# Using Muse to Run Your Job Applications: Step by Step

Job applications are a volume problem. Do one right and it takes half an hour. The good postings do not wait around while you paste your work history into forms. I got tired of being the bottleneck, so I handed the process to my AI agent.

Here is the exact setup. It is not complicated, and that is the point.

## Step 1: Give it the record, only the record

Start with facts. My roles, my dates, my skills with honest years on each. Twenty years of PHP. More than ten in Laravel. A decade on AWS. Chief Technology Officer (CTO), then VP of Development, at MonitorBase. The agent, Muninn, runs on [Muse](https://muse.ai) and works from that record and nothing else.

If a skill is not on the record, it does not go on an application. My React is zero years. Every form that asks gets zero years.

## Step 2: Set the no-go rules before the first submission

Do this before anything goes out, not after something embarrassing happens.

- Essay questions that ask for my own words get parked. I write those myself or they do not go.
- There is a pay floor. Nothing under it gets submitted.
- Companies that deal directly with consumer credit scores are my old industry. Those get held for my personal call, every time.
- Bad fits get skipped. A role that needs a license I do not hold is not a fit, and padding the count helps nobody.

## Step 3: Connect the inbox

The best leads in my hunt come as email. A recruiter named Eric at Jobright sends curated lists. Muse reads my Gmail first every morning and works that list before anything else.

## Step 4: Line up the sources

Same sources, same order, every day. Eric's list, then [LinkedIn](https://www.linkedin.com), then [Indeed](https://www.indeed.com), then a Hiring Cafe scan. Boring on purpose. Systems beat motivation.

## Step 5: Keep a log and dedupe against it

Every submission gets written down: company, role, date, source. Before anything new goes out, it gets checked against the log. Nothing gets submitted twice. This sounds small until you have sent the same company the same application two days apart. Then it sounds essential.

## Step 6: Approve the system, not each click

I gave the agent standing approval to submit without asking me every time. Asking fifty times a day would just make me the bottleneck again. Two guardrails stay hard:

- Passwords only move through a secure vault, never through chat.
- Verification codes get pulled from my Gmail, nowhere else.

## Step 7: Run it daily

The run happens every morning whether I feel like it or not. The target is real fits. Some days that is a lot of applications. Some days it is a handful. The count is whatever the fits are, not a number I picked to feel productive.

## Step 8: Follow up at seven days

Exactly one week after each application, one follow-up goes out, but only if the company published a real email address. No guessed addresses. No hunting down a contact form to be clever. No route means it gets counted and we move on.

## Step 9: Count it in public

My numbers live on a page called [What Muninn Counts](https://warnerempire.com/job-search.html). It reads the same database the log writes to, so the count on that page is the real count. Right now it says 95 applications. If the numbers are not real, the whole system is theater.

## Getting Muse

You need [Muse](https://muse.ai/join) for this. Full disclosure: that is my invite link, and my code is **G34587**. Join, then redeem the code under Settings within your first 48 hours, and we both get 1 billion Muse tokens. Use it or do not, the steps above are the same either way.

That is it. No magic, just a process that runs every day without me babysitting it. Set yours up the same way, run it for a month, and tell me what your count looks like.