually-effective-but-highly-disliked system... well things are just going to suck for them.
A lot of people doing the latter camp are people with the knowledge of the former camp, and who are sufficiently happy with the speed and guard rails to no longer worry about "molding the solution in their hands"
I'm not speaking from personal experience, this is what friends are doing at their startups
But I am not surprised at all, because the building blocks of major applications are all out there as boilerplate code - heck half the time AWS has the example you need for you, assuming you know what you want to stitch together and why
If you know the major AWS tool chains and how and why to use them and how to design a product in microservices, then theoretically Claude has no idea what the whole shebang is up to but happily writes all the parts
For me it is hard to invest too much time into being in first crowd.
This space is moving too fast for me and I have current job to do that is paying my bills.
I can invest time to watch/follow people from the first crowd.
But no one is going to give you a medal and it is not a „better crowd”.
I might need to pay for this expertise some day, but I guess it will be OpenAI or Anthropic that takes my money just like so far all the advances were introduced to frontier models or their own tooling.
- One crowd is using to research algorithms, libraries, write boilerplate code, write test harnesses, introspect and integrate with APIs, do hands-off refactoring, and automating what would otherwise be boring tasks. They still think about architecture, best practices, understanding things in detail and the general shape of the solution is in their hands.
- Another crowd is curating prompts, setting up autonomous agents, creating tooling and guardrails around it, anything else but getting actually involved in how the sausage is made. They are working on meta tasks around the problem, in the hope the solution will write itself.
These two crowds are currently living in very different worlds, and getting very different results. We'll see what survives soon.