The first <sigh> does sound a lot like a moan. OP linked to the timestamp so I missed it when it first played. I was also confused but on second playback I heard the first <sigh> and also thought wtf.
> I can't speak for other products, but on the type of features I work on, AI is blatantly incompetent
I do chat with the web interfaces of LLMs quite a bit as my workflow. I’m always baffled by how rarely it arrives at the correct solution.
I pose a question with code snippets, it gives me some incorrect answer. I pose a better question, and it gives me another wrong answer. Wash rinse repeat several times; not until the LLM gets the right answer, but until I pose the question specific enough that I can answer it myself.
I do find this loop of question refinement quite beneficial, but man it makes me terrified of vibe coded applications. 9/10 times the LLM doesn’t even get in the ballpark of the problem, let alone the solution.
Not only that, but even when you get the LLM to give you the right answer, if you ask it to turn that answer into actual code committed to your repo, it usually destroys something else in the process.
The bias in this comparison is insane. “Tinkerer” is a pejorative, normally used to downplay one’s (own) ability or impact. It’s often used to mean “unserious” or “amateur”.
The same comparison could be said the other way around: there are entrepreneurs, who care deeply about the product, inside and out; and scammers, who don’t care about the problem itself or the beauty of a a neat solution, but only the sales of the end product.
Entrepreneurs currently are in hell, and a scammers are in heaven.
I don’t think either is more apt or more offensive, they just appeal to whichever side of the fence your on.
I've always though that about tinkering as in "being in love to work with inner working of things and understanding the internals, mastering stuff". Nothing pejorative.
Tinkering is more or less synonymous with dabbling, which is used more these days. Synonyms with tinkerer would be amateur, hobbiest, dabbler. There’s definitely an implied interest in how things work, but also strongly implies that one is not serious about what they’re doing.
More in line with your definition would be artisan or craftsman; although fwiw craftsman can have a negative slant in modern (American) English as some people use it in lieu of blue-collar workers.
Since there is an entire AI community[0], claiming 135,000 members, that labels themselves 'tinkerers', any pejorative meaning in tinkerer is long gone. Probably in the same way as 'hacker' was once a rather pejorative term to the general public.
The definition of tinkerer is “An unskilled person who tries to fix or mend”. Someone making a slack channel with a diminuative moniker does not change the meaning of the word.
You would not want a surgeon who says that they’ve tinkered around with surgery. You would not want a pilot who says they’re tinkered around with piloting.
Synonyms are quite literally beginner; dilettante; novice; pretender; uninitiate.
I am not challenging the definition, only your "pejorative" connotation, which is simply incorrect. Everyone is a beginner at some point. Does that make "beginner" pejorative?
Depends. If you’re using is disparagingly, then yes, it is a pejorative.
For example from the original analogy, if you told an entrepreneur that because they didn’t use AI, they were not an entrepreneur, but instead “unskilled” or a “beginner”, that would be disparaging; pejorative.
Idk about that. I work for myself and I cannot stand coding with LLMs. Even if it were faster (I have only noticed slowdowns long term when using them), the experience of arguing with a slot machine is far more stressful and soul-sucking than even the gnarliest bug.
Yes coorporate “ship as fast as possible” is draining, but that doesn’t mean LLC coding isn’t also to blame.
I freelance too and I have been struggling personally for a while for other reasons but I absolutely recognise this. I have only about eighty lines of code in production that I fully did not write (thanks Muse Glimmer).
The difference between corporate salaried programming and freelance programming is who is on the hook in a direct sense for the bug the LLM introduced, or for maintaining the massive pile of code or the cognitive burden of ownership.
My experience of freelancing is that most of the people I am competing against for the work care less about the code (and things like quality, security, appropriateness, user burden), are faster, are happy to nickle-and-dime the bill, will charge more to fix their own bugs, churn out meaningless requirements specs that they cannot stick to, and rely on lack of client knowledge to get through.
So there has always been a problem of competing against the slapdash; they always win the quote. I care about longer-term maintainability, cost control and quality, and I lose out because of it. (I'm often asked to maintain or finish the work of people who beat me to the quote and then failed; nowadays I say no.)
