How long is a blog post? Some people write short posts with one clear thought. Others write longer essays.

I tend to start out writing a short post and then get distracted by all the rabbit-holes I want to go down. Drafting my thoughts on blogging leads to adding lots of blogs to my reader, writing some code that takes an OPML and makes a table of blogs and their most recent posts, fixing the org-html-themes setup for my Emacs configuration, breaking out this chunk as its own post, drawing a bunch of mindmaps, doing a braindump, tweaking my workflow for processing braindumps to use faster-whisper and whisper-ctranslate2 instead of WhisperX because of this issue, so that I can try the whisper-large-v3-turbo model, experimenting with workflows for reviewing the PDF on the iPad… Definitely lots of yak-shaving (wiktionary definition). I still want to write that post. I already have the sketch I want to include in it. It's like Chilli in the Bluey episode Sticky Gecko (script): "The door: It is right here. All we need to do is walk out of it: it's so easy!" The thought! It's right there! Just get to it, brain! But I wander because I wonder. I suppose that's all right.

It might be fun to play around with the sizes of things I share: shorter when my attention is fragmented or squirrely, longer when I can think about something over several days or years. Here are some ways to tinker with that.

Breaking thoughts down into smaller chunks so I can get them out the door:

• When I notice that something is a big blog post (like this reflection I've been working on about blogging), I can break out parts of it into their own blog posts and then replace that section with links.
• I can post interesting quotes and snippets to Mastodon and then round them up periodically or refer to them in blog posts. TODO: It might be good to have a shortcut for an accessible link to a toot using a speech bubble or similar icon.

Taming my tangents and ideas: I'm sometimes envious of blogs with neat side notes, but really, I should just accept that the tangents that my mind wants to go on can take paragraphs and are more suited to, say, collapsible details or a different blog post. Something I can experiment with: instead of gallivanting off on that tangent (soo hard to resist when there's an idea for an Emacs tweak!), I can add a TODO and leave it for my future self. Maybe even two TODOs: one inline, where it makes sense in the text; and one in my Org Mode, with a link to the blog post so that I can go back and update it when (if!) I get around to it. Who knows, maybe someone might comment with something that already exists.

Saving scraps: It's easier to cut out half-fleshed-out ideas if I tell myself I'm just saving them somewhere. Right now I capture/refile them to a scraps heading, but there's probably a better way to handle this. Maybe I can post some thoughts to Mastodon and then save the toot URL. Maybe I can experiment with using Denote to manage private notes.

Connecting thoughts and building them up:

• I tend to write in small chunks. (TODO: I could probably do some kind of word-count analysis, might be neat.) Sketchnotes and hyperlinks might help me chunk thoughts so I can think about bigger things. I can link to paragraphs and text fragments, so I can connect thoughts with other parts of thoughts instead of trying to get the granularity right the first time around. The shortcuts I made for linking to blog posts and searching the Web or my notes are starting to help.
• I sporadically work on topic maps or indices. Maybe I'll gradually flesh them out into a digital garden / personal wiki.
• Sometimes I don't remember the exact words I used. Probabilistic search or vector search might help here, too. I don't need an AI-generated summary, I just want a list of related posts.
• I can figure out how to add backlinks to my blog, or simplify the workflow for adding links to previous posts. Maybe something based on this guide for 11ty or binyamin/eleventy-plugin-backlinks. I might need to write something custom anyway so that I can ignore the links coming from monthly/weekly review posts.

Connecting to other people's thoughts: For the purposes of conversation, it'll probably be good to let people know if I write something about their blog post. Doesn't happen automatically. Pingbacks and referrer logs got too swamped by spam a long time ago, so I don't think anyone really uses them. Idea: It might be neat to have something that quickly lists all the external links in a post, and maybe a way to save the e-mail addresses or Mastodon handles for people after I look them up so that I can make that even smoother, and some kind of quick template. I can send email and toot from within Emacs, so that's totally doable… (No, I am not going to write it right now, I'm going to add it to my to-do list.)

