I love seeing an `AGENTS.md` in open source projects.
It's now my #1 heuristic to know if the team is on the right track.
(I need to start adding them to all my projects.)
Ghostling
61–70 of 72 posts
Re: Ghostling
#62Earlier quoted context omitted.
I think your github readme is really missing a picture/screenshot to quickly understand what is the experience like. I.e. if your app is mainly about adding the chrome (as in the surrounding UI pixels) around the TUI, then it would be good to show what is the chrome like.
Nah, I think it’s pretty clear. It would look like a terminal emulator. Just like how Electron looks like a bunch of browser widgets - because it’s literally a single-web-app browser.
A screenshot would still make it clear that that's all you're getting - and no extra chrome etc
Re: Ghostling
#63This is amazing and might be exactly what I’m looking for my own weirdo retro tooling that sometimes needs to run over ssh but also expect a “GUI” experience… any metrics on the overhead this might add to, for instance to a hello world type program?
Re: Ghostling
#64I built my own macOS terminal app over the last two weeks using swift/appkit + ghostty's zig library. I basically just wanted vertical tabs and notifications for when AI agents (claude code, codex) are finished. I already use it as my main terminal over iTerm. It's a fun project since I use my terminal all day, so I always have ideas for how something could be improved or polished. AI can do the chore work of figurin…
I hardly think you can say YOU built it went all your did was manage 10% of the project. An LLM built it, you just test it.
Re: Ghostling
#65I built my own macOS terminal app over the last two weeks using swift/appkit + ghostty's zig library. I basically just wanted vertical tabs and notifications for when AI agents (claude code, codex) are finished. I already use it as my main terminal over iTerm. It's a fun project since I use my terminal all day, so I always have ideas for how something could be improved or polished. AI can do the chore work of figurin…
I hardly think you can say YOU built it went all your did was manage 10% of the project. An LLM built it, you just test it.
Figuring out what to build, how it should work at a high level, and what it should be like to use it is the hard part and the defining aspect of any tool or application. (Or a fun game, for example)
Now that the cost of writing code is ~$0, I think the entity who figures that out is the one who built it, not the entity that does the trivial chore once the hard decisions have already been made.
Re: Ghostling
#66Earlier quoted context omitted.
> iirc no way to focus the originating terminal pane on notif click. Yeah mine has the same issue - it just brings up the script editor. It's not as much of an issue for me since I'm rarely running more than a single Opencode instance at a time. Could definitely see that being useful if you were running quite a few agents to let you hone in on the correct terminal window/tab quickly though.
Yeah, exactly. I have so many agents running these days. 1-2 per project. Often one is polishing a plan file for the next feature/bugfix (claude), and the other is reviewing it to give feedback (codex). And that is my life now. But on the upside, it's trivial to maintain sideprojects now.
Nice. You have a similar setup to mine. I've found OpenAI's gpt-5.3-codex (at xhigh) to be pretty dismal at actual implementation but it's surprisingly adept at performing a high level feature branch review and catching issues that are missed by Gemini/Claude.
Re: Ghostling
#67Earlier quoted context omitted.
Yeah, exactly. I have so many agents running these days. 1-2 per project. Often one is polishing a plan file for the next feature/bugfix (claude), and the other is reviewing it to give feedback (codex). And that is my life now. But on the upside, it's trivial to maintain sideprojects now.
Nice. You have a similar setup to mine. I've found OpenAI's gpt-5.3-codex (at xhigh) to be pretty dismal at actual implementation but it's surprisingly adept at performing a high level feature branch review and catching issues that are missed by Gemini/Claude.
Yeah, through trial and error I found that Codex (now gpt-5.4) has really good feedback once there's a plan file.
Kinda like someone on the internet who suddenly becomes an expert once you make concrete claims for them to shred into.
I have an AGENTS.md prompt that specifies how to review a plan. Something like: for every finding, rank the top solutions internally and then recommend a solution. And if there's a simpler, better pivot that the plan should take, pitch it at the end.
It's really actionable, and I basically can paste the review back into Claude, then tell Codex to re-review the plan, and repeat this until it has nothing more to add, and now I have a really good plan.
Re: Ghostling
#68The C file is small enough to read (over a few minutes.) I got to about line 5 and realized: I’ve never seen quite that technique for embedding a font via an autogenerated header before. I’m more used to Windows resources; this seems to generate a byte array in CMake code. I’m somewhere between horrified and impressed, in that I feel we’ve finally discovered a cross platform binary resource embedding solution.
> I’m somewhere between horrified and impressed, in that I feel we’ve finally discovered a cross platform binary resource embedding solution. Maybe I'm misunderstanding your comment about having "finally discovered" that... For to me embedding binary resources in source files is nothing new at all? It was definitely done in the early JavaScript days for a variety of reasons. Arguably early basic listings that had lot…
"Finally discovered" was more that there are so many difference ways, all of which work differently. Maybe this is just adding one more technique, xkcd-style, but it struck me as more possibly cross-platform-identical than several techniques I've seen. If you embed Windows resources, for example, that's highly Windows-specific; this one would work regardless.
Re: Ghostling
#69Earlier quoted context omitted.
Nice. You have a similar setup to mine. I've found OpenAI's gpt-5.3-codex (at xhigh) to be pretty dismal at actual implementation but it's surprisingly adept at performing a high level feature branch review and catching issues that are missed by Gemini/Claude.
Yeah, through trial and error I found that Codex (now gpt-5.4) has really good feedback once there's a plan file. Kinda like someone on the internet who suddenly becomes an expert once you make concrete claims for them to shred into. I have an AGENTS.md prompt that specifies how to review a plan. Something like: for every finding, rank the top solutions internally and then recommend a solution. And if there's a simpl…
Nice. I have a similar loop (Opus 4.6 plan Codex xhigh review) that I kind of run in an iterative feedback loop until meaningful actionables converge down to zero.
I initially gave some thought to automating this because CC/Codex do have SDKs... but every once in a while Codex will propose some absurdly over-engineered stack advice that I have to manually reject before passing it back to Claude.
Re: Ghostling
#70Earlier quoted context omitted.
Agree on all the points, except 4. There are even people out there who use lynx as their primary browser :) Although while I usually like tabs for most apps, I don't use tabs for terminal and rely either on window manager or tmux. I guess the difference is that I often want a mix of tabs and having multiple terminals side by side, whereas I don't really need that for a browser (or very seldom)
Which window manager do you use? Sway had the better, though often tedious, WM tab solution that I've tried. Niri had a useless one. I really tried to love sway splits and tabs for terminal windows. But I finally admitted I'd rather just alt-tab to a few different terminal apps, each with its own concern (maybe one per project, this one for my remote machine), and best of all, each with its own internal tabs. That sa…
My favorite is still AwesomeWM, although I haven't used it in years. Nowadays I'm on mac and my "tiling manager" is about 6-7 custom functions I wrote in hammerspoon. Basically just tile left/right, full size, "reasonable" size + centered, and a shortcut that distributes all my apps in the virtual spaces I've assigned to them.
One of the reasons why I use tmux rather than i.e. iterm2 tabs is that I don't have to change any of my habits or learn extra shortcuts when I'm working with ssh.