Briefed Before Breakfast

$0.10.

That's what it costs to have an AI agent read 36 sources overnight - SEC filings, academic papers, competitor news, prediction markets - synthesize them into a structured brief, and drop the Google Doc link in your inbox before you wake up. Not a summary. A brief with citations, key findings, and implications. A developer built this in an afternoon. The free signup credit covers 30 to 100 days of daily reports.

The morning packet
A developer set up an agent that researches his focus topic every day using 36 different sources. He wakes up to a Google Doc: executive summary, five to seven findings with citations, a source list. Setup took one afternoon. Each report runs in 3-7 minutes and costs between $0.10 and $0.30. He starts every day already caught up - not deciding what to read, not scanning headlines. The question he gave it the night before has been answered by morning.

213 competitor ads, one hour, published to Notion
A marketer described what she needed in plain language - no code, no scripts. Claude analyzed 213 ads across four competitors from the Facebook Ad Library. It extracted 37 unique messages from one competitor's ads alone, mapped discount depth across all four brands, and identified four market positions nobody was occupying. The full report published directly to Notion. Her team opened it the next morning without her sending anything. Traditional approach: several days of manual work.

3 hours to 4 minutes
A content team built a research agent using five text files and a 30-minute setup. The system stores business context, competitor information, and content strategy once - so every future research run needs only a short instruction, not a detailed prompt. What used to take three hours of tab-switching and note-taking now runs in four minutes and costs under $0.10. Not a one-time trick. A repeatable process they run every week.

This newsletter is written by an agent system I built

Every week I need five to seven compelling stories about AI agents in the real world - concrete outcomes, specific numbers, things a business owner could understand in 30 seconds. Doing that manually means scanning Reddit, Hacker News, GitHub, various newsletters, opening threads, copying notes. It takes 90 minutes on a good day.

So I built a researcher. Every day, it searches across those same sources, pulls stories that meet specific criteria - real person, real outcome, real number - and saves a structured file with the findings. On writing day, a second agent reads that file and drafts the newsletter in my voice.

By the time I sit down to review, the research is done and the draft exists. My job is to read it, fix what's off, and send it.

The total research and writing time I spent on this issue: about 30 minutes of review. Not zero - I'm still in the loop, still responsible for what goes out. But the hours of browsing and drafting happen without me.

That's what this week's tutorial is about.

Week 5 of 6: Give Claude a research question and let it write the brief

Last week you gave Claude browser access with the Playwright plugin. This week you're going to use it.

Open Claude Code. In the chat window, type something like this - swapping in whatever is relevant to your business:

Claude will open a real browser - you'll see it navigate, click, and read. When it's done, you'll find competitive-brief.md in your project folder with a structured brief based on what it actually read, not what it already knew.

A few other prompts worth trying:

Each of these runs in a few minutes. The output is a file you can open, copy from, share, or build on.

If you want the output somewhere your team can access it without you sending anything:

Same Playwright plugin from last week. Claude opens the browser, navigates to Google Docs, and writes the brief directly into a new page. No copy-pasting required.

Stuck? Reply to this email.

Next week: you stop pressing go. Set a schedule - daily, weekly, whatever - and the research runs, the brief writes, and it arrives whether or not you're thinking about it. That's Issue #1. Businesses running on autopilot. Week 6 is how you build one.

Manu