☎️ I used to be the assistant. Now I have two.

Hello there 😎

In October 2011, I was a floating assistant at a literary management company in Hollywood. Also known as a "floater."

A floater is a temporary assistant. You cover someone's desk when the regular assistant is on holiday, sick, or at a dentist appointment (read: interviewing for a new job). If you manage not to screw anything up too badly, it's a good sign you're ready for your own desk.

The worst was the phones. The worst worst was merging phone calls. You'd be in the middle of rolling calls for your executive (i.e., calling everyone back you owed a call), and someone would call you back, and so you needed to "merge" the two calls together into one line. High-tech stuff, for the time.

My worst nightmare was dropping a call during a merge. So I practiced. I'd stay late, making imaginary calls like a toddler with his Paw Patrol taken away from him, so now he has to use his imagination. I'd practice merging calls between two different lines to make sure I didn't mess it up when two executives wanted to get on the same line and bullshit about some deal point or rescheduling their massages.

The next summer I had my own desk. Three of them, actually: three executives, three calendars, and nine phone lines, plus managing this FileMaker Pro database where I was logging every call that went in/out.

Run to Hamburger Hamlet and pick up lunch for the office? Sure! Read two books over the weekend for Monday morning? Yes, please. Mail the checks. Send the holiday cards.

It was 80% glorified executive assistant work (calendar, email, answering the phone). The other 20% was cool Hollywood stuff and why I moved to LA (listening to calls, reading scripts, redlining literary option and purchase agreements).

(And for the record: looking back, it was tedious and boring and such an inefficient use of time... but I loved this era of my career.)

🙋 I'd have to hire the worst version of me

The alternative to using my agents and prompts isn't nothing. It's hiring someone to do what I did in 2011.

These agents do this stuff way better than I could. I'm thrilled they're doing it for me instead of hiring that 2011 version of myself. That 2011 me was a great person, had a good eye for narrative, and was getting better at deal-making, but he was a shitty executive assistant. He was always daydreaming about writing or thinking about the book he just read, or literally thinking about anything else other than answering the phone for three different bosses all day.

⏳ This used to be my full-time job

I did this all day and spent most of my brainpower on it. And now, anyone with a $20/month subscription to ChatGPT or Claude (you can start on the free tier, but you'll hit usage limits quickly) has the equivalent of that 2011 version of me (but faster! smarter! better!) in their back pocket.

Here's an example: for any of my executives, the 2011 version of me might book a call with a producer or writer or another executive. Then I'd do research on them and pull up notes from previous calls. Then I'd brief the executive. Then I'd make the call for them, take notes while I listened in, summarize them after the call, and book a follow-up call a month later.

Last month, I had an agent build me a custom networking tool to help me stay connected with people. If I want to book a call, they schedule on my Calendly. The night before we meet, the agent writes a few bullet points about what we talked about last time and any updates on their life I want to ask about.

☀️ My two main skills and what they do

I have two main skills (think: files that tell the agent what to do). Most likely you can start with something simpler (I'll provide a starter version at the end), but just to give you a sense:

Skill one is "Start Day". What it does:

• Yesterday's note. What was due, what needs a reply, what got carried forward.

• Email. A week of inbox sorted into four prioritized buckets.

• Health. The last three nights of sleep, how my weight has changed, yesterday's calories and macros, and my workout for the day.

• Writing. What's on deck this week.

• Runbooks. Any open projects I've got going.

• The ask. "What are your top 3 priorities for today?"

Skill two is "End Day". What it does:

• The journal. Reads the day's note, pulls out what went well and what I can improve.

• What got built. Sweeps the day's sessions for what shipped and what was learned.

• Birthdays. If a kid's birthday is within 14 days, reminds me to think about getting a lai see or gift ready.

• Content. Proposes one to three ideas based on my journal and notes.

• Hand-off. Sends the open loops to my OpenClaw agent (think: an always-on AI assistant that works overnight).

🧭 Forwards, backward, and around corners

At first, these skills focused on briefing me on my day and preparing me for what's next.

As I built them out, it became about helping me look around the corners of life and anticipate what's coming up. It was like having a full-time assistant who's always looking forward, backward, and around corners, so you're ever vigilant of the caprices of life.

When you're a parent, when you're doing a portfolio career, things are unpredictable. You're trying to bring some stability to it.

It's so easy to lose sight of someone's birthday or a homework assignment that came over Remind, or a soccer game added via SportsEngine, or a new deadline sent over Slack. It's like having someone watch your back, keeping everything top of mind when you don't have the bandwidth.

More examples:

• It sees that I started a refund, but the last day to return the thing is today.

• It flags that I have only $1,000 left in the account I pay bills from, and I need to move money before an auto-payment hits.

• It looks ahead two weeks, sees my brother's birthday, and tells me to think about a gift.

• It connects the dots between client work and uses one project I'm working on to inspire how I approach another client.

It moves backward and forward in time, connecting everything I've done with everything I still want to do. It's like playing three-dimensional chess for me, every day.

🙌 Then they started offering to do the work

In the next phase, I had it surface things it could just do for me and get done.

• Any website or tech maintenance. If I flag a problem in Gmail, it takes a look and fixes it.

• Groceries. If it doesn't see a receipt from Instacart or my meal delivery service that week, it reminds me to place the order and asks if I want it to do it.

• The email I've been putting off because I'm not sure how to respond. It offers to write the first draft.

🧠 The secret sauce: it's the context, not the agent

There's a lot of hype around the new agents (I've tried Instinct, and I run an OpenClaw agent too). They're good! For now, they haven't bested me at running these skills, which I can tweak to exactly what I want.

My setup is loaded with context about me: more or less everything I've written over the last 20-odd years, plus everything I've read and taken notes on. I've been writing and journaling for so long, and it has all of it, so they can be intelligent about how I approach life, what I want to mind, and what I want to write about.

Ultimately, it comes down to the wealth of context you arm your system with.

🚀 Start with one prompt, then layer

If I were starting today, here's the prompt I'd start with.

Copy this into any AI tool that can read a file (Claude, ChatGPT, or Claude Code if you keep notes in a folder):

When it works, you have a note with three priorities and a short list of what's owed before your first cup of coffee.

Then you can start adding layers. In no particular order:

• Email triage

• An agent's overnight report

• Family birthday checks

• Health monitoring

• Project management across everything you're doing

• "Three things I can do for you today"

Start running this every day. As it gets better, layer in the next thing.

Have fun, and let her rip.

🤝 When you’re ready, here’s how I help you land your next job:

I love meeting new people. Say hi on LinkedIn 👋