GPT’s differentiator is they focused on training for “thinking” while Gemini prioritized instant response. Medium thinking is not the limit of utility
Re: overparameterization specifically Medium and High are also identically parameterized
Medium will also dynamically use even higher thinking than High. High is fixed at a higher level rather than leaving it to be dynamic, though somewhat less than Medium’s upper limit
Memories of years ago on Stack Overflow, when it seemed like every single beginner python question was answered by one specific guy. And all his answers were streams of invective directed at the question's author. Whatever labor this guy was doing, he was clearly getting a lot of value in return by getting to yell at hapless beginners.
SwiftUI is also now an API for writing Android native apps, too.
I don't know why you skipped over that part? It's maybe like when Google rewrote Java. Win32 is a bad comparison because there's no Win32-compatible API for native apps on other platforms (that I know of?) except for emulation, but the Android SwiftUI project is not using emulation, it runs the code natively and the result is native Android UI.
Apple has not released a version of the SwiftUI framework for Android.
There are third party attempts to create something that lets SwiftUI code run on Android, just as Wine is third party software that allows you to run Win32 apps on Linux.
No one said it's first-party. And it's a good thing that it's not first-party! We want multiple options for deploying our code.
As I said, Wine provides emulation. But the SwiftUI on Android project does not emulate - it runs your SwiftUI code natively (as Swift that is compiled for Android), and maps it to native Android components, fully accessible and meeting platform expectations.
Completely different result and experience.
This also means that you can extend the SwiftUI on Android code with Android-specific code that will not run on iOS, to add other Android-specific UI. This is impossible with Wine + Win32.
I like LLVM, and I enjoy a good UI focused-language like Vala or Obj-C. Building with or contributing to Swift is a waste of my time as a Linux developer, it was in 2018 and it still is in 2025. Foundation will not fully support Linux until the late 2030s, and even a fully-implimented SwiftUI translation is still ignoring basic GNOME HIG and lagging behind best-practices. I would not be developing apps I want to use, or ship to users on other platforms. Electron would be preferable to cross-platform SwiftUI, and deep down you know it.
And that's my sympathetic opinion, as a Linux developer who loves their native UI trinkets and pseudopolish. Windows developers have dozens more options and likely won't find out Swift ever existed until Swift 2 is announced during a keynote presentation. Broader adoption of Swift has simply failed. If the language disappeared tomorrow, I wouldn't know as nothing on my system consumes Swift as a dependency according to nix-tree.
Obj-C is unusable for many new Apple platform features. Not suitable for building anymore.
Electron - not available on iOS, so it is out of the question.
I make a living off my iOS/macOS apps so I am interested in ways to diversify without giving up the platform that makes me my money. These cross platform solutions for Swift are interesting for those targeting Apple platforms. I agree they are not compelling if you do not prioritize Apple platforms.
I can't make a living off Linux like I can on Apple. Android is also much less profitable. So Apple continues to make business sense for me, for what I build and who my customers are. And thus Swift.
Frustrating that OpenAI support does not help with this, and no one on the team is available to resolve critical issues that leave customers unable to use quota they paid for...
edit: If you are downvoting because you are also anti-AI, my comment is not about whether supporting AI is good. I'm only remarking that they are aggressively negative about the topic. The aggression is obnoxious and less tolerated with other topics.
Lobsters is in general very anti grift and marketing. A huge portion of daily hype submissions are low signal fluff.
As it happens AI the the hype of the day. Yes it is useful but also it attracts the same insufferable people who were pushing NFTs 4 years ago. So Lobsters have separate AI tag for technical pieces to do with actual development of AI systems and "vibecoding" for softcore user experience entries. Lots of people mute the latter. This, and the fact that the site refers to their blog posts as lowly vibecoding irks some of submitters.
Even if they were anti-AI (which I don't think is true), it's not a big deal. There isn't exactly a shortage of write ups about GenAI on the Internet.
There are film photography forums where digital photography is off-topic. There is at least one machinist forum where the topics of drone-making and 3D printing are explicitly banned. This is fine, these are niche places for like minded people who don't want to be drowned out by the masses promoting whatever is hot.
> Even if they were anti-AI (which I don't think is true)
The entire site is constantly talking about how AI sucks and how it's not worth taking seriously. The AI tag was exclusively "vibecoding" despite it being extremely annoying for anyone who wanted to talk about it. Posts about AI being bad are upvoted far more, despite being very obviously low content/ uninteresting. Sentiment is very obviously anti-AI on lobsters.