(Also, there's another thought here about books and The Great Conversation, and blogs and smaller-scale conversations, and William Thurston and mathematicians and understanding, and cafes…)

Hmm. I think that getting my brain to make smaller chunks and get them out the door will be a good thing to focus on. Synthesizing can come later.

Related:

• Old-school blogger

See discussion on Mastodon

Nudged by Dave Winer's post about old-school bloggers and my now-nicely-synchronizing setup of NetNewsWire (iOS) and FreshRSS (web), I gave Claude AI this prompt to list bloggers (with the addition of "Please include URLs and short bios.") and had fun going through the list it produced. A number of people were no longer blogging (unreachable sites or inactive blogs), but I found a few that I wanted to add to my feed reader.

Here is my people.opml at the moment (slightly redacted, as I read my husband's blog as well). This list has some non-old-school bloggers as well and some sketchnoters, but that's fine. It's a very tiny slice of the awesomeness of the Internet out there, definitely not exhaustive, just a start. I've been adding more by trawling through indieblog.page and the occasional interesting post on news.ycombinator.com.

It makes sense to make an HTML version to make it easier for people to explore, like those old-fashioned blog rolls. Ooh, maybe some kind of table like indieblog.page, listing a recent item from each blog. (I am totally not surprised about my tendency to self-nerd-snipe with some kind of Emacs thing.) This uses my-opml-table and my-rss-get-entries, which I have just added to my Emacs configuration.

(defun my-opml-table (xml)
  (sort
   (mapcar
    (lambda (o)
      (let ((latest (car (condition-case nil (my-rss-get-entries (dom-attr o 'xmlUrl))
                           (error nil)))))
        (list
         (if latest
             (format-time-string "%Y-%m-%d" (plist-get latest :date))
           "")
         (org-link-make-string
          (or (dom-attr o 'htmlUrl)
              (dom-attr o 'xmlUrl))
          (replace-regexp-in-string " *|" "" (dom-attr o 'text)))
         (if latest
             (org-link-make-string
              (plist-get latest :url)
              (or (plist-get latest :title) "(untitled)"))
           ""))))
    (dom-search
     xml
     (lambda (o)
       (and
        (eq (dom-tag o) 'outline)
        (dom-attr o 'xmlUrl)
        (dom-attr o 'text)))))
   :key #'car
   :reverse t))

(defun my-rss-get-entries (url)
  "Return a list of the form ((:title ... :url ... :date ...) ...)."
  (with-current-buffer (url-retrieve-synchronously url)
    (set-buffer-multibyte t)
    (goto-char (point-min))
    (when (re-search-forward "<\\?xml\\|<rss" nil t)
      (goto-char (match-beginning 0))
      (sort
       (let* ((feed (xml-parse-region (point) (point-max)))
              (is-rss (> (length (xml-get-children (car feed) 'entry)) 0)))
         (if is-rss
             (mapcar
              (lambda (entry)
                (list
                 :url
                 (or
                  (xml-get-attribute
                   (car
                    (or
                     (seq-filter (lambda (x) (string= (xml-get-attribute x 'rel) "alternate"))
                                 (xml-get-children entry 'link))
                     (xml-get-children entry 'link)))
                   'href)
                  (dom-text (dom-by-tag entry 'guid)))
                 :title
                 (elt (car (xml-get-children entry 'title)) 2)
                 :date
                 (date-to-time (elt (car (xml-get-children entry 'updated)) 2))))
              (xml-get-children (car feed) 'entry))
           (mapcar (lambda (entry)
                     (list
                      :url
                      (or (caddr (car (xml-get-children entry 'link)))
                          (dom-text (dom-by-tag entry 'guid)))
                      :title
                      (caddr (car (xml-get-children entry 'title)))
                      :date
                      (date-to-time (elt (car (xml-get-children entry 'pubDate)) 2))))
                   (xml-get-children (car (xml-get-children (car feed) 'channel)) 'item))))
       :key (lambda (o) (plist-get o :date))
       :lessp #'time-less-p
       :reverse t))))

(my-opml-table (xml-parse-file "~/Downloads/people.opml"))

I'm rebuilding my feed list from scratch. I want to read more. I read the aggregated feeds at planet.emacslife.com every week as part of preparing Emacs News. Maybe I'll go over the list of blogs I aggregate there, widen it to include all posts instead of just Emacs-specific ones, and see what resonates. Emacs people tend to be interesting. Here is an incomplete list based on people who've posted in the past two years or so, based on this work-in-progress planetemacslife-expanded.opml. (I haven't tweaked all the URLs yet. I stopped at around 2023 and made the rest of the elements xoutline instead of outline so that my code would skip them.)