The problem is, those people now use LLMs. For the immediate future, LLMs may actually be improving their output, certainly at the prototype stage, and they will now be including one-shot, vibe-coded demos with their quotes.
So what is a freelancer who still has a sense of ethics to do?
The long bet is that those people who do not care for quality will still be disappointing their clients, but how do you avoid using LLMs while still getting, at least, back to where you were in competing with them?
I don’t freelance anymore, so I can’t give any freelancing advice. Although freelancing has always been fraught with offshoring, price arbitraging, and generally lower quality (not saying all freelancing, just that it was an issue when I was freelancing a decade ago).
I’m sure LLMs have only made it worse, possibly pushing it over an edge where hand coding as a freelancer is not viable. I do not know.
Personally, I tell myself that we’re in the lowest of the low at the moment. Either AI will get substantially better, or it will get substantially worse. The amount of debt the industry is loaded up with is not sustainable. Something is going to have to crack. I expect in another 2 years for things to “settle down” on some side of the fence, although which side I do not know.
My quipy advice though is to build something for yourself. Be the boss that doesn’t require AI. We need more builders behind products, more indie products built because they’re cool and the founder is passionate about the problem; not more business types and cash grabs.
> My quipy advice though is to build something for yourself.
This! Groups with strong software skills (like HN) miss how LLMs unblock people who previously were unable to build what they wanted because they lacked the software skills.
I know an ME at work who have begged software folks for a year to build specific software he wants and was getting a mix of "we are too busy", "it does not work this way" and quick prototypes that were not quite what he wanted. Then he spent his free cycles with Claude building what he wanted. It took some time and some iterations and it may be ugly under the hood, but he wowed many folks with different types of demos of our system. The view he presented resonated with many users. And he feels empowered how that he can build tools he wants to try himself and do some projects that before he had to beg for, himself.
The software is being disrupted and disruptions are painful. But maintaining the status quo of several big tech names controlling everything we do on most electronic devices and optimizing for the doomscrolling time is not a worthy goal either. My 2c.
> This! Groups with strong software skills (like HN) miss how LLMs unblock people who previously were unable to build what they wanted because they lacked the software skills.
That isn't me. I have the software skills. The problem is not lack of ability building complex stuff from scratch, I do it over and over again, and I have had my own products in the past.
The problem is that LLMs do not actually improve the psychology of sustainable lone freelancing, which is more burdened with the problem of actually owning the entire problem and its consequences than someone vibe-coding their fun ideas in the lunchtime safety of their salaried job or even a single developer within a corporate team setting (because they can quit and walk away when they hate it). If you're still getting paid as you fix it when it breaks after deployment, you're in a different world; it doesn't matter to me how empowering LLMs are in that context!
When you are up against people who are bluffing/bodging, or small agencies who think coding is just a little step up from web design, or over-extended hobby coders working in the evenings, LLMs have a very different set of impacts on your life. I can't just run an experiment with some LLM code and roll it out, because I have to face the customer directly and because nobody is going to pay me to fix it; I am not on a salary that allows that kind of experimentation.
So I am left trying to work out how to use LLMs to catch up with the above people who are, without losing control.
Back in the day, folks without software skills were able to build certain things themselves; from standalone Microsoft Access databases with simple Visual Basic 6 frontends, to finance maintaining huge spreadsheets as pseudo databases, etc.
The issue now is that LLMs increase the blast radius. Before it was one person lighting a firecracker. Now it's one person setting fire to the fireworks factory. For example, the LLM exposes the ME's tools and internal company data to the Internet.
I do not understand this blast radius, firecrackers and explosions talk. No offense intended, but this seems to assume that those not in software engineering are like kids that should only be allowed to play with new toys under adult supervision.
I do not agree to that. Being able to make computers do your bidding is a very powerful skill, but is not universal. Software is a small part of the real world. If LLMs help non-software types tinker and try their own ideas more effectively, vive les LLM! If some folks who write CRUD apps or ad distribution algorithms feel threatened by LLMs, ... well, that is a trade worth taking. My 2c.
I agree generally. At least, I have a better time using LLMs when I treat them like StackOverflow or Google, where I'll ask a question non-specific to my code, and figure out how to fit the answer into my project. When I give them access to a codebase, cognitive debt is incurred far too quickly. I know people online will say I'm just not using it "right" -- I do not really believe them.
