a browser is a very interesting piece of software because well let me explain why I just thought of this and what the connection is
so basically I realized that in a similar way that like websim is powerful [and] useful as a method to interface with a language model, both for you and then from the other end for Claude, or for whatever the model might be.
it’s a powerful interface.
so is the web browser itself, obviously.
obviously, web browsers have completely changed how humans interface with computers and how it feels to use code as an end user, or like to use the product of code, right?
but if you’re a user who could write code, then building a tool for yourself in the browser, or not in the browser, like the point is that like, the browser changes not just the end user who doesn’t codes relationship to the computer, but it changes the same thing for a developer as well.
workflow
so like, just now I noticed that something that I often do I do almost every time that I that I push a commit that is the result of like some you know like an hour of work or something or more
almost every time I do that I it’s the result of having created an issue in Linear and you know maybe some comments maybe a lot of comments maybe it’s had a lot of work done in the issue in addition to some ChatGPT chat or something
and the thing that I often do is that when I’m done with that task I will
push the commit
and I will go to GitHub and find the commit in my browser and copy the URL to the commit
and go back to linear and paste the URL, right?
but in the process I also
a few things I check,
I see like if that was the first commit for the day I’m like cool, I filled my square
and I review the commit in GitHub, which is like a second stage of review after seeing the code in whatever the editor might be, like the Git editor, or just in the code itself.
that’s like the first pass, and then the next pass is in GitHub, where I might then realize some issue or something, or some change I want to add on.
and so the thought is, this could be automated with some scripts.
I could make a script that would allow me to just like, I mean, I could, and that’s obviously what linear is.
one of its main things is that you can integrate with GitHub issues and things.
so it can do it at that level
but I think what I would want to do rather than that is like have a command in my command line that I just type or actually like directly in my git editor thing my git ui I would like to have just a key that I can press and copy the url to the github commit just made and then just immediately paste that into the issue or even go as far as like just add it as a comment onto the issue like something like that
insight
but there’s something that you lose there in the same way that if you’re prompting Claude in the Claude UI or over the API or with some interface that is different than using websim because it allows you to express to Claude in a different way
you have this like box yes but also you have like sometimes your prompt is not in the websim url bar but it’s in some box that Claude has written whether you asked for it or if Claude decided to put it in
and also you might be changing you know like html objects on the page and then sending submitting some form and that’s your prompt, right?
and Claude gave you more bandwidth to express back to Claude.
sure, you could have expressed in high bandwidth using your own words, using your brain, exerting that effort to think of something and express it.
but the medium of the browser has allowed you to you, as in both you and the model to express yourself in a different way and for the model to express itself it’s like a feedback loop almost, it could be, it can be
comparison
and in the same way the user, myself and GitHub there’s certain feedback loops that occur when I do that
sometimes I get, I will get distracted even I’ll go on GitHub and like just browse stuff for like too long, maybe I won’t, I’ll stop working like right now, it’s kind of what I’m doing
but there’s something to be said for that.
conclusion
that’s the point.
browsers are interesting, websim is interesting.
it’s not necessarily more or less efficient than using Claude.
it’s just different.
it’s like when computers were invented, everything was in the command line and then we had UIs and GUIs and eventually the web browser.
it’s not that one is better than the other.
in fact, I mean, they have different strengths.
but it’s not just about strengths, it’s that, like, there’s literally, it’s like, it’s like a sculpture versus a painting, like, you know, or versus, like, a scrawl, a sketch on a notepad.
like, those are totally different mediums, and you cannot express many things, at least not in the same way, and without the same richness, without the same dimension, without moving between those mediums, so.
outro
yeah, I don’t know, it’s like both a realization about browsers, it’s a realization about language models, it’s a perspective, it’s a thought.
that I’m going to now transcribe with my thing and I’m gonna see how long that takes me.
I’m gonna like literally try to stay completely focused on just that task.
right now the time is 8.31, I’m gonna press stop, I’m gonna go, I’m gonna transcribe the file, I’m gonna process the file, I’m going to open the file in my text editor as a Markdown preview thing in my browser, screenshot it and send it to you.
computers are amazing, aren’t they?
there’s a lot of stuff there.
that you have like this conscious control over like an instrument like you know like a saxophone is the one that comes to mind because it has keys and it feels expressive in a certain way that like the guitar doesn’t quite feel it’s more I mean but I guess you can see you can see them similarly anyway it’s an instrument computers are instruments that’s cool okay bye