How to Automate Newsletter Based on SecondBrain and Bluesky Changes

I automate the gathering of content for my Newsletter, but not the writing of it. Everything mechanical that I used to do by hand — pulling links to recently-updated Second Brain notes, finding books I’ve read, copying top Bluesky posts — runs from a small Python script. The personal writing stays mine.

Inspired by Simon Willison’s Substack automation see also Simon Willison, but where Simon does it fully programmatically with SQL and Datasette into Substack, I go markdown-first into Listmonk via my listmonk-rss repo. No notebook, no AI in the loop.

GitHub Repo

Find at sspaeti/listmonk-rss. This repo also contains an automatic newsletter sent via listmonk-rss.py based on new blog posts (documented at Listmonk), as well as the automation newsletter.py and templating.

This is not to replace my own writing with my subscribers, but for me as I do so much work on my second brain, to surface latest changes, to give the readers a possibility to keep up with what has changes since the last email in a very simple and easy to scan way, and for me to reduce the workload to go through all my changes and posts manually.

A generated email with gathered information from my second brain, my recent book notes I read or currently reading, from Bluesky post and engagment (can be pulled via DuckDB and simple query - see Querying Bluesky with DuckDB and SQL):

With make newsletter-send it will be scheduled on Listmonk automatically:

1
2
3
4
5
6
7
8
9

make newsletter-send
uv run python newsletter.py send drafts/newsletter-2026-06-02.md
INFO:httpx:HTTP Request: GET https://list.ssp.sh/api/lists "HTTP/1.1 200 OK"
INFO:httpx:HTTP Request: POST https://list.ssp.sh/api/campaigns "HTTP/1.1 200 OK"
Campaign draft 97 successfully created!
INFO:httpx:HTTP Request: PUT https://list.ssp.sh/api/campaigns/97/status "HTTP/1.1 200 OK"
Campaign 97 successfully scheduled with 45 mins delay!
INFO:httpx:HTTP Request: POST https://api.pushover.net/1/messages.json "HTTP/1.1 200 OK"
✓ .last_newsletter advanced

And this is how it looks:

All in beautiful Markdown - and the preview looks like this - beautiful, no? 🙂

The script (newsletter.py) has two commands: gather and send.

gather collects, since the last sent newsletter:

• Blog posts: capped at 2 — RSS auto-announces them via the older listmonk_rss.py workflow already.
• Brain notes: git log --numstat inside the second-brain-public/content/ submodule. A line-count threshold (default 20 added lines) filters out commits that are just grammar fixes. Notes are split into Major (≥100 lines added) and Smaller updates, each with a placeholder paragraph for me to write framing prose.
• Books: scans my Books/ folder for any book whose Created, Started reading, or Finished reading date falls in the window, and pulls the > callout and ## Notes During Reading. The folder is copied to a local .copy/books/ snapshot first so the script never touches my live vault.
• Bluesky: DuckDB query against app.bsky.feed.getAuthorFeed, top 15 by engagement. See DuckDB + Bluesky AT protocol. see ATProto queries

Output is a markdown draft with editorial slots (intro, per-section framing, closing). I edit in Neovim, then make newsletter-send pushes to Listmonk as a scheduled draft so I get a Pushover notification and a final review window.

• No AI in the loop. AI would flatten my voice. The script gathers; I write.
• Date advances only on send. Re-running gather between sends always shows the same backlog — nothing is consumed until I actually send. If I skip a few weeks, the next draft has all of it.
• Filename-as-slug for URLs. Hugo derives URLs from the filename, not the frontmatter title, so the script slugifies the file stem (e.g. zen mode for writing.md → /brain/zen-mode-for-writing/, even when the title is “Zen Mode for Writing (Obsidian, Neovim)”).
• Wikilink handling depends on source. Brain-note intros keep wikilinks converted to ssp.sh/brain/ URLs (they’re all published). Book notes get wikilinks stripped to plain text — they may reference notes in my private vault that don’t exist on the web.

1
2
3
4
5

make newsletter                        # gather a fresh draft
make newsletter-since SINCE=2026-04-01 # custom window
make newsletter-edit                   # open latest draft in $EDITOR
make newsletter-send                   # push latest draft to Listmonk
make bsky-engagement                   # ad-hoc: print top Bluesky posts

• Listmonk
• Neomd

Origin: ssp.sh Newsletter Drafts, Simon Willison
References: Newsletter