I love it, although in my experience, one needs very strict agent rules that tightly define how the AI should develop. The problem is that 1) these rules take a lot of effort to build and 2) AIs always overengineer, complicate, overstructure, overdesign… ;)
Same. I miss the excitement around programming. New languages, new tricks, weird hacks, odd projects.
I have little to no interest in AI coding - I try it every couple months but always regret it, git reset, and go back to coding by hand.
It feels quite lonely these days. The general energy and enthusiasm from hn was a big source of motivation, or maybe community. Now using hn feels quite isolating.
I do see the occasional post about going back to hand coding though, and their frequency seems to be on the rise, so hopefully that correction continues
(I don’t need to see no AI, to each their own, just would be nice to see an equal amount of hand coding content one day)
Yeah personally I think he is right about a lot of the current state of affairs. I think he has a decent grasp of the situation despite being incredibly biased.
His predictions though? I don’t think I’ve ever read anything from him to get his read on how things will play out.
Just because one makes the wrong predictions does not mean that the data used to make them was wrong. A lot of people seem to dismiss his reporting because of his takeaways
Idk Google Maps is pretty bad. Clicking on the search bar force opens the App Store, and trying to load photos does the same and straight up lies to you saying it’s only available in the app. But if you scroll down to the second carousel of images it loads fine in web.
Google Maps has a better excuse for wanting to be an app than most since a primary use case on mobile is turn-by-turn navigation. That's possible to do in a web page now, but several features that improve the experience are not available in browsers.
I'm annoyed by a different thing ... It used to be when you searched for something, maps would always be a tab for results. Now it doesn't seem to be anymore...
What you need to do, in the modern Web, is to flip that around.
Anytime I need to search an entity with physical presence, I use Google Maps first. Maps is very good about getting the first result as intended and not showing superfluous non-matches. Maps will not chirp up with AI-generated responses. Maps will take me right to the place I want, and I can see a lot of profile info at-a-glance, and if I wanted their website, then Maps will give me their main website, and nothing else.
Whether it's a corporate HQ, a clinic, a local restaurant, or a geographic area, Maps can come up with a lot of relevant info and I don't waste my time in an enshittified Search. So I'm thankful for the Maps app and its mobile integrations.
I’m not sure how that’s a justification. If you want to app-wall the turn by turn directions, that would be one thing. The fact that turn-by-turn directions is better on mobile doesn’t rationalize lying about being unable to show photos, and auto opening the App Store three times every time you want to search for an address somewhere
However, I live with Activity Monitor open at all times. The computer constantly grinds to a halt, and I have to force quit apps at least 10 times per day.
Not really Apple's fault though, just most apps use around 4GB of RAM these days even when they're idle. Some use upwards of 16GB.
I've watched the computer slowly get obsoleted over time, not because the hardware is too old, but because software just get's worse every year.
Install Linux to unleash the full potential of your M1 (or to liberate any other system with a BloatOs, like macOS or Windows).
I resucitated and saved from the landfill a laggy and almost unusable laptop with Window 10, 4GB RAM and a Celeron processor, how?... Installing Manjaro (Arch), XFCE and kernel 6.18.
ZSwap and MGLru do wonders with memory and cache management.
It's fast enough with Thorium SSE4. I can easily open 15-20 tabs.
I mean, you could make a better comment about it, but in my opinion the sarcasm is justified.
AI has greatly increased the number of security bugs found on a weekly basis (which is great) and we're getting one of those one-shotted and then immediately abandoned open source "projects" every day (which is useless).
And yet there is no indication that any of this new software development capacity is going into improving the quality and efficiency of the software we use on a daily basis.
Why is widely used professional software not getting better and more efficient if bug fixing is supposedly so much cheaper now? Is it not worth putting any effort into optimisation at all?
I think the reason could be that AI can only do more of what we have done manually before, i.e. more lowest common denominator code.
Or am I just too impatient and this is all coming?
> Or am I just too impatient and this is all coming?
I doubt its coming, at least not within the timescale OpenAI or Anthropic would like to believe. If we were close, then we'd see the fruits of that, which like you said, we are seeing the opposite currently.
