About Yesterday in AI
Useful news. Room to think.
A daily field guide for the people deciding how AI fits into real organizations.
Yesterday in AI is a publication of Ted Tschopp’s Public Workbench. It follows enterprise AI through the decisions it creates: what to build, what to buy, what to secure, and what to question.
The daily edition opens with a short briefing, then makes room for researched analysis. A separate live wire keeps the wider news landscape in view. The name is Yesterday in AI; the logo reads YesterdAI.
Editorial method
Discovery starts with the curated AI news feed and ranked Top Stories feed. Supplied summaries and ratings help identify useful stories. We then follow the reporting, check dates and claims, and link the evidence.
The morning edition is anchored at 5 a.m. Pacific and looks back 36 hours. Original publication or a verified substantive update determines eligibility. A date-only feed field is not evidence of an exact publication time. Stories with uncertain timing at the boundary stay out of full coverage.
Each edition includes up to eight analyses, normally 800–1,200 words each. We publish fewer when the available evidence calls for it. Overlapping windows do not justify repeating a story; a new article about the same subject needs a substantive development.
Where AI fits
Codex assists with research, writing, editing, and publication. Every full article includes an original AI-generated editorial illustration. Images illustrate an idea; they are not documentary photographs. Bylines identify the publication rather than implying personal review of every automated edition.
The editorial process distinguishes reported facts, attributed claims, and our analysis. Vendor benchmarks and forecasts remain claims, not independent validation. Links let you read the underlying material and make your own assessment.
The historical launch editions
The first seven editions were reconstructed from historical feed snapshots. Their reporting cutoffs describe the period covered. Their preparation and article publication dates record when the work was actually created. They do not claim to have been published in the past.
Read, correct, inspect
Read through the website, RSS, or JSON Feed. The source repository contains the site, articles, image prompts, source records, and publishing instructions. Report a correction through GitHub Issues.
The live feed refresh runs hourly on GitHub. Daily editorial production runs through a local Codex scheduled task; it depends on the host computer and app being available. If a production run fails, the last verified edition remains online.