A context provider, not a blog

Most industry writing is published once and left to rot. This is built the other way round: a small number of durable explanations of how things work, and a dated stream of what changed, kept separate on purpose.

How it works

A pipeline follows a set of named sources per topic — publisher feeds, primary documentation, podcasts and practitioner communities. New items are de-duplicated and screened, then summarised into a short analysis with an explicit impact rating and, where there is one, a recommended action. Those are collected into a brief for the day they were published.

The guides are separate. They are written from primary documentation, and they are revised when the stream repeatedly contradicts them — rather than accumulating addenda.

What it does not do

It does not republish other people’s reporting. Every story is a short original analysis that links prominently to the source, and those outbound links are the point rather than a footnote. If you want the detail, the link is there and you should click it.

Reading it as a machine

Every page is static HTML with schema.org structured data. There is an RSS feed, a sitemap, and an llms.txt describing the site for language models and agents. Nothing requires JavaScript to read.

Corrections

Anything here can be wrong, and the frontier sections are the most likely to be. Where practitioners and a platform disagree, both positions are stated rather than resolved.