I don't even think this should be contentious. The site is very openly anti-AI. When discussions about this have come up in meta posts it is overwhelmingly the case that users say "yes, we are anti AI, that's fine".
> it's not a big deal.
It stopped me from posting on there and I deleted my account because the conversations were so stupid. It comes up frequently that users want to be able to filter out "AI fluff" separately from "interesting discussion of AI". You can have a post about the internal algorithms and optimizations of an LLM and it will still be labeled "vibecoding" on the site - if that doesn't blatantly indicate a site-wide anti-AI bias I don't know how.
HN comment sections are full of anti-AI remarks, but there’s enough volume of contents that you can still find some quality info here.
On Lobsters it feels like the angry anti-LLM mindset is woven into the site’s culture, like you’re breaking some unspoken rule if you accidentally say something non-derogatory about AI.
As you wrote earlier it's a culture war thing. Lobsters is very big about being on one side of the culture wars there and the American social justice side is anti-AI, so the culture on the site has adopted the same view.
Just regret waiting until I was 35 to attempt it and now I'm already going on 40. Wasted good years on easy mode.
edit:
But this part is wild to me: "I use AI for some things. It helped me fix a few bugs a couple of times"
I can't imagine being solo indie and not leaning hard into Codex, CC, or Composer at this point. To use it only sometimes for the rare bug or copy editing sounds tragic. It's been an incredible boon for me at least - extending, refactoring, prototyping etc. within a complex codebase I wrote myself and in new ones that I guide it on.
> I can't imagine being solo indie and not leaning hard into Codex, CC, or Composer at this point.
Some of us don't want to because we like our artisanal programming, and being indie is great because we don't have to capitulate to management forcing us to use AI all the time.
What's that saying? "If you love what you do, you never work a day in your life."
I use LLMs every day, but not to write my software. I use them like really good personal assistants. They are now a standard (and invaluable) part of my daily workflow.
I'm not exactly "indie" (retired at 55, and working for free), but I can relate to a lot of what's being discussed, here.
Similar situation here. Laid off last year (at 45), can just about afford to retire to a frugal life (which I wouldn't mind at all).
But now that I am finally free to write only software that I want to bring into the world, I cannot imagine playing roulette with LLMs all day for something as mundane as productivity, giving up all the intellectual joys of the craft as well.
And the relief of not having to justify to a manager why things are taking longer than expected! It's like the giant finger that had been pressing down on me almost all my working life has finally been lifted. Sweet semi-retirement, how I love thee.
That's nice for you that you made enough money from your careers to not optimize your productivity, and to enjoy spending your time with code as leisure.
You two are retired or semi-retired and interested in the pleasure of coding apparently as an unpaid hobby. But I am speaking only to the utility of LLMs for business needs as an indie - the blog post is also about making a living as an indie and speaks to LLM use in terms of what they found productive. Indie in this case means independent business, not independently retired from business...
I should have put "productivity" in double-quotes. What I meant is I don't have to worry about justifying why I'm not using AI to "optimize" my "productivity" to a pointy-haired boss as my FAANG friends seem to be having to do.
As I said I'm only semi-retired. There's a very expensive fortnightly life-saving medicine I depend on which I'm currently getting for free on a government programme. But it's tied to my current place of residence and other places in my country either don't offer it at all, or not to people who haven't been living there for decades. I would like to afford ending my dependence on the government programme because it might end someday anyway, and also so that I can travel a little.
All this is to say, if the newest AI tools actually increased my productivity, I would be stupid not to use them. But I haven't found that to be the case. It's fine to use for autocomplete and to look things up now and then, but every time I've had it generate a substantial amount of code it has messed things up and in the long run cost me the same amount of time I could have done the job in myself. And doing it myself has always been more engaging, fun and satisfying, and has left me with a much better understanding of a complex system that I'm going to have to maintain and extend for years.
BTW: I didn't actually make that much money. Kids coming out of college, make more than I ever did, at my peak. I just lived very frugally, avoided personal debt, and saved and invested as best as I could.
I guarantee that a lot of folks here, would sneer at the way that I lived, and live now.
I wouldn't say I wasted the good years on easy mode, but I did have a misguided plan that I needed to become a team lead and then I could start on my side projects.
I thought I'd have more time because I saw just how little my team leads actually did at work, and no, they weren't doing it behind the scenes, they wouldn't even show up to meetings with the business.
Of course I cared too much, burnt out my best years, quit and now have jack to show for it on the wrong side of 40 when I could have built up my own indie empire.
reply