Kindle newsletter digest

One-liner: Newsletters pile up unread and get lost to distraction. Forward them to a dedicated address; a daily job cleans, summarizes, and categorizes what came in, and publishes it as one dated Issue — grouped by category, magazine-style. Phase 1 ships that as a webpage; Phase 2 sends it to Kindle too.

Origin: New idea, came from the user directly (not sourced from an old repo).

Phases:

1. Phase 1 — intake → process → publish to reads.aiste.io as a browsable Issue archive.
2. Phase 2 — take the same Issue and deliver it to Kindle via Send-to-Kindle (email an EPUB/doc to the Kindle’s personal address; Amazon auto-converts it, no custom device software needed).

Architecture (settled — see CONTEXT.md for domain terms, docs/adr/ for the full reasoning):

• Intake: a Mailgun/Postmark inbound-parse route on reads.aiste.io receives forwarded newsletters and immediately commits the raw email to the repo — no Gmail OAuth, no polling.
• Compile: once a day, a scheduled GitHub Action cleans, summarizes (bullet highlights, plus the full cleaned text kept for a “read full piece” expand), and categorizes everything received since the last run via the Anthropic API, and commits it as Markdown with frontmatter.
• Publish: Jekyll (GitHub Pages’ native build) renders the Markdown into the site. Full archive of past Issues, not latest-only.
• DNS: reads.aiste.io carries the GitHub Pages CNAME; intake is nested one level deeper at mail.reads.aiste.io with its own MX record (a CNAME and an MX can’t coexist at the same DNS name). Both added at Namecheap, no DNS host migration and no change to the existing aiste.io landing page.
• Categories are inferred per-article at compile time, not fixed per newsletter source — the same source can land in different categories on different days.

Phase 1 prototype scope — newsletters to build/test against:

• Stratechery
• Lenny’s Newsletter
• TAAFT (There’s An AI For That)
• HealthCare Huddle

Status: Raw, design fully grilled — architecture and phase split settled via /grilling. Next: create the dedicated repo and start Phase 1 prototyping.