I've been using gstack for the last few days, and will probably keep it in my skill toolkit. There's a lot of things I like. It maps closely to skills I've made for myself.
First, I appreciate how he implemented auto-update. Not sure if that pattern is original, but I've been solving it in a different-but-worse way for a similar project. NOT a fan of how it's being used to present articles on Garry's List. I like the site, but that's a totally different lane.
The skills are great for upleveling plans. Claude in particular has a way of generating plans with huge blind spots. I've learned to pay close attention to plans to avoid getting burned, and the plan skills do a fair job at helping catch gaps so I don't have to ralph-wiggum later. I don't find the CEO skill terribly effective, but I do like the role it plays at finding delighters for features. This is also where I think my original prompting tends to be strong, which could be why it doesn't appear to have a huge impact like the other skills.
I think the design skills are great and I like the direction they're going. DESIGN.md needs to become a standard practice. I think it's done a great job at helping with design consistency and building UIs that don't feel like slop. This general approach will probably challenge lots of design-focused coding tools.
The approach to using the browser is superior to Claude's built-in extension in pretty much every way (except cookie management). It's worth it for that alone.
For people who don't understand this...think of each skill like a phase of the SDLC. The actual content, over time, will probably become bespoke to how your team builds software, but the steps themselves are all pretty much the same. All of this is still early days, so YMMV using these specific skills, but I like the philosophy.
I find that I don't have a lot of sympathy for people angry at this type of behavior, even though I share the disdain for someone else's AI output. The people doing this kind of thing are not the kind of people to be reading this manifesto. We've been creating bait content for a long time, and humans have never been given the tools to manage this in any sophisticated fashion. The internet was not a bastion of high quality content or discourse pre-AI. We need better tools as content consumers to filter content. Ironically, AI is what may actually make this possible.
I do find it interesting that people don't mind AI content, as long it's "their AI." The moment someone thinks it's someone else's AI output, the reaction is visceral...like they're being hoodwinked somehow.
I suspect the endgame of this is probably the fulfillment of Dead Internet Theory, where it's just AI creating content and AI browsing the internet for content, and users will never engage with it directly. That person who spent 10 seconds getting AI to write something will be consumed by AI as well, only to be surfaced to you when you ask the AI to summon and summarize.
And if that fills people with horror at the inefficiency of it all, well, like I said, it isn't like the internet was a bastion of efficiency before. We smiled and laughed for years that all of this technology and power is just being used to share cat videos.
> I do find it interesting that people don't mind AI content, as long it's "their AI." The moment someone thinks it's someone else's AI output, the reaction is visceral.
Isn't it obvious? If I'd wanted to see AI response to my question, I'd ask it myself (maybe I already did). If I'm asking humans, I want to see human responses. I eat fast-food sometimes, but if I was served a Big Mac at a sit down restaurant I'd be properly upset.
> If I'm asking humans, I want to see human responses
I find this fascinating, honestly. It shouldn't matter as long as it addresses your ask, yet it does. I also wish I could filter social media on "it's not X. It's Y"
Because it's probably not actually about the content but the sense of connection. People want to feel like they're connecting to people. That they're being worthy of someone's else's time and attention.
And if that's what people are seeking, slack and social media are probably not the platforms for it (and, arguably, never were).
> It shouldn't matter as long as it addresses your ask, yet it does.
If the LLM output is concise and efficient I don’t actually care that it’s LLM output.
My problem is that much of the LLM prose feels like someone took their half-baked idea and asked the LLM to put a veneer of quality writing on top of it. Then you waste your time reading it to parse out the half-baked idea hiding among the wall of text.
If a person has a shitty idea that sounds good, they start writing about it. If they exercise some care in their writing, the act of writing itself is enough to make them realize that their idea is shitty.
By the way, it happens to me all the time! Even just on HN, I’ve bailed halfway through writing a comment because I realized that I didn’t know what I was talking about, lol.
But an LLM will gladly take that shitty idea and expand it into a very plausible article/message/post, that seems reasonable if you don’t think very critically about it. And it’ll be done with such a high-seeming level of care that any human author would’ve been fact checking themselves the whole time.
