Editorial policy

How posts are made

Daily digests, weekly trend essays, newsletter issues and podcast scripts are produced by an automated pipeline. Each post says so under the byline, in a “How this was made” note.

  1. Sources are collected automatically from research labs, company engineering blogs and paper feeds. An AI model (currently OpenAI GPT) picks the items and writes the drafts. The text you read is AI-generated.
  2. Daily digest items are checked against the source's full text. Numbers must appear in the source, items that cannot be verified are dropped, and a source is not repeated across issues.
  3. Weekly essays cite at least eight primary sources, and their numbers table is checked against those sources.
  4. Podcast scripts are fact-checked and approved by the editor before recording.
  5. The newsletter is built from these posts.

Who is responsible

Posts are published under the byline Applied Signals Editorial, a pseudonymous editor. The editor runs the pipeline, sets its rules and is accountable for what it publishes, including corrections.

Sources

  • We work from primary sources: papers, official engineering blogs and release notes. Each digest item links to the source it summarizes, and weekly essays cite at least eight primary sources.
  • Numbers in daily digests and in the weekly numbers table are checked against the cited sources.
  • Summaries can still be wrong or miss context. Read the linked source before relying on a result.

What we do not publish

  • Investment advice
  • Hype language or marketing summaries

Sponsorship and affiliate links

Sponsored placements are clearly labeled, and sponsors have no input on editorial content. Affiliate links are marked.

Corrections

When a post is wrong we fix it and record the change on the corrections page. To report an error, email [email protected] with the page link and what is wrong.