Update for the week ending on Friday, Jul 24, 2026

Refactoring English

• Published Why I Stopped “Creating Content”
  • Made it to HN front page
• Backported into the ebook improvements to the design doc section I made when I excerpted it for a web version
• Deleted the section on design reviews from the book
  • It felt too outside the scope of the book
  • It’s still available as a blog post
• Trimmed low-value text from passive voice chapter
• Started editing project for a reader
• Updated book release version to 0.9.9
• Ended early access discount
• Fixed some missing countries / incorrect calculations in PPP pricing

Feedback app

• Add chart-based analytics
  • Screenshot
  • Excluded the admin user (me) from analytics
• Fix book’s internal chapter list to sort in the same order as the book based on book’s “next chapter” navigation links
  • Fixed subtle bugs where chapters have tricky links that look like navigation links but aren’t
• Tried to do bluegreen deployment
  • Turns out it can’t work because the SQLite database sits on a volume, and each volume is bound to a single machine, so I can’t have the new machine take ownership of the SQLite database because it’s on another machine’s volume

Scraper bot attack

• Dealt with a scraper bot sending 50k/requests/day to the HN popularity contest
• Fortunately, they’re super easy to spot because they rotate through thousands of IPs but use the exact same rare user agent everywhere
• Unfortunately, Netlify doesn’t allow blocking by user agent without an edge function, so I’d pay for every edge function execution
• I tried throttling in Bunny to 1 KB/s based on user agent, which slowed the attack down but didn’t solve the problem
• I tried faking results in Bunny to trick the attacker into thinking there’s less to scrape, but that didn’t make much difference
• Made a custom script that generates IP blocklists based on bunny access logs

HN Popularity Contest

• Updated processing code to Python 3.14.4
• Added more unit tests
• Updated metadata and excluded a bunch more ineligible domains
  • The fewer the supported domains, the less there is for a scraper bot to scrape
• Modified my plausible config to record the domain for single-domain views
  • I didn’t realize the Plausible drops URL query params by default for privacy
  • Took a lot longer than I expected because it turned out that Plausible changed the JS include but didn’t notify me, so I was using the outdated JS include that didn’t support their documented way to include query params

Little Moments

Little Moments is an open-source web app I’m working on to replace the current family photo sharing app (TinyBeans) I used (which I strongly dislike).

• Simplified the e2e test fixture so that I could actually understand the SQLite database management logic
  • And added a clearer comment
• Fixed ESLint config to also lint inline scripts in template files
• Trimmed unused go test scripts and made the flag parsing stricter
• Ported improvements from go-app-starter
• Adjust capitalization in logInAsAdmin e2e test helper
• Reformatted the e2e test part of flake.nix

Weekly Snippets

• Ported git repo from GitHub to Codeberg
  • I had to delete a bunch of large images that bloated the repo beyond
• Ported CI from CircleCI to NixCI

keeperfx-zig

• I vibe-coded a Zig build of KeeperFX, the open-source reimplementation of Dungeon Keeper.
  • Works under Linux, but so does the real KeeperFX now.
  • Was mainly just curious if it would work and what it looks like to do LLM-translation of C modules to Zig.

NixOS config

• Updated my laptop to NixOS 25.11
• Figured out why all my USB devices stopped working
  • I at some point added a hardening step to disable auto-loading kernel modules, and I didn’t realize that disabled kernel modules for all my USB devices.
  • I made this change a couple of months ago, so I must not use USB devices often.

Misc

• Experimented more with self-hosted git
  • Tried self-hosted CI with Gradient but I think it’s too complicated to self-host