So it forces the reader to think even more critically, rather than letting our subconscious try to judge authenticity of the writer through the language they use.
For example, someone who says “my WiFi is broken” when referring to the fact that their computer is dead, we can quickly judge them as “not an expert at computers”. But if they say that “my M.2 drive has gone bad”, we inherently assume they have some understanding. —- when the first person uses LLMs to write, they sound as informed as the second person even if they are completely clueless and wrong
In my case, it's because it doesn't address my ask, which is why I didn't ask an ai in the first place. The only person I know who does sloppypasta is my brother in law. I know he means well, but when I ask his opinion I want the perspective of someone in his demographic. If a generic ai response met my needs, I wouldn't be asking him.
> It shouldn't matter as long as it addresses your ask
But it doesn't? I'm more than capable of using Google and chatgpt myself. If I was looking for a machine generated answer to my question I would have already found it myself and never made the post in the first place. If I went to the effort of posting the question, it means that either the slop answer is not sufficient for some reason or that I want to hear from actual humans that have subjective experiences that an LLM cannot.
Posting an AI response verbatim basically says "I think you're too stupid to click a couple of buttons, so let me show you how it's done". I think it's very reasonable to get upset at the implication.
As an example of this, I am currently comparing two different models of Android e-readers, from a Chinese brand where the tech specs are all published but there aren't a lot of good comparative reviews. Plus, the specs like battery life are close to the same mAh, but for e-readers especially with Android optimization/drivers/etc make a gigantic difference.
So I have been Googling for "Reader X vs Reader Y review"(/comparison/etc) hoping to find Reddit comments or non-spam blog posts from people who actually own both to compare screen and battery life. I found a reddit thread comparing them directly and lo and behold the first comment is someone saying "I own both but honestly you could just ask ChatGPT for this". Fortunately a couple other people responded...
When I ask Gemini or ChatGPT, all I get is regurgitation of the tech specs (that are all mostly identical) plus summarized SEO spam reviews (that were probably written by another LLM based on those same tech specs) and it's totally unhelpful. So for this, I absolutely do NOT want an OpenClaw bot to respond as if they've physically used the devices and it would be actively enraging to learn a "helpful" comment "answering" the question was actually just an LLM impersonator.
I think it is reasonable, yes, but I don’t think it’s ever been reasonable to expect reasonableness on the internet. We have a difficult enough time showing each other decency.
I think it should matter. When you ask the AI something you are in a frame of mind, you have a specific context, the question also holds value and context that might completely change the parsing of the answer or at least the difficulty of it.
What I'm asking and the response from AI through an intermediary lose some context (the prompt), it's like the telephone game where the data becomes more and more distorted, that's why people don't have an issue with their own AI generated answers.
Another issue is that when I'm talking with someone and parsing through what they've said I'm considering them, as a person, taking all available context (some of this might happen unconsciously).
In any case I don't think there is an easy solution to the problem.
> shouldn't matter as long as it addresses your ask, yet it does. I also wish I could filter social media on "it's not X. It's Y”
The people copy-pasting slop almost never excerpt the relevant response. As a result, you get non-concise text you have to triple check. This is functionally useless to the point of being fine to skip.
We can tell by your fury that you’re a slop poster.
I don’t want a random person’s use of an AI to be slopped at me. I don’t know what they asked it, a lot of the words are made up, and I have to go through the effort of decoding it.
If I wanted an AI answer I would ask an AI. AI slop is made up. It’s like handing me a paste of google search results. It’s creating work for me.
But the internet has had slop long before AI. It's in the same class as clickbait. AI just made it worse and given the slop a distinct flavor. You can be furious about this if you want, but to me this seems like a waste of energy, which is the whole point of my original post.
We need better tools for managing our attention. Perhaps the effort of decoding it can be offloaded to AI.
Is that lame? Yes, but at least it's somewhat effective.
