Notable and persistent and extraordinarily petty is their refusal to read AGENTS.md files, forcing the inclusion of their branding into the source of repos directly.
to be clear, I don't mean that Claude refuses to read AGENTS.md files, but that Anthropic refuses to bake AGENTS.md into the harness as a first-class feature.
Claude the model by now has learned what an AGENTS file is, and itself is not petty enough to ignore it for partisan purposes, but claude-code the harness doesn't natively support it.
I have lots of choice (I own the company) but I'm still not going to switch from Claude until I see evidence that the alternative is meaningfully better. So far I don't see that evidence. In the past I've looked at using competitive products and it turned out to be a painful experience (Cursor didn't work at all on my computer, Google thing -- whatever it was at that time -- required dependencies I wasn't willing to install). I'm sure these issues have been resolved since but why would I spent time kicking the tires of another product just to have it work "as well"? Claude's cost to me is minimal so there's no cost savings to be made.
fwiw nobody "marketed to me". I picked Claude because friends were using it with great success and they helped me get started with suggestions on prompt style. Before that I'd played around with various LLMs for coding but not done any actual production work.
Unfortunately the browser still can't make the kind of network connection needed to transport a terminal session to a remote computer natively. afaik all the tunneling solutions are pretty clunky/insecure.
Have you looked into VibeTunnel? I got the terminal working in my browser and it runs on my computer. I can access it on my phone since we’re on the same Tailscale network. I use Ghostty-Web and tried to switch to Wterm but it didn’t work. I think it’s because Ghostty-Web renders a canvas and wterm normal div tags.
Yeah ChromeOS has the same problem. They have a Javascript-native terminal emulator, and a wasm (formerly PNaCL) implementation of open-ssh. But they have to use ChromeOS specific browser extensions in order to allow native TCP connections to port 22 from wasm, and Google only gives themselves this privilege, not any random dev like us.
I'm curious to see if this weird stack gets ported to the Googlebooks or if they just make a mouse-friendly Android app instead.
It doesn't take much to support a WSS to Telnet/SSH bridge or directly hosting a PTTY shell... for that matter, you can leverage path/querystring to include signed credentials (like a JWT) as part of the path statement for the underlying connection.
There's even some decent options to bridge further the other way, from a terminal to wss back to a terminal based, server hosted, application.
Lot but not enough still. Most web tech is like that, almost there but not really. Webaudio prob being the worst one. Webgpu being weird thing that nobody really knows who it is for.
reply