pytebeat VS tlssh

In this way I estimate I create about 32 new programs a day for only my own personal use. But in many cases I only use each one once. ^R makes it easy to use them again a few times, but in other cases I save them into a shell script to make them easier to distribute to other machines, parameterize, and maintain. The one I most often use is probably a user interface for YouTube via youtube-dl or yt-dlp that consists of a few such shell scripts.

My main editor is Emacs. If I want to do the same thing repeatedly (e.g., delete a line containing the string .LVL) I create a keyboard macro with F3 and F4 (or C-x ( and C-x )) when I do it once, then run it repeatedly with C-x e. I probably write about 4 programs a day in this way.

Emacs has a M-: command to evaluate Lisp expressions, which are programs. Recent programs I have written in this way include (/ 43.2 1.7), (* 9.3 1.2), and (+ 8 3 2.50 3.50 3 7 3.50 3 1.50 4.50 6 5.50 6.50 6 3 2.50 2.50). Probably I write about 1 program a day in this way but I only use each one once. Longer Lisp programs like this can be written in scratch and executed with C-j or in .emacs (or .emacs.d/init.el) and executed with C-x C-e. For example, (global-set-key [f5] 'recompile). I use my .emacs file constantly every day but probably only add something to it about once a month. An outdated version is at https://github.com/kragen/kragen-.emacs.d/blob/master/init.e....

Sometimes I write bigger programs for my own personal use too.

A few years ago I wrote https://github.com/kragen/pytebeat for a livecoding performance of bytebeat in a bar. I finished writing it in the train on the way to the bar.

In https://github.com/kragen/laserboot I wrote a simple parametric 2-D CAD system for laser cutters in PostScript.

For Dercuano http://canonical.org/~kragen/dercuano I wrote a kind of shitty HTML rendering engine that generates a PDF file, as well as a simple CMS for generating a tree of HTML files from a directory of Markdown.

The other night I wrote a bytecode interpreter with a graphical display in C as a sort of mockup for the operating system of a small computer I recently got the parts for; it's in http://canonical.org/~kragen/sw/erika.git.

I've also written compilers, interpreters, ray tracers, database engines, parser generators, graphics libraries, logic circuit optimizers, 2-D game engines, etc., for my own use. Some of the recent ones are in http://canonical.org/~kragen/sw/dev3.

* My own RSS reader (not published. It will never be end-user friendly enough to compete with other ones. But it's better for me)

And then plenty more than I use occasionally, and some I no longer use. E.g. for a while I used my own SSH replacement, in order to get TPM-backed keys (https://github.com/ThomasHabets/tlssh). Nowadays I use yubikey instead (https://blog.habets.se/2016/01/Yubikey-4-for-SSH-with-physic...).

Those are just the main ones (as in not small, and used every day). I find myself fixing problems all the time by writing code.

A curious question. Aside from my day job, this seems like a thing I do all day. :-)

I'm not sure what the motivation for your question is. Do you feel like everything's been invented and built already, and it's just a matter of (at most) plugging the things together?

I find myself constantly thinking "this should exist". I don't have time to make them all exist.

https://github.com/ThomasHabets/arping

Nothing like it existed at the time, and I wanted to send ARP requests as easily as sending ICMP ping.

https://github.com/ThomasHabets/simple-tpm-pk11

I wanted to use a TPM chip for SSH client keys, and couldn't find anything like it.

https://github.com/ThomasHabets/tlssh

I wanted to explore what it would be like to have SSH, but with identities not based on providing username, but an x509 cert. (and TPM chip protecting the key)