> People want to feel like they're connecting to people. That they're being worthy of someone's else's time and attention
They are achieving the exact opposite. I don't connect with the person who sends me slop. And they send me content that is a waste of my time and attention, because I have to vet it. Why would I trust someone - how can I ever connect with them - when the only thing I know about them is they take shortcuts?
>I find this fascinating, honestly. It shouldn't matter as long as it addresses your ask, yet it does. I also wish I could filter social media on "it's not X. It's Y". Because it's probably not actually about the content but the sense of connection.
It's also about the content. Generic slop I can get on demand from an LLM myself, vs a novel insight.
I am really into this approach of AI being used as a user-agent.
In particular, I've been thinking a lot about educational content, and what I'd love to ask educational providers for is not AI-generated content, but rather carefully human-built curricula offered in a structured manner, which my own AI could then use to create dynamic content for me.
> The moment someone thinks it's someone else's AI output, the reaction is visceral...like they're being hoodwinked somehow.
Reading AI generated prose, even if it’s my prompt, always gives me the same feeling as when I read a LinkedIn post: Like a simple concept was stretched into an unnecessarily long, formulaic format to trick the reader into thinking it was more than it was.
Everyone taking their scraps of thoughts and putting them into an LLM likes it because the output agrees with them. It’s flattering. But other people don’t like it because we have to read walls of text to absorb what should have been a couple of their scattered bullet points.
Just give me the bullet points. Don’t run it through the LLM expander. That just wastes my time.
Everybody wants to use LLMs to produce things and absolutely nobody wants to consume the things that LLMs produce and this is the fundamental reason this is all going to collapse unless we find a way for producers to pay consumers to consume their LLM output.
Gotta disagree. I've found several great new YouTube channels that clearly use ai for everything but the script writing. I assume it's an enthusiastic and smart niche expert who lacks the charisma to make videos in addition to doing the research. In very glad ai is filling in those people's weak spots.
How would you know it’s an enthusiastic and smart expert creating the content you’re consuming, do you have the subject matter expertise to judge that?
The odds are far higher it’s somebody who knows very little about anything but wants to make money from the gullible.
I suspect the voice is AI, the thumbnails definitely are, but unsure of the rest of the content. It's really interesting and the content in the video seems legit (looked up the "forgotten vegetables" and was surprised to find that we did forgo a lot of tasty things to satisfy the machine harvesters).
Guessing it's likely a phd student with too much time on their hand.
Great case in point. I think the whole "forgotten/banned garden plant" genre appeared post ai. Probably too difficult to get together enough visuals for a full length video on an obscure plant before.
One person's slop is another person's treasure, I guess. I've seen a lot of slop on Youtube, and I block the channels putting it out. It's pretty awful. They use AI narration that can't pronounce simple common phrases correctly. I'm not wasting my time with that garbage, I'd rather give views to actual people producing good content. I don't have time for slop.
>I do find it interesting that people don't mind AI content, as long it's "their AI." The moment someone thinks it's someone else's AI output, the reaction is visceral...like they're being hoodwinked somehow.
The problem is that getting an AI to answer a question is trivial. If I wanted to know what an AI has to say about the topic, I would just ask myself. Sending AI output has, as the author writes, the same connotation as sending a LMGTFY link. It does not provide me any value at all, I know how to write a question to an AI, just as I know how to use Google.
I'm starting to realise I might have completely misunderstood the whole lmgtfy thing. I thought it was a semi-rude way to call someone out for asking lazy questions instead of trying to find the answer themselves.
No, you completely had it correct. And sending an AI response to a question is the same semi-rude way to respond.
The context here is that the person logging the ticket (or asking the original question by using AI to do it) is the one who is ALSO being a lazy piece of shit, and deserves and equally lazy useless response in the form of a LMGTFY or AI response, because they were too lazy to actually think about their original query and spend time to craft a succinct but useful ticket/query.
I am sorry, but in what way is everyone letting the "We've been creating bait content for a long time" comment slide?
