Many native web browser capabilities exist because adventurous developers came up with the patterns themselves. There is no way we’d have CSS scroll-driven animations available to us if front end developers hadn’t already established the requirement and pattern. Browser standards, CSS specs etc follow what _we_ hack together and make ubiquitous (sort of like desire paths), not the other way around. That is why we shouldn’t blindly “use the platform”; vendors should adjust to our demands, not the other way around. Of course, if a native element or capability matches exactly your requirements in some scenario, using it would be sensible.
I try to do more and more in the terminal, but even that isn’t immune to the same syndrome, programs tying up tons of threads and memory, terminal emulators themselves being GPU accelerated and performance intensive. Perhaps we will see a cultural hard fork and a trend towards small, simple, utilitarian programs in personal computing; if the price of basic components continues its upward trend, it may even be necessary. Devices that are purposefully limited, hacked, owned, customised, privvy to closed networks that follow the same philosophies.
What interests me is what the “stink” looks like and feels like to you and why you find it so off putting. Let’s get into the weeds. I realise I sound like a therapist.
I really, really wish that posts like this would give _specific examples_ of generated code they thought was sub par and what they did to remedy it. It’s not that I don’t believe that they found the output lacking in some way, it’s more that I read the post thinking, “I feel like Claude et al are way better programmers than me but this person is saying they’re more skilled, I’d love to see what their specific issues with generated code were and how they fixed them”. I mean, even one or two short examples would be amazing; without them I just come away with so many questions.
I don't understand how this is different from oai "structured output" (and whatever the similar paradigm was on Sonnet ~3.7 back then) which everyone moved on from. On their gh they say:
"Jev is TypeSafe's closed service for runtime-defined semantic decisions. This project reproduces that interface pattern with open models; it does not reproduce Jev's undisclosed model or training"
As someone else pointed out it isn't actually Jev... can someone enlighten me
I would like more information on what Degoogled means and whether it’s still possible to install 3rd party apps by some other means. Nice looking device, I expected it to cost more. Also, does this only work in the US rn?
The point being I liked the general offering including the AI _to begin with_. Once they went OTT with it it killed the product for me to the point of not wanting it with or without
Warp was so promising to begin with - super clean and polished UI, AI features that actually made sense, really good dev outreach & PR. Over time, they started cramming more and more AI in everywhere, took over half the input area and you could just see there was no coming back. This is why VC backed software will always stink in the end. I use ghostty now.
Well, when the author eventually gives up trying to sell people names of apps and the domain expires, I’m sure it will present good value when it gets bought and repurposed as a porn site; the first two apps under quick utilities are CreamShot and DripCleaner?!
Exactly, text content goes in a <span>, icon is an <svg>, flex align items center on the wrapper gets you close, nudge the icon to perfection with transform: translate
It does matter, but how long do you think it takes to get right? It's a follow up prompt or a few tweaks by hand. I also have an /a11y skill for it that's tailored to exactly the things it sometimes doesn't get right first time round. Further, while it may not one-shot that stuff every time, with a little setup and the right AGENTS/CLAUDE md - it's usually not far off.
Another thing that helps is pointing it to patterns in an existing codebase (e.g. "use the box-link pattern for cards, as shown in [..]").
EDIT: The point being that even if they make mistakes that are easy to spot and fix _now_, you'd have to assume that in the very near future those kinks will be ironed out - I mean, the capabilities are only going in one direction.
It really feels as though my 20 year career as a front end developer is coming to a very abrupt end; at least as I have know it these past two decades.
I just had need of this, dropped it in - works great. The only slight issue I had is that if there's a tiny amount of overflow, you basically don't get to see the fade because it's designed to disappear as you reach 100% scroll. In practice, that means I've noticed single chars getting lopped off with no visible gradient. I don't think there's much way around this beyond adding a bit of right hand padding to the track
If there’s one thing I’ve learned these past couple of days, it’s to resist the temptation to jab the escape button and start waving my arms! I wonder how much of this cyclical self doubt / self congratulating I go through in my own thoughts without even realising it. If you could verbalise or articulate all the half thoughts, snatches of ideas, feelings and ruminations the human mind goes through on some tasks it might be even more bizarre (or could just be me)
Your post made me laugh because I experienced the same as you but the other way around. I switched from Claude to a multi model harness a couple of days ago and the first model I tried was GLM5.2.
I gave it some simple code porting exercises and watched dumbfounded at the reasoning, which was more like the ravings of a lunatic - but lo and behold, after much confusion and a dizzying number of eureka moments the task was completed very successfully.
I tried Kimi on a similar task, much faster, a little more reassuring somehow in its ramblings, also surprisingly good results.
To be clear, I’m not surprised the results were good because they’re not GPT or Claude, but because the line of reasoning was so bonkers. Coming from Claude, I was just not used to seeing this, but I’ll bet it’s just as nuts with the frontier models and we’re just not allowed to see it (I’m about to read the links you shared).
Agree wholeheartedly that transparency is of grave importance.
Strange, I managed to sign up with an email alias and pretty much always have my proton vpn turned on, never had any issues with Claude code or web - maybe I’m about to? :D
Been thinking of switching to a multi model harness anyway.
That the frontier models and subs are all big tech is probably what bothers me most about “AI” right now, but I’m bullish on advancements in the capabilities of local models. I suspect and hope that, in time, the field will level and we will have very capable local, offline models and the landscape will be much as it is now with subscription compute in the cloud for enterprise and self host / local first for indies / hackers etc.
I used Kagi for a bit but began to question what I was paying for, there are plenty of other search providers basically doing the same thing, DDG, Ecosia, Brave search etc. I mostly just use DDG now because search as a paid offering that scrapes results from other indexes doesn’t seem like good value