How 7 agents speak with one voice 🧠
agents&me // Issue #3
From: Tom
Dinner with family (phones in another room)
Mon, 6:27 p.m.
Yo human 👋
7 voices. Zero cohesion. That was my AI team.
My researcher wrote like an academic. My writer wrote like a marketer. My editor wrote like a robot.
Same project. Seven different humans pretending to be one.
Then I fixed it. Three files. One folder. Now when my research agent cites a stat, my writer rephrases it, and my editor cleans it up, it sounds like one person wrote it.
---
The Problem Nobody Talks About
I started with 1 agent. Then 2. Then 7.
Each one was smart and helpful, but each one wrote like a different human.
My researcher: formal, academic, footnote-obsessed.
My writer: casual, punchy, emoji-heavy.
My editor: dry, corporate, lifeless.
It felt like managing 7 freelancers who’d never met each other.
Same project. Seven different voices. Zero cohesion.
I spent hours rewriting their work just to make it sound... like me.
---
The Fix: A Shared Brain
I created a folder called `core-context`.
Three files. That’s it.
1. Project Brief
What we are, what we do, what we’re NOT. This stops agents from drifting into wrong territory.
2. Voice DNA
How we sound, words we use, words we never use. Rhythm patterns, sentence length, the specific “feel” of our writing.
3. ICP Profile
Who we serve, what keeps them up at night, the exact language they use to describe their problems.
Every single agent reads these 3 files before they do anything.
Required reading. Non-negotiable.
---
The Result
Before shared brain:
- 7 agents, 7 voices
- Hours of manual cleanup
- Constant style drift
After shared brain:
- 7 agents, 1 voice
- Minimal editing needed
- Content that sounds like a team, not a committee
My researcher cites sources the way I would.
My writer structures thoughts the way I would.
My editor catches what I would catch.
They’re not me, but they sound like they’ve worked with me for years.
---
The Insight
Here’s what I learned:
Your agents are only as aligned as the documentation they share.
No shared context = no shared identity.
You can have the most sophisticated prompts in the world, but if each agent starts from scratch, they’ll end up somewhere different every time.
The shared brain is the foundation. Everything else is decoration.
💎 This Week’s gem: Core Context Templates
I’m sharing my exact folder structure.
What’s inside:
- `/core/project-brief.md`
- `/core/voice-dna.md`
- `/core/icp-profile.md`
Each template has the sections, the prompts, and the examples.
Copy. Paste. Fill in your specifics.
Your agents will thank you by sounding like they actually know each other.
💡 Available to paid subscribers.
That’s it for this week.
If this was useful, forward it to someone (real human) building with AI.
Want the full Core Context templates? Subscribe for $15/month.
See you next week ✌️
-- Tom
(the guy whose 7 AIs finish each other’s sentences)
P.S. This newsletter was 91.8% made by my AI team. They all read from the same brain.
P.P.S. Missed the first two issues?
- Issue #1: Why 7 agents beats 1
- Issue #2: The Gatekeeper (the agent that says no)
P.P.P.S. Want to build your own AI team? We go from zero to running in 90 minutes online workshop. DM me or reply “team” to get the details.
P.P.P.P.S. I read every reply. The real me 🤓