Did you even read the article? It is about person to person interactions. The three examples weer:
* Someone butting in to an ongoing discussion with a solution (but it's generic and misfitting AIslop)
* Someone being asked for their expertise and responding (but it's generic and misfitting AIslop)
* Someone comes with a problem thesis looking for help (but it's generic and misfitting AIslop)
The only one of these that existed prior to AI was the middle one, and the article very specifically calls out how transparent it used to be, because it had the shape of a google link.
The first one would be impossible because the person would have to either write an unhelpful response, and they wouldn't find the words at length. You could ignore them or pick it apart easily. The last one would be impossible unless if they were copy pasting from a large PDF, which would look nothing like a chat message.
What kind of workplace hellscape do you work on where people posting low effort bait on SLACK was the norm? The premise of this reply is entirely non-sensical.
>I find that I don't have a lot of sympathy for people angry at this type of behavior, even though I share the disdain for someone else's AI output. The people doing this kind of thing are not the kind of people to be reading this manifesto. We've been creating bait content for a long time, and humans have never been given the tools to manage this in any sophisticated fashion. The internet was not a bastion of high quality content or discourse pre-AI.
Which is irrelevant. TFA is talking about personal communication (and the examples are from a business setting).
And their concern is not the mere quality or lack thereof, but also its origin, and this is something new.
>I do find it interesting that people don't mind AI content, as long it's "their AI." The moment someone thinks it's someone else's AI output, the reaction is visceral...like they're being hoodwinked somehow.
No, many of us hate "our AI" content too, and wouldn't impose it to other people, same way we wouldn't fling shit at them.
> The internet was not a bastion of high quality content or discourse pre-AI.
I have read thousands upon thousands of pages of AI-related discourse, watched hundreds of videos since 2022, maybe even a thousand now on it. NEVER at any point in time did people opine for the "high quality" internet of before. They opined for the imperfect HUMAN internet of before. We are now seeing once pristine, curated corners of the internet being infected with sloppypasta.
This is quite a broad brush to paint the internet with. It's like saying The Earth is not a bastion of warzones/peaceful places to live. That is HIGHLY dependent on location.
Even before AI, the human social internet was loaded with bots and disingenuous actors. You want the imperfect human internet that is also pristine and curated. I've been socializing on the internet since 1994, and I feel fairly confident in sharing that this never existed, except in nostalgia.
If that's what you're pining for, you're going to have to find a highly protected part of the internet that is walled off from untrusted actors. However, that's always been the solution, and AI doesn't change that.
And since the foundation of the internet, the correct response to bots and disingenuous actors has been to a) ignore them b) ban them and c) ostracize then. We're talking about basic behaviors that have been understood since Usenet, something you surely should be aware of since you grew up in that era.
I think the difference was that before all this, there would be additional information embedded in the way a person types, or the way they'd written their code, that you could use to build a larger picture of the situation.
Right now it's as if everyone started wearing digital face masks that replaced their facial expressions with "better" ones. Sure, maybe everyone's faces weren't perfect before, but their expressions contained useful information.
I don't think that "it's more of the same" is a good way to think about it. The internet contained a lot of low-quality content, but even low-quality content used to be fairly expensive and time-consuming to produce. Further, you could immediately discern bottom-of-the-barrel content-farmed nonsense by the writing style alone. Now, LLMs make it practically free to generate unlimited amounts of slop that drowns out human-written stuff, and they can imitate the style hints we used to depend on for quick screening.
Yet how are the alternative ways of thinking about it better? Spending your time angry about what others can do? In any era, that’s a poor life philosophy.
The problem is the same as it has always been. Figure out how to use your time and attention effectively,
Is it possible to be critical without being angry? Are the only options here misplaced ire or total queiescent fatalism? Does the site here even seem excessively angry?
You're reading too much into my words if you think I'm suggesting we should take it in stride.
I think we should accept that trying to enforce social norms is a waste of time as that will only work on the politest part of the internet. Instead, focus on what you can control: better mechanisms for managing your attention and time.