(my-opml-table (xml-parse-file "~/Downloads/planetemacslife-expanded.opml"))

Making this table was fun. It's nice to see a lot of people also writing and learning out loud. This reminded me a little of EmacsConf - 2020 - talks - Sharing blogs (and more) with org-webring. TODO: Could be fun to have a blogroll page again.

I notice I tend to like:

• posts about adapting technology to personal interests, more than posts about the industry or generalizations
• detailed posts about things I'm currently interested in (Emacs, personal knowledge management, some Javascript), more than detailed tech posts about things I've decided not to get into at the moment
• "I" posts more than "You" posts: personal reflections rather than didactic advice
• curiosity, fun, experimentation

Looking forward to discovering more!

Related:

• Old-school blogger

See discussion on Mastodon

I was curious about the shape of my blog over the years, excluding Emacs News and my link-heavy weekly/monthly reviews. It started off with lots of little posts like the way other weblogs were also quick links and notes. As weblogs morphed into blogs with more text, I also settled down into fewer, longer posts with lots of code (analyzed by looking for <pre> blocks). I wrote much less after A+ was born. Interestingly, I've been shifting towards longer posts with more images.

• Blog posts exclude permalinks that match emacs-news|review|week-ending, which casts a bit of a wide net but should give me the general shape of things.
• Total words per year and average words per post both exclude code snippets.

Here's how I got those numbers:

(append
 '(("Year" "Posts" "Total words" "Words per post" "Posts with pre" "Posts with images")
   hline)
 (cl-loop for i from 2001 to 2024
          collect
          (let* ((default-directory (expand-file-name (number-to-string i) "~/proj/static-blog/blog"))
                 (exclude (shell-quote-argument "emacs-news|review|week-ending"))
                 (files (format "find . -name '*.html' | grep -v -e '%s' | " exclude))
                 (posts (string-to-number
                         (string-trim
                          (shell-command-to-string (concat files "wc -l")))))
                 (words (string-to-number
                         (replace-regexp-in-string
                          "TOTAL: " ""
                          (shell-command-to-string
                           (concat files "xargs ~/bin/count-words | grep TOTAL")))))
                 (posts-with-images
                  (string-to-number
                   (string-trim
                    (shell-command-to-string (concat files "xargs grep -l '<img' | wc -l")))))
                 (posts-with-pre
                  (string-to-number
                   (string-trim
                    (shell-command-to-string (concat files "xargs grep -l '<pre' | wc -l"))))))
            (list i
                  posts
                  words
                  (/ words posts)
                  posts-with-images
                  posts-with-pre))))

And here's how I plotted the charts:

import pandas as pd
import matplotlib.pyplot as plt
df = pd.DataFrame(data[1:], columns=data[0])
# Create a figure with subplots
fig, (ax1, ax4, ax2, ax3) = plt.subplots(4, 1, figsize=(10, 12))
fig.suptitle('Blog Statistics by Year', fontsize=16)

# Plot Posts
ax1.bar(df['Year'], df['Posts'], color='lightblue', label='Other posts')
ax1.bar(df['Year'], df['Posts with pre'] , color='darkblue', label='With preformatted blocks')
ax1.set_title('Number of posts per year')
ax1.set_ylabel('Posts')
ax1.legend()

# Plot Posts
ax4.bar(df['Year'], df['Posts'], color='lightblue', label='Other posts')
ax4.bar(df['Year'], df['Posts with images'] , color='darkgreen', label='With images')
ax4.set_title('Number of posts per year')
ax4.set_ylabel('Posts')
ax4.legend()

