I have a confession. For the past two years I have been subscribed to more newsletters than I will ever read, following more RSS feeds than I check, and saving more articles than I open. Every morning I sit down with a vague sense that something important happened and I missed it. The problem is not a lack of information. It is too much of it. AI, AI governance, product leadership, agents, MCP, new tools, the space moves fast and the signal-to-noise ratio is brutal. Miss a day and you feel behind. Read everything and you feel buried. There is no good default.

The irony is not lost on me. I spend my days thinking about how AI changes the way humans make decisions. And I was drowning in information exactly like everyone else.

So I built something.

What it does

The Personal Executive Briefing Agent runs every night at 3am. It pulls from my Gmail, my Zoho business email, a curated list of RSS feeds and Substacks, a set of Reddit communities, and a handful of GitHub organizations I track. It scores every item against my actual priorities, not generic interests, but the specific things I am working on right now. Then it sends me one email.

By the time I sit down in the morning, the briefing is waiting. Twelve to fifteen items. Each one with a one-sentence explanation of why it matters to me specifically. A two-sentence TL;DR. And a recommended action: read now, save for later, reply, ignore.

Not “this is an interesting AI development.” More like: “this agent failure pattern is the exact scenario your JDD framework addresses and you should use it as a basis for a blog post.”

That specificity is the whole point.

What changed

I have been running it for a few weeks now. The mornings feel different. I open one email instead of twelve tabs. I read what matters instead of skimming what arrived. The cognitive overhead of deciding what deserves attention, which turns out to be surprisingly expensive, largely disappears.

What I learned building it

A few things I learned building it:

The pre-filtering step matters more than the LLM. Getting from 160 raw items down to 33 before any API call is what makes the economics work and what keeps the briefing focused. Garbage in, garbage out – and garbage is expensive.

Diversity caps are essential. Without them, whatever source had a busy news day floods the briefing. Reddit during a major release will crowd out everything else if you let it.

The “why it matters to you specifically” prompt is the hardest part to get right. The difference between a useful briefing and a generic one lives entirely in that one sentence. It took several iterations of the prompt to get it consistently specific rather than generically positive.

Configuration as Markdown is the right call. Being able to edit my topics, weights, and assumptions in plain text, without touching code, means I actually tune the system as my priorities change. A config buried in code is a config you stop trusting.

How it runs

The whole thing runs on GitHub Actions and costs roughly $7-9 per month in Anthropic API costs. Resend handles email delivery on their free tier. The repo is open source, designed to be forked, and built so that all personal configuration lives in Markdown files that never touch the engine.

If you want to run your own version, the setup guide in the README will get you there. The Gmail OAuth step is the most involved, budget twenty minutes for that part. Everything else is straightforward.

Try it

The project page is here: executive-briefing-agent

And the source is here: github:executive-briefing-agent

One last thing

One thing I would add before I close. This project is a small demonstration of something I keep writing about. The agent does not decide what matters. It surfaces candidates and explains its reasoning. I decide. The judgment stays with the human. The mechanical work of collection, filtering, and summarization moves to the machine.

That division of labor is not accidental. It is the design.