Strategic, directed anger is an important component of using your time effectively. It sends a clear signal that certain kinds of behavior are unacceptable and people who'd like continued access to your time had best not engage in them. You shouldn't go around yelling at people every time you get a bit frustrated, but you should and I do express anger when someone signs their name to LLM-generated Slack responses.
> I do find it interesting that people don't mind AI content, as long it's "their AI." The moment someone thinks it's someone else's AI output, the reaction is visceral.
Somehow nobody that replied to you mentioned this. The issue is reciprocity. If I spend two hours manually researching and using my expertise to reply to a ticket, then 10 minutes later I get a novel-length AI reply in response...I now have no respect for the person replying with AI, because they can't even be bothered to spend a few minutes and summarize their "findings" and I suspect they didn't even read what their AI wrote. Especially in a professional setting, where you were hired for your (supposed) skillset, not your prompting skills.
If I'm sending out AI content, then sure, give me AI content in return.
The biggest vendor I work with uses "AI" for all email communications. It's like they use it to sanitize and corporate-speakify their communications, and I really hate it. They can never communicate like a real human being in email. But when we have actual zoom calls they speak like real humans, but in email it becomes so robotic. It's frustrating to feel like I'm speaking with a robot.
I acknowledge that those likely to copypaste slop aren't likely to find this article themselves, but I built the page to be shared or guide discussions around etiquette like nohello.net or dontasktoask.com. IMO a common understanding of AI etiquette would provide social pressure to halt some of these behaviors.
I honestly don't mind someone else's AI as long as I can trust it/them. One problem I have with sloppypasta specifically is that it reads as raw LLM output and the user isn't transparent about how they worked with the AI or what they verified. "ChatGPT says" isn't enough; for me to avoid inheriting a verification burden, I'd also need to understand what they were prompting for, if they iterated with the AI, and if/what/how they validated.
(the other problem is that dumping a multi-paragraph response in the midst of a chat thread is just obnoxious, but that's true even if its artisanal human-written text)
Good point: RTFM and (wall of slop) are two ways of telling someone that responding to them is not worth your time that are both ruder and more time-consuming than simply saying nothing. Explaining the culture of RTFM, i.e. "if there was any way you could possibly have found the answer otherwise, you should never have asked the question" to non-tech friends usually results in disbelief.
But the slop-wall is even worse, as it wastes the questioner's time in figuring out that they're just getting slop. At least RTFM is efficient.
I think you will find you will get farther by offloading this unpleasantness to an AI and open sourcing it rather than teaching etiquette to the internet, a place not known for its decency.
There’s a certain very satisfying force to turning something into a static website that you can point people at. The Internet equivalent of “don’t make me tap the sign”; especially in an era of AI-slop.
Because they are. It would be like if I bought some trinket off aliexpress and told you I made it by hand just for you. You wouldn't mind if you bought it yourself, but the fact that I lied about it to make it seem like I care is deceptive and immoral.
Sending someone AI generated text without disclosing so is incredibly offensive. It says you don't care about wasting the receivers time and don't care about honesty either.
I agree. I think it's interesting that, even if AI handles the conversation effectively, we're still repulsed.
I'm curious what will happen once AI generated text gets good enough that people can't tell the difference. Will we just assume everything is AI and remain suspicious, or will we stop caring?
My hunch is we'll all retreat to places of the internet where we can feel sure we're talking to real people and there are chains of trust. For example, I spend most of my time on discord servers where people are real life friends or friends of friends, and increasingly assume "public" internet to be AI default, and therefore use our own AI to browse and summarize for us.
I like to think that a lot of the current internet is just going to die and we will return to more in person interaction. And I think awareness of this is continuously rising. Terms like "chronically offline", talking about quitting social media, reducing phone use, etc are hot right now. But I'm still yet to see awareness and talk convert to action. People are as addicted to social media as they ever have been. We just widely recognize much like junk food and cigarettes where we know it's bad but keep doing it anyway.
What I'm fairly confident of is people will not just enjoy being deceived in to talking to bots. We have seen this before, companies and customer service platforms have been using templated messages to imitate real human conversation for a while. When you load a website and the Intercom chat box pops up with a message that looks like a real person from the company is trying to talk to you, initially it might have worked but very quickly you learn it's fake and tune it out.