# Plot Total Words
ax2.bar(df['Year'], df['Total words'], color='lightblue')
ax2.set_title('Total words per year')
ax2.set_ylabel('Total words')

# Plot Words per Post
ax3.bar(df['Year'], df['Words per post'], color='lightblue')
ax3.set_title('Average words per post')
ax3.set_ylabel('Words per post')
ax3.set_xlabel('Year')

# Adjust layout and display
plt.savefig(f)

I want to write more, and I want to enjoy going through my archive. Some posts are long, especially those that come from transcripts. If I sat with the ideas for longer, I might be able to make them more concise or break them up into more atomic notes; but I also want to get things out faster in order to learn from potential conversations. So I'm thinking about text structure and margins, since I want to re-read my blog more and I sometimes glaze over when there's lots of text.

More headings are a good start. Org Mode makes it easy enough to add them: M-RET calls org-insert-heading.

I'm experimenting with sticky tables of contents on large screens: one for "on this page" on the left, and one for long posts on the right.

On the individual post page, it'll just be the table of contents for the post, like this one.

It feels a little busy. If I write some Javascript, I might be able to use IntersectionObservers to highlight where we are. Maybe I can even squeeze the article's TOC into the "on this page" TOC if there is one, which means it can stay on one side.

I want to do other things with the margins. Doodles for fun? My cargo bike post started with the doodles pulled all the way into the margins, and then I moved them back into the text so that I don't have to worry about bumping into the table of contents.

Sometimes I use a sketchnote to help me think through or summarize a topic. It might be fun to use the sketchnote as a table of contents or overview, maybe even highlighting different sections of it as I scroll. handwritten.blog uses mix-blend-mode for hyperlinks. If the SVG isn't too big, maybe I can use the same kind of technique I used in animating SVG topic maps with Inkscape. Alternatively, I could put extracted regions from the sketchnote in the margins for context and visual variety.

Sidenotes? I like how A Scripter's Notes has both an active, expanding TOC on the right as well as side notes on the left.

Karthinks uses a sticky TOC and sidenotes:

A Blog With Relevant Information uses just sidenotes, so the rest of the page feels pretty clear:

Maybe keywords, like the Cornell method of note-taking? Kind of like sidenotes, but more structural, for skimming. I'm having a hard time finding a blog example, though. If I figure out side nodes, I could probably just use a different style to indicate those Cornell-style cues.

But there's so much more I want to do with the space. I like the stacking of https://notes.andymatuschak.org , and I like that you can link to a particular stacked state.

Then every so often, I come across a blog that is just clean and refreshing and then I want to get rid of everything in the margins.

There are plenty of CSS and JS resources out there. Figuring out what I want is the tough part.

Nudged by Thierry Stoehr's toot about my 23rd blogiversary, I've been thinking about how much I've learned thanks to blogging, and how I can get even better at learning out loud. I'm curious about what this could become over the next twenty years, when I'm in my sixties.

The first part of the text from the sketch is duplicated and expanded in the list below. There are a lot of different aspects I want to get better at, so I'm not going to try to work on all of them in one go, but it's fun mapping out so much room for growth.

Besides, maybe one of these aspects will resonate with you as either something you're learning or something you've figured out something about, and then you'll get in touch, and then we'll both learn more. Wouldn't that be cool?

I'm experimenting with getting stuff out in smaller chunks, so this is part 1 of 4: Starting.

I want to get better at making sense of things and sharing what I'm learning. Nudged by Chris Maiorana's post on Second Brain, Second Nature, I borrowed How to Take Smart Notes by Sönke Ahrens (2017). Here are my notes.

The book goes into detail about Niklas Luhmann's Zettelkasten or slipbox system. Lots of people have written about Zettelkasten and various implementations. There's even a whole micro-industry around Notion templates. So I won't spend a lot of time right now describing what it is or what the key aspects are. I can focus instead on what that means to me and what I want to do with it.