The bottleneck was never writing code, and it still isn't. Writing code is like 15% of the job, but the labs can't tell investors "Yeah we make everyone only 15% more efficient" and still be able to raise trillions in capital. Whatever AI accelerates, the bottleneck is still code review, comprehension, and security debt.
Vibe coded slop gets abandoned because non-devs were never going to be able to use AI to make commercial software beyond an MVP. Once your app needs multi-tenant auth, session invalidation, concurrency handling, API changes, graceful degradation under load, etc. natural language prompts will fail. You still need a mental model of how things work.
When building a house, lowering the cost of bricks and getting them delivered faster does nothing to speed up the rest of the process that still relies on engineering discipline.
I don’t get it either. What are these people doing that keep saying that Mac OS is snappy. It’s the same slow garbage as windows.
One problem is that you need Firefox because safari doesn’t peppery do ad blocking. But once you have 20 tabs open or so, it uses 5GB GB ram. But perhaps your also need safari for something else. That’s another 5GB of RAM. Then once you count a finder replacement (2GB ram), and some office stuff (probably 1-4GB for each app window), 16GB of ram are almost as sluggish as an old windows laptop. Add MS Teams, and the thing is brought to it’s knees. And Mac OS has a lot of quirks and the window management is atrocious (alt-tab doesn’t switch between app windows, a bunch of apps to get alt-tab behavior don’t work properly either).
I feel like more and more that Mac Users are being forced by the atrocious window management to only have one window open per software, so then the system is fast. Problem is when you have multiple tasks going on in parallel, then you need many windows and desktops.
I don't really think it's all that much about macOS. I think it more has to do with the state of web development - claude.com or chat.com or grok.com all can use up to 5GB of RAM as an idle webpage in the background regardless of browser. Most apps these days seem to be electron apps which can fire off multiple 4+ GB node processes doing who knows what.
I think most true native apps don't have these issues. Unfortunately there is very little incentive to build on the macOS platform these days, which is where I point more of the finger at Apple. UX, UI, & DX are all quite a bit worse for Mac native.
I live with safari and finder and all my office ware is native (mail.app) or web based (Google Docs). Chat is via slack, but I use the website and not the native app. I've adapted to the quirks and I'd bet that's a lot of the reason why it doesn't feel slow to me. I feel I have plenty of windows open at once but likely less than you do.
If you have to use specialized software to make your system snappy, then it is, as a system, not snappy. The attitude about the MacBook is that it never slows down … what, unless you use MS office?
my point was I'm not using specialized software, I'm using the software shipped with the system. MS Office isn't Apple Software, so holding apple responsible for Microsoft code quality seems counter to your first point.
If I have Office stuff open and the whole system crawls, then the system isn’t snappy.
It’s like the claim is “get a Mac and you’ll never see a slow down ever again! Just as long as you run software that won’t make the whole system come to a crawl”
Awesome will check this out. EUrouter is too cost prohibitive for me, and now that open router has been acquired by stripe it’s only a matter of time until it’s unusable
Depends on what’s being written, and who the audience is. Anything of any length would be hard to simulate in a way that would fool an author - writing has a certain flow to it. A cadence. The editing and restructuring, deleting of words, typos you don’t catch until some random reread, rephrasing of sections because you want to use the original phrasing later in the piece.
Could you simulate something be typed? Trivially. Could you simulate something be drafted? Honestly, even if you wanted to put in all that time and effort, I’m not even sure LLMs are sophisticated enough to send the logical drafts, loops and edits that would pass a writers sniff test
I think you could simulate something that passes a sniff test. A writer would probably spot implausibilities in the simulation if they paid attention, but then we're back to square one, because you can spot that something was written by an LLM after you've wasted your time reading it and realize that you've been led around in circles with superficial information and no coherent train of thought, but by then your time has already been wasted.
To add to this, one can’t ignore the relationship between signal and receiver. I’d imagine most people on HN have enough pre-LLM reading experience to have a decent sense of what was written by an LLM versus a human.
And as LLMs get better at producing human-like text, that same pre-LLM reading experience, which helps people tell the two apart, will become less and less common.
I would not want that in my product.
reply