← Feed

Posts

• View post

  Did some blogging this week. First was a look at the uptime of various GitHub Pages alternatives. I've been plotting to change platforms for a while and ran a simple test to track uptime. https://alexsci.com/blog/static-hosting-uptime/

• View post

  I built a serverless web clock this week. A Cloudflare header transform rule adds the time at the CDN edge. https://clock.alexsci.com/

• View post

  I do front-end projects just infrequently enough to constantly get bit by framework churn. I just found out that create-react-app has sunset.

• View post

  I set up a landing page for Blog Quest today. Bulma keeps being my go-to CSS framework for quick projects like this. https://blog-quest.alexsci.com/

• View post

  About a month ago I noticed that NIST's time.gov website had a measurement error that caused it to miscalculate the offset of your device's clock relative to UTC. I reported the issue and they've released a fix. Full details in my latest post: https://alexsci.com/blog/how-time-gov-works/

• View post

  @sethmlarson has a clever approach to generating his human.json, which is to automatically generate it from the RSS feeds he follows. https://sethmlarson.dev/ive-added-human-dot-json-to-my-website

• View post

  J3 is a git repository backed by Amazon S3 storage. I think it's a really clever approach to rethink how git data is stored and it should scale exceptionally well using that foundation. https://code.lord.io/j3/

• View post

  Woah, I never considered how many DNS queries you could generate from a single lookup. > Ondřej’s example was teams.microsoft.com, which took his Bind resolver running 9.16 247 queries! https://blog.apnic.net/2026/06/02/cold-start-dns/

• View post

  @mnot here is a Feed Menu implementation that lists Atom feeds for IETF Mailing Lists. https://ietf-feed-menu-demo.alexsci.com/ In the future, IETF could host their own, but Feed Menu seems to support curation on unrelated domains.

• View post

  It's interesting to see AI maximalists interpret people's aversion to reading LLM output as a writing-style problem. They say: people will like the LLM output if you prompt it to avoid using tropes associated with LLMs. I see LLMs as digital con men: lying or hallucinating with confidence, well-written prose, and smile. I don't want LLM output that hides the tell-tale signs of AI generation any more than I want a con man to avoid sounding sleezy. I just don't wa...

• View post

  It's wild how poor Google search has become for power users. A search like `site:feeds.bbci.co.uk inurl:rss.xml` finds tons of pages on duckduckgo but zero when I try Google. Seems like an unforced error by the historical king of search.

• View post

  I've been using the Tranco list (https://tranco-list.eu/) for some recent work and was surprised to find how much churn the lists experience. Comparing 74V4X (August 25th) to JZ2VY (July 1st) I see 5.7M domain names total, but only 3.1M occur in both lists.

• View post

  I realize I've been reading the LLM disclaimers wrong. When they tell you to "double check" answers they don't mean "perform a second check", as no first check was ever performed. It's a liability disclaimer asking you to check it once and check it twice.

• View post

  I just noticed that Google News has search-based #RSS feeds, like https://news.google.com/rss/search?q=devops I don't love that they rewrite all the URLs for tracking, but that's their style I suppose.

• View post

  I track git forge web hosting at https://stats.uptimerobot.com/g9hjQmdr0w GitHub, closely followed by GitLab, performed the best. Bitbucket gives two-nines of uptime, with Codeburg at only one-nine. My tangled page has been down for months, although tangled is still an "alpha" release.