I spent a long time tinkering with the tooling, which meant that writing always took a back seat or was put off. As a transition, I decided to use Bear Blog [0] for writing, and when I eventually find a self-hosted solution that works for me, I'll just switch over. And Bear Blog is in line with my values, unlike so many other platforms.
I also came with a similar resolution which is that building just for the sake of building will not motivate me to actually build.
I'm adding CSS, JavaScript, PHP, whatever, to my website because it's required to accomplish what I want. Not because my website must have all of that to be on par with what's whateverGPT is able to produce.
Building a CI pipeline which will automatically convert the linted markdown I'm pushing to a beautiful HTML blog post will surely be helpful to my technical growth.
But I know for sure that I won't have any motivation to do it without having written an actual post to share to the internet.
People worry about tooling because they don't want to create a future mess they have to unpick: or the process might be hard enough they just won't do it.
For my private blog for example, how to easily - as in drag and drop - insert images was a big thing that needed to work. So was reasonable code rendering.
I settled on the requirement "must be able to publish a Jupyter notebook" since that format roughly handles all those requirements while still being markdown mostly.
Then you have hosting: I want low touch so how whatever I pick interacts with that matters a lot too suddenly.
It still took some tweaking of Nikola[1] to get the process right for me, though not much.
Keep it simple doesn't help much if simple is actually just missing features that enable you to write efficiently and often (if that's what you want to do, I think I mostly just keep technical notes for myself).
I totally agree that the most important thing is to write more and curate our own parts of the internet.
I think a lot of these things become simpler as AI abstracts and commodifies the functionality and code required to develop and ship new functionality.
> If you truly dislike writing HTML, that is fine too. Write in Markdown, AsciiDoc or whatever plain text format you find pleasant and convert it to HTML using Pandoc or a similar tool.
Funny, but note that Knuth had already published three volumes of TAOCP and a second edition of Volume 1 (1968–1973), won a Turing Award for them (1974), and had the second edition of Volume 2 in galley proofs, and even then it was only a combination of two factors (the publishers moving away from hot-metal typesetting to phototypesetting with a decline in quality, and the emergence of digital typesetting that he felt more comfortable handling) that led him to take up the problem. And even then, he simply wrote down a design and left it to a couple of grad students to implement over the summer while he was gone, and it was only when he came back and saw their (limited) progress that he realized the problem was harder than two good Stanford grad students could handle, and decided to take it up himself. And even then he basically started in mid-1977 and was done in a year or so (TeX78 and MF79), and only when it became very popular and incompatible ports started cropping up that he decided to (re)write a “portable” TeX and METAFONT himself (1980–1982, ??–1984). And after that there was a constant stream of feature requests so he decided on an “exit strategy” and froze the programs. And continued to do research and publish papers on the side during the years he was working on TeX/MF “full-time”.
So yeah the moral I guess is, tooling may take longer than you expect, but you must at least be trying to get away from it and back to writing. :)
Yes I like that page too, and I guess I was responding to it from memory, more than to your comment :) There's an element of truth to it but also a misunderstanding, so the story can be told in both ways. Maybe I should write this up as a webpage/blog post.
> Are reposts ok? If a story has not had significant attention in the last year or so, a small number of reposts is ok. Otherwise we bury reposts as duplicates.
[0]: https://bearblog.dev/
reply