01 / 07
How we build with Claude

The model is
the easy part.

Everyone has the same Claude. What makes it useful is what you wrap around it. That is what we are building today.

Scroll, or use the arrow keys

First, do not get lost

Three tabs.
We only live in one.

What an agent actually is

Three pieces, or it is
just a chat window.

ModelThe brain
+
ContextThe brief
+
HarnessThe keys
=
It works for youNot just answers you
In plain English

A brain, a brief,
and the keys.

01 / Model The brain

Clever, quick, and knows nothing about you. The same one everybody gets.

A brilliant new starter on day one. Has not met a customer yet.

02 / Context The brief

Your prices, your customers, your voice. Everything it needs to sound like you.

Hand over the folder. This is the bit that is always missing.

03 / Harness The keys

Your systems, and permission to use them. Reading, writing, sending, on a schedule.

A desk and a login. This is what a chat window never has.

The map

Your setup is four layers.

Why most people get nothing out of AI

Build them in this order.

MMemory

It cannot write about your business until it knows your business.

AApplications

It cannot pull your numbers until it can reach your systems.

SSkills

Now it can do a real job, because it knows you and can reach your data.

RRoutines

Once a job comes out right, put it on a timer and stop thinking about it.

Most people start at Skills with an empty Memory, then wonder why it sounds like a robot.

underseage.studio

Four layers.
That is the whole map.

The layers never change. Only what you put in them does. So you are never starting from scratch, you are just adding to one of the four.