Reworking the blog, and adding alerts before they are needed

Most of the last two days went into one file: src/content/blog.ts. We reworked all posts against a single standard — bounded language about what AI actually does, a worker-first framing, and German that reads as native German rather than a translation. That last point matters more than it sounds. A rendered blog is only as trustworthy as its weakest sentence, and translated-sounding copy quietly costs the reader their confidence. So the German posts went through a proper register pass, quote-style and grammar included, until they read the way a German writer would have written them in the first place.

Alongside the copy, we did some housekeeping that readers will never see. Blog entries now carry sensible workday dates, a duplicate post was removed, and AI-coauthored posts are marked with a clearer icon. On the operations side, we added alerting on the dispatch ledger for three failure modes — runaway spikes, tier misuse, and cost surges. It watches the numbers so a person does not have to stare at them.

None of this is dramatic. It is the ordinary work of keeping a site honest: say what is true, say it clearly in every language you publish, and put a tripwire in front of the mistakes you can predict.