How It Works

1. Sources

Twice a week (Tuesday and Friday mornings), the system searches the following sources for new opinions, filings, and commentary:

Federal Court Opinions

• CourtListener API — Searches the full CourtListener opinion database using a battery of keyword queries covering Section 230, First Amendment, content moderation, platform liability, AI, and related topics. Returns published opinions from all federal courts.
• RECAP Archive — Also via CourtListener, searches RECAP for district court filings (motions, orders, pleadings) that may not yet have published opinions but are actively litigated.
• Federal Circuit RSS Feeds — Monitors the opinion feeds for all 13 federal circuits (1st through 11th, D.C. Circuit, and Federal Circuit) plus SCOTUS.
• Google Scholar Alerts — Reads Scholar case alert emails from a monitored Gmail account, pulling newly indexed opinions that match configured search terms.

State Court Opinions

• Justia RSS Feeds — Monitors opinion feeds for all 50 state supreme courts, plus intermediate appellate courts in states where First Amendment or platform liability cases are most active.

Legal Commentary

• Eric Goldman's Technology & Marketing Law Blog — One of the leading academic commentators on Section 230 and internet law.
• Techdirt — Covers platform liability, First Amendment, and tech policy with a legal focus.

2. Relevance Analysis

Every item retrieved is evaluated by Claude (Anthropic's AI) before it enters the database or newsletter. The analysis has two steps:

Step 1 — Is it relevant?

Claude reads the case name, court, and a text excerpt (up to ~5,000 characters of the opinion or filing) and decides whether it falls within one of three topic pillars:

• Section 230 — Cases interpreting or applying 47 U.S.C. § 230, including immunity defenses, exceptions (FOSTA, CSAM), and circuit splits.
• First Amendment — First Amendment claims involving technology platforms, government compulsion, content moderation, prior restraint, compelled speech, or related doctrines.
• AI Liability — Civil liability claims arising from AI or ML systems — defamation by AI output, negligence, product liability, copyright — and regulatory actions targeting AI.

Cases that don't clearly fit one of these pillars are marked not relevant and excluded. The threshold is intentionally strict — only cases with a genuine legal development (new ruling, significant filing, notable outcome) are included.

Step 2 — Summarize

For relevant cases, Claude generates a structured summary covering:

• Issue — The specific legal question at stake.
• Posture — Where the case is in litigation (e.g., motion to dismiss, summary judgment, appeal).
• Development — What happened in this particular filing or opinion.
• Platform/Actor — The platform or defendant involved, if applicable.
• Significance — Why the case matters for the broader legal landscape.

Blog posts receive a shorter summary focused on the key legal argument or takeaway.

3. Deduplication

The same case can appear across multiple sources — CourtListener, a circuit RSS feed, and a Google Scholar alert might all surface the same opinion. The system deduplicates by case ID and case name before sending anything to Claude, so each case is assessed at most once regardless of how many sources found it.

4. Newsletter & Website

After analysis, relevant items are:

• Composed into an HTML newsletter and delivered by email every Tuesday and Friday morning.
• Stored in a SQLite database and published here, where you can browse by topic, search by keyword, or use the AI-powered search to ask questions across the full case archive.

The newsletter archive is also available on this site under Archive.