Exactly this. The non-composability and non-standardization of GUI tooling is my main issue with them ; having the same toolkit available to solve every problem takes some doing but is ultimately more efficient.
That being said, it's a hard sell. It's not easy to grok the simplicity of the commandline tools until you've used them to solve what would otherwise be an intractable problem.
Yes, any GUI that doesn't provide any kind of RPC or composable API layer is almost a lost cause to me. I'd personally refuse to use it. With a caveat, though: on MacOS, which I absolutely abhorred in the beginning, having now been required to use for work (instead of Linux), I found the beauty of OSA (Open Scripting Architecture). Holy mother of Turing, I wish there was anything similar on Linux. I had no idea how amazing this thing was. I can fearlessly dig things out of even complex UIs for automation in ways that I can't ever dream to be able to do on Wayland. That still not an excuse - any serious UI should provide a functioning API.
Floating point value stored multiplied by 10^8. That gives you a huge integer, but it's extremely accurate, especially for US denominated currencies. Easily transformed into floating point numbers for reporting/etc.
One of the reasons why a source-based system like Gentoo is particularly nice is that you can compile your binaries with debug flags, so if you hit bad behavior you can inspect, write a patch, compile into your running system, and then push the same patch upstream.
I barely have to do it, but imho, this is how software should work and what running a computer should feel like.
It's worth noting that even more staid distributions like Debian provide you with the means to do this. It's arguably bit more complicated, but saves you a lot of time and hassle on the happy path.
I use OpenBSD and it’s actually the same thing with the additional niceties of binary packages. A bug or an issue with any program (including the kernel and drivers)? Patch and rebuild.
I think the "net" does a lot of heavy lifting for a box like this - e.g. you do all the important work on a remote server, and only do basic maintenance work on the laptop itself.
4.8 also seems like a regression and using it from the chat GUI results in 4.6 no longer showing up. If someone from anthropic is here, is it possible to readd 4.6 in the "other models" dropdown ? I feel like I got a bit baited/switched here.
Yeah, I was using 4.6 way more than 4.7. Pulling 4.6 from the web chat also means we lose access to Extended Thinking there. So they're saving on compute. It's hard not to assume this was part of the motivation behind the 4.8 release timing.
On web and mobile I can still select Opus 4.6, after a chat using 4.8, listed under other models. Extended thinking is a toggle in the effort menu
When I select 4.7 or 4.8 Extended thinking is replaced by adaptive thinking, but maybe I've understood the comment wrong and you meant 'when they pull 4.6 from web chat'?
I think it's just nice to have things in a central place ; no one's really gotten decentralized tech right and things like discoverability, interaction, job running, etc. is really nice to have in one place.
Mastodon and email are the closest I've felt to a distributed system that works, but for oss stuff ... I think we're getting closer, but it's still a very hard problem to solve.
Opus definitely pushes me to ignore problems. I've had to tell it multiple times to be thorough, and we tend to go back and forth a few times every time that happens. :)
"I see the tests failing, but none of our changes caused this breakage so I will push my changes and ask the user to inform their team on failing tests."
There's something interesting I've found about my interactions with the AI - I use it as a thought-partner. I don't ask it to solve a problem for me (well, first at least!) I think about it as a tool to work with, engage with the problem, and spit out a result that I then test and review.
I see it as part of the feedback loop, and it speeds up some of the mechanical drudgery, while not removing any of the semantic problems inherent in problem solving. In other words, there's things machines are good at, and things humans are good at - if we each stick to our strengths, we can move incredibly fast.
The whole "integrated development" experience. Take it or leave it, but old farts like me go all the way back to poring over code on printouts since your only window into it was one file at a time in an 80x25 terminal - not terminal window, actual terminal or, by then, terminal emulator.
That does affect later habits like, for example, hating the information overload from syntax highlighting. And don't even get me started on auto-indent.
Whereas younger colleagues, whose habits were formed in the era of much more sophisticated tools, have dozens of files open in tabs and can follow, say, a "where is this defined" or "where is this driven" (this is RTL code, not normal software) in an instant. Keep in mind some oldtimers had really fancy emacs setups that could do that, and vi users had things like ctags.
That being said, it's a hard sell. It's not easy to grok the simplicity of the commandline tools until you've used them to solve what would otherwise be an intractable problem.