How the briefing
gets made.

Everything starts with 80+ hand-picked sources — vendor engineering blogs, research labs, trade press, and practitioner writing. The list is maintained by a human: sources earn their place by being consistently useful and get removed when they stop being. We don't scrape the open web.

A pipeline crawls those sources on a fixed schedule throughout the day and stores every new article it finds. Collection is fully automated — no story is typed in by hand, and nothing is collected outside the curated list.

Each collected article is read by Claude, an AI model from Anthropic. It scores relevance for our audience on a 0–100 scale, assigns the story to one of our six desks, and pulls out the key facts and figures worth surfacing. The scoring prompt encodes our editorial brief: enterprise and tech AI, practitioner-relevant, signal over volume.

Only stories that clear a relevance threshold make it onto the site. Most collected articles never surface — that filtering is the product. When a day produces too little that clears the bar, that day's edition stays thin or empty. We never pad an edition to look busier than the news actually was.

Cleared stories are assembled into a dated daily edition — one anchor story plus briefs — that becomes a permanent archive page. A human reviews the published output and the pipeline's choices on a regular cadence, adjusting sources, thresholds, and prompts. We do not claim every individual story is hand-checked before it appears; the human work is in tuning the system and correcting it when it's wrong.