Talking about bait, good job getting 42 responses on hacker news! Your opinions are controversial enough to draw out people who need to correct them, yet genuine enough to not be passed off as a troll and downvoted.
It's been pretty amusing seeing the total upvotes for my comment go up and down.
I wasn't expecting it to be so controversial. Reading and responding to many of the replies, I think many people are strawmanning me as being in support of AI slop.
Not the author, but did a LOT of research on this during my time at Disney while working on Disney+ prior to its launch.
This is, effectively, no different than a carousel of algorithm-recommended content. However, UX studies have found users reluctant to watch something recommended to them. It requires making an affirmative decision on time investment. Most people have the experience of a friend recommending a movie or book and still being reluctant to dive in.
The problem is very similar to dating apps, if you think about it. This is why Tinder's innovation on "swipe left/right" took off the way it did. In UX terms it's better to drop users into something and make the cognitive effort be choosing to get out of it rather than choosing to get into it. It's a big part of why TikTok works.
The reason this isn't more common in video apps has more to do with UX norms at this point. Another important thing I learned about streaming at Disney was that no one really cares how innovative the browsing experience is. They just want to watch Frozen. They're used to carousels now, and they're easy to program. This, I think, speaks more to your sensibilities.
I guess it's really not for me though. First thing I do is turn autoplay off, and I'd refuse to use a service that doesn't give me that option. OTOH, I do sometimes find it fun to hunt for good stuff among the recommendations.
I encourage everyone to RTFA and not just respond to the headline. This really is a glimpse into where the future is going.
I've been saying "the last job to be automated will be QA" and it feels more true every day. It's one thing to be a product engineer in this era. It's another to be working at the level the author is, where code needs to be verifiable. However, once people stop vibing apps and start vibing kernels, it really does fundamentally change the game.
I also have another saying: "any sufficiently advanced agent is indistinguishable from a DSL." I hadn't considered Lean in this equation, but I put these two ideas together and I feel like we're approaching some world where Lean eats the entire agentic framework stack and the entire operating system disappears.
If you're thinking about building something today that will still be relevant in 10 years, this is insightful.
This is a very strange statement. People don't always announce when they use AI for writing their software since it's a controversial topic. And it's a sliding scale. I'm pretty sure a large fraction of new software has some AI involved in its development.
I strongly agree with this. The only place where AI is uncontroversial is web search summaries.
The real blockers and time sinks were always bad/missing docs and examples. LLMs bridge that gap pretty well, and of course they do. That's what they're designed to be (language models), not an AGI!
I find it baffling how many workplaces are chasing perceived productivity gains that their customers will never notice instead of building out their next gen apps. Anyone who fails to modernize their UI/UX for the massive shift in accessibility about to happen with WebMCP will become irrelevant. Content presentation is so much higher value to the user. People expect things to be reliable and simple. Especially new users don't want your annoying onboarding flow and complicated menus and controls. They'll just find another app that gives them what they want faster.
Apps are a strange measure because there aren't really any new, groundbreaking ones. PCs and smartphones have mostly done what people have wanted them to do for a while.
There are plenty of ground breaking apps but they aren't making billions of advertising revenue, nor do they have large numbers. I honestly think torrent applications (and most peer to peer type of stuff) are very cool and very useful for small-medium groups but it'll never scale to a billion user thing.
Do agree it's a weird metric to have, but can't think of a better one outside of "business" but that still seems like a poor rubric because the vast majority of people care about things that aren't businesses and if this "life altering" technology basically amounts to creating digital slaves then maybe we as a species shouldn't explore the stars.
