The work happens at 3am. I am asleep.
The Unattended Playbook is how one engineer runs cold outreach, builds and deploys a working demo per prospect, and drives a fleet of browser tabs with coding agents, on one laptop, for almost nothing, without being in the room.
The essays give the shape of every method and the mistakes that produced it. The exact scripts, the rules file and the setup are not in them. Those are what I install on your machine, in a day, so you do not spend the year I spent finding them.
conversations booked per hundred cold pitches, after every pitch shipped with a deployed, in-stack demo
agent sessions sharing one laptop, one browser and two small servers at the same time
monthly hosting bill for forty-odd deployed projects, with every model call inside a free tier
The six essays
In order, though each stands alone. They are long, specific, and built from the incidents that made each rule necessary. If you read one, read the first.
- The Browser That Works While I Do
My coding agent drives a real, logged-in Chrome all day, in the background, and I never see it happen.
- Cold Email Sent at 2am That Still Reads Like a Person
The agent finds the people, finds their real address, and drafts the note. The rules that keep it from sounding like a machine took longer than the code.
- I Ship a Working App on Your Stack Before We Ever Talk
The pitch is a URL. The agent builds and deploys it overnight. Response rates went up roughly tenfold, and the reason is not what I expected.
- My Coding Agent Has a Constitution. Every Rule Is a Scar.
The file the agent reads before every session is longer than most codebases' READMEs. Each line in it is dated, and each date is a mistake.
- The Agent Must Refuse, Not Guess
A form-filling agent answered five questions it had no answer to, confidently, on one form. What it took to make it stop, and the two rules that fell out of it.
- Ten Agents, One Laptop, and a Hosting Bill Under Twenty Dollars
What it actually takes to run a small fleet of coding agents on a single machine, sharing one browser and two tiny servers, without them killing each other or the machine.
Have it installed on your machine
Two ways to get the working version rather than the description. Both include the unredacted playbook: the actual rules file, the browser wrapper, the lock and ledger tooling, and the report and reflection templates, shaped to your machine and your workflow.
Operator Call
Bring the one workflow you want to run without you. On a screen share, we set it up on your machine, live, and you keep everything.
- Your rules file, tiered, with the hard rules that matter for your situation
- The attended-browser lane wired to your real Chrome, background tabs only
- The end-of-session report and reflection protocol, installed and running
- The unredacted playbook, to keep
Operator Install
The whole system, on your machine, by the end of the day. Outreach, demos, browser fleet, memory and reporting, with your accounts and your hosting.
- Everything in the call
- The outreach lane: signal-driven targeting, real-address resolution, a warmed sending subdomain, the do-not-contact gate
- The demo pipeline: discover, dedupe, research, build, deploy, with the capacity ledger
- Lane locks and the health check, so a second and third session can run safely
- Thirty days of follow-up by email while it beds in
Before you pay for anything
Why not just publish the prompts and scripts?
Because they are the least transferable part. Every script on my machine is shaped by my browser, my accounts, my hosting and my rules file. The essays give you the shape of each method, which is most of the value. The install gives you the working version on your machine, shaped to yours.
Does this work with any coding agent?
The methods are agent-agnostic: a rules file the agent reads, a browser it drives over the DevTools protocol, locks, ledgers and reports. I run it on Claude Code and that is what I install. The structure ports to any agent that can read a file and run a shell.
What do I need before a call or an install?
A Mac or Linux machine you control, a coding agent you already pay for, and one workflow you want to run without you. Outreach, applications, research, deploys, anything repetitive with a browser in it. We start from that workflow, not from a template.
What happens after I pay?
You get a receipt from Stripe and an email from me within one business day with times to pick from. Calls are on a screen share. Installs are a scheduled working day on your machine over a screen share, followed by thirty days of email support.
Is this for teams or for individuals?
It was built by one person for one person, and that is where it fits best: founders, freelancers, one-person consultancies, anyone whose week has a repetitive browser-shaped job in it. Small teams work too. Large ones should start a conversation instead.
Will you run outreach for me?
No. I set the system up on your machine, with your accounts and your voice in the rules file, and you run it. The essay on cold email explains why the voice cannot be outsourced.
Written by a senior software engineer in the Seattle area who has spent the last year running a one-person practice this way. The rest of the site is sixty-odd essays on Rails, Postgres, Go and production systems, and this page is how to hire that engineer for the ordinary kind of work.