Challenge

Second Ladder Intel publishes three times a week to ~50,000 subscribers. Each issue requires competitive scanning across 9+ newsletters, story research with sourced citations, a structured draft with 10 sections, a community post, an Airtable content record, a Notion archive page, and an exclusion list update to prevent topic repetition. Done manually, this is a 3-4 hour process per issue. At three issues a week, that's 10+ hours of production work that doesn't scale.

The goal: compress the entire workflow into a single automated session that produces publication-ready output, synced across every system, with zero manual data entry.

What I Built

A fully automated newsletter production system that runs as a scheduled task inside Cowork. One trigger kicks off the entire pipeline: competitive scan, story research, story options with recommendations, full newsletter draft, community post, Airtable record creation, Notion archive sync, and exclusion list update. The output is three files (newsletter draft, story options, community post) plus live records in Airtable and Notion, all cross-linked and ready for review.

How It Works

The system runs in sequence through seven stages.

First, it scans 9 competitor newsletters (Morning Brew, The Hustle, Exec Sum, The Daily Upside, Chartr, The Rundown AI, Ben's Bites, The Neuron, TLDR AI) to see what they covered in the last 2-3 days. It pulls actual headlines and angles, not just "this was covered." This feeds into a coverage landscape block for each candidate story.

Second, it researches 18+ story candidates across three signal categories (AI impact, market moves, career opportunities) and scores them against editorial rules: no stacking negative stories, no leading with layoff angles for the Big Story, minimum 3 bullets per Opportunity Flow category (Hiring, Funding, Contracts).

Third, it surfaces story options with star-marked recommendations in a structured options file. Six options per signal, six Worth Reading candidates, six Predict This candidates. I review and swap picks before the draft is finalized.

Fourth, it drafts the full newsletter with all 10 required sections: subject line options with preview text, The Rundown, Quick Signals (5 items, 2-4 sentences each with inline citations), Opportunity Flow (sourced bullets across Hiring/Funding/Contracts), The Big Story, Making Moves, Try This Out, Predict This, Worth Reading, and Footer. Every data point is researched and cited inline. No placeholders.

Fifth, it generates a GroupApp community post tied to the lead story, with an engagement question.

Sixth, it creates the Airtable ContentOS record with all required fields (title, status, content pillar, format, upload location, publish date, internal doc URL, cover image prompt, brand) and links it to the Notion archive page.

Seventh, it archives the full issue to Notion under the organized hierarchy (Newsletter parent > Month page > Issue sub-page) and updates the exclusion list with every topic covered, preventing repetition across future issues.

The system also handles real-time feedback loops. When I flagged that the Contracts section was light (only 1 bullet), the system researched and added 2 more sourced signals (Dell/Microsoft $9.7B Pentagon deal, CACI $212M Space Force contract), then updated the draft, the Notion page, and the exclusion list in a single pass.

The Stack

Cowork scheduled task (orchestration), Beehiiv MCP (publication stats and performance checks), Airtable MCP (ContentOS record creation), Notion MCP (archive sync and page creation), web search and web fetch (competitive scanning and story research), Claude in Chrome (fallback for JS-rendered sources). The entire stack runs through MCP connectors, no custom code, no API scripts. Repeatable for any newsletter operation.

Results

This was the June 15, 2026 issue: "AI just split the job market in two, and only one side pays more." The system produced a complete publication-ready newsletter with 17 sourced stories, 11 inline citations, synced records across 3 platforms (local files, Airtable, Notion), and handled a live editorial correction mid-run. Designed to cut per-issue production time from 3-4 hours to under 30 minutes of review time. Real performance data (open rates, CTR, time savings) will be added once the issue is sent and tracked.