I think this might miss the point. We put off upgrading to an new RMM at work because I was able to hack together some dashboards in a couple days. It's not novel and does exactly what we need it to do, no more. We don't need to pay 1000's of dollars a month for the bloated Solarwinds stack. We aren't saving lives, we're saving PDFs so any arguments about 5 9s and maintainability are irrelevant. LLMs are going to give us on demand, one off software. I think the SaaS market is terrified right now because for decades they've gouged customers for continual bloat and lock in that now we can escape from. In a single day I was able to build an RMM that fits our needs exactly. We don't need to hire anyone to maintain it because it's simple, like most business applications should be, but SV needs to keep complicating their offerings with bloat to justify crazy monthly costs that should have been a one time purchase from the start. SV shot itself in the face with AI.
To be fair, Claude Code is vibe-coded. It's a terrible piece of software from an engineering (and often usability) standpoint, and the problems run deeper than just the choice of JavaScript. But it is good enough for people to get what they want out of it.
But also, based on what I have heard of their headcount, they are not necessarily saving any money by vibecoding it - it seems like their productivity per programmer is still well within the historical range.
That isn’t necessarily a hit against them - they make an LLM coding tool and they should absolutely be dogfooding it as hard as they can. They need to be the ones to figure out how to achieve this sought-after productivity boost. But so far it seems to me like AI coding is more similar to past trends in industry practice (OOP, Scrum, TDD, whatever) than it is different in the only way that’s ever been particularly noteworthy to me: it massively changes where people spend their time, without necessarily living up to the hype about how much gets done in that time.
Then it should have talked about the rest - instead of starting with rather graceless and ugly LLM-written generic prose about AI topics that to many readers is already tiresomely familiar and doubtless was tldr for even the readers who aren't repelled automatically by that.
I am as enthusiastic about formal methods as the next guy, but I very much doubt any LLM-based technique will make it economical to write a substantial fraction of application software in Lean. The LLM can play a powerful heuristic role in searching for proof-bearing code in areas where there is good training data. Unfortunately those areas are few and far between.
Moreover, humans will still need to read even rigorously proved code if only to suss out performance issues. And training people to read Lean will continue to be costly.
Though, as the OP says, this is a very exciting time for developing provably correct systems programming.
LLMs are writing non-trivial math proofs in Lean, and software proofs tend to be individually easier than proofs in math, just more tedious because there's so much more of them in any non-trivial development.
Some performance issues (asymptotics) can be addressed via proof, others are routinely verified by benchmarking.
This assumes everything about current capabilities stay static, and it wasn't long ago before LLMs couldn't do math. Many were predicting the genAI hype had peaked this time last year.
If you want it to be a question of economics, I think the answer is in whether this approach is more economical than the alternative, which is having people run this substrate. There's a lot of enthusiasm here and you can't deny there has been progress.
I wouldn't be so quick to doubt. It costs nothing to be optimistic.
If you give an agent a task, the typical agentic pattern is that it calls tools in some non-deterministic loop, feeding the tool output back into the LLM, until it deems the task complete. The LLM internalizes an algorithm.
Another way of doing it is the agent just writes an algorithm to perform the task and runs it. In this world, tools are just APIs and the agent has to think through its entire process end to end before it even begins and account for all cases.
Only latter is turing complete, but the former approaches the latter as it improves.
Maybe not an agent exactly but I can see an agentic application is kind of like a dsl because the user space has a set of queries and commands they want to direct the computer to take action but they will describe those queries and commands in English and not with normal programming function calls
My read was roughly that agents require constraining scaffolding (CLAUDE.md) and careful phrasing (prompt engineering) which together is vaguely like working in a DSL?
Sigh. Is there any LLM solution for HN reader to filter out all top-level commenters that hadn't RTFA? I don't need the (micro-)shitstorms that these people spawn, even if the general HN algo scores these as "interesting".
Every job in engineering is changing right now. Managers aren't immune. I've been an EM for almost 20 years in some flavor or another, and I've been thinking a lot about how I want to adapt to this era.
This is the first time I've seriously considered swapping out of management. Not for any of the reasons the author says, but because:
- I don't feel as confident mentoring others through this period given how much the work is changing
- I find myself enjoying the work more
- EMs tend to have more difficulty justifying their existence at the best of times let alone a period of change like this
The AI world will still need EMs. It's just unclear what those EMs will be doing every day and how it will work.
I've been thinking about this as well, and I'm glad the author is talking about it. However, I don't think he took it far enough.
It is correct to say there's near-infinite demand for AI, and supply is limited. It stands to reason that wealthier people will pay more, and therefore get more, out of AI.
However, this has always been true, but historically instead of AI it's been workers. The economics of labor haven't changed. So it will, as always, be a game of how you deploy the workers you hire. Are you generating useless morning briefs or are you actually generating value for yourself and others with the AI you buy? If you generate more value that the tokens you burn, you'll get ahead.
This will be true in academia as well, the area of interest to the author. He writes like, before AI, grad student level intelligence came for free.
I got my first tech job in 2001. I've been doing this a while and ridden all the waves.
There are two kinds of waves. The ones that don't require collective belief in them to succeed, and those that do.
The latter are kinds like crypto and social media. The former is mobile...and AI.
If no one else in the world had access to AI except me, I would appear superhuman to everyone in the world. People would see my level of output and be utterly shocked at how I can do so much so quickly. It doesn't matter if others don't use AI for me to appreciate AI. In fact, the more other people don't use AI, the better it works out for me.
I'm sympathetic to people who feel like they are against it on principle because scummy influencers are talking about it, but I don't think they're doing themselves any favors.
I think it is a reasonable moral stance to acknowledge such things are possible, yet not wanting to be a part of it. Regarding making it technically impossible to do...I think that is what Anthropic means when they say they want to develop guardrails.
If you read the statement, they explicitly state these guardrails don't exist today, and they want to develop them.
Though I have a feeling we're talking about different things. In Claude Code terms, it might want to rm -rf my codebase. You sound like you might want it to never run rm -rf. Anthropic probably wants to catch dangerous commands and send them to humans to approve, like it does today.
That's my point. They formed anthropic under the sole mandate of "guardrails first," now seemingly don't have them at all. So they're just another ai company with different marketing, not the purely altruistic outfit they want everyone to believe
The ability of some people to never be happy, and to find a way to twist a good situation into bad, will always impress me.
Here we have a company doing something unprecedented but it is STILL not enough for people like you. The DoD could destroy them over this statement, and have indicated an intent to do so, but it's still not enough for you that they stand up to this.
I wonder what life is like being so puritanical and unwilling to accept the good, for it is not perfect! This mindset is the road to a life of bitterness.
I have noticed, if I hit my session quota before it resets, that Claude gets "sleepy" for a day or so afterward. It's demonstrably worse at tasks...especially complex ones. My cofounder and I have both noticed this.
Our theory is that Claude gets limited if you meet some threshold of power usage.
If only we could look into the future to see who is right and which future is better so we could stop wasting our time on pointless doomerism debate. Though I guess that would come with its own problems.
First, I appreciate how he implemented auto-update. Not sure if that pattern is original, but I've been solving it in a different-but-worse way for a similar project. NOT a fan of how it's being used to present articles on Garry's List. I like the site, but that's a totally different lane.
The skills are great for upleveling plans. Claude in particular has a way of generating plans with huge blind spots. I've learned to pay close attention to plans to avoid getting burned, and the plan skills do a fair job at helping catch gaps so I don't have to ralph-wiggum later. I don't find the CEO skill terribly effective, but I do like the role it plays at finding delighters for features. This is also where I think my original prompting tends to be strong, which could be why it doesn't appear to have a huge impact like the other skills.
I think the design skills are great and I like the direction they're going. DESIGN.md needs to become a standard practice. I think it's done a great job at helping with design consistency and building UIs that don't feel like slop. This general approach will probably challenge lots of design-focused coding tools.
The approach to using the browser is superior to Claude's built-in extension in pretty much every way (except cookie management). It's worth it for that alone.
For people who don't understand this...think of each skill like a phase of the SDLC. The actual content, over time, will probably become bespoke to how your team builds software, but the steps themselves are all pretty much the same. All of this is still early days, so YMMV using these specific skills, but I like the philosophy.
reply