Alfred Black: The Agentic Butler That Builds Itself

Hey guys, this is David from the

screenless dad and today I'm going to

show you the world's first agent that on

boards itself. You do not need to

provide any context about your life

because Alfred Black has been designed

to get it for you. So when you go on the

Alfred Black website, you're going to

sign in with Google because email is

still the single most representative

thing that can show anybody the surface

of your life. Although it's very noisy,

but thankfully Alfred is really good at

discerning signal from noise. So, we're

just going to s sign in with Google. And

when we do that, what happens is that a

dedicated agent computer is provisioned

and Alfred is born. This is going to be

your specific personal Alfred and nobody

else gets access to it. It's not a cloud

container thing. It's your dedicated

agent computer that is powering Alfred.

And next, what's going to happen is that

now your Alfred is going to start

looking through your last 100 days of

email inbox. All the emails you

received, all the emails you sent, all

the things that you read, all the things

that kept on unread, and all the things

that you kept rescheduling, what are the

things that you were actively avoiding,

all sorts of those things. What happens

here is that there are a couple of

machine learning libraries that create a

profile of all the different behavioral

patterns that it can build from the

metadata of your emails. And

concurrently it starts extracting facts

about you using large language models so

that we also understand who is part of

your life. What are the the different

companies that you interact with? What

kind of services do you use? Who are the

different people in your life? And

usually from about 100 days of email,

Alfred can extract about 4 to 500 unique

facts about you. And out of those four

or 500, we're going to extract 10 to 12

that are key identity facts. This is our

first attempt at grounding Alfred so it

gets the basics right. Who are you?

Where do you live? How does your family

look like? What's your wife's name?

What's your daughter's name? Your son's

name? When do you like to get your

morning briefing? What's your company's

name? These are basic personal

identity facts that we want to use so

that Alfred grounds them. In my

experience, Alfred gets these right

about 80% of the time. So, there is just

one or two that you may need to edit

yourself. If you want to add it, just

click on edit and confirm and that's it.

You can also add some extra sentence

context if you want, saying, let's say,

yes, I do decline meetings before 9:00

a.m., but that's because this and that.

And that will give Alfred the extra

context. And then once that's done,

Alfred is going to compose your first

brief. This is important because I want

you to have a feel for what it's going

to look like to work with Alfred. So,

you're going to get an email that is

specifically written for you and it's

going to give you an introduction from

Alfred, but also a quick run through of

your life right now as seen by Alfred.

So, you can see that all the different

stuff about today and what are the

things that have been waiting for the

for you. what are the different drafts

that may be there in the system and a

couple of quiet notes that Alfred has

identified and just a simple

introduction. You're going to get one of

these every single day from now on. And

it's it's a good way to get a feel for

what it likes to have Alfred in your

life. Now, when you go into your

household, this is the last part when

you want to confirm your the surface of

your life. So Alfred thinks in terms of

four things but the fourth is not shown

here and I'm going to explain to you

why. The first is matters. Now this is

not a task. A matter is not a project.

Matter is a matter. It's an ongoing

concern that matters to your life

because I have about 35 different

project. But I only work on one or two

of them actively. And yes, I guess you

could say it's an active project that

has active tasks. But the thing is that

sometimes I have like short-term

projects and and I have completely

unrelated stuff that occupy the real

estate in in my in my mind and I want

Alfred to have the same understanding of

what's in here than I do. So I want

Alfred to have a good understanding of

what are the matters that occupy real

estate in here. What are the matters

that demand my attention? So, for

example, you might have the Carter

sprint. That's important. Eliza may be

having a wedding in September, which

obviously is something that you're

preparing for. And then there is an

ongoing correspondence with the Hol

Council. And this is just an example

demo, but I think you get the idea. And

sometimes Alfred will find out some

matters that make no make no sense. You

just remove them. Or maybe Alfred

figures something out, but it's wrong.

Like Eliza's wedding may not be in

September. It may be next September.

It's actually not that important. So I

can just say it's next September. So not

relevant yet. And then right and then

all I need to do is just edit these

things. Click on edit. So now it's

saved. I can remove the Carter sprint

because that's not important yet. And

then also chores. These chores are the

recurring tasks that are important to

you and these chores are going to be the

baseline for Alfred to start building

its own chores. So these are the things

that Alfred realizes. Let's say you like

to take a walk between 2 and 3 weather

permitting. So, Alfred will create a

chore for itself to check the weather

every every day at 150 and then send you

a notification to nudge you to go to a

for a walk or not nudge you because say,

"Yeah, weather is not nice. I don't

think you should go for a walk." These

chores are automatically created by

Alfred. And again, if you want, you can

add notes. You can remove stuff from

here. You just confirm the baseline. And

then there are a couple standing rules

that Alfred populates. These are the

things that it looks like kind of

ongoing standing rules that will make

sure that Alfred's work with you or or

is smooth. The fourth thing, you can

always also edit these as well or remove

them as you wish. The fourth thing that

does not exist in here yet, but it will

be built over time are patterns. And I'm

going to talk about patterns in just a

minute. But once you have this, you can

go to the today section where Alfred

sets the table, sets your desk, and this

is how Alfred sees your day every single

day. Now, it's important that this is an

anti- dashboard. In an ideal world, you

should not open this ever. But if you

want, you can still use this as a

control center. But you don't have to

because Alfred can always come to you

where you are. Alfred creates your daily

briefing every morning and Alfred can

deliver that daily briefing where you

want it to. You will get the web app

that I'm demoing right now. You also get

Alfred's own email address. So you can

always email Alfred and Alfred also can

email you. So if you want, you can get

your daily briefing via email and you

can just respond to that. Same can

happen through a text and same can

happen through a voice. So, my preferred

way of working with Alfred is to get a

voice call from Alfred. When he calls me

on the phone, I pick it up and then

Alfred walks me through the daily brief

and asks me if there are any decisions

that I want to make. So, basically, I

don't need to open this because Alfred

can just walk me through these items and

ask, "What do you want to do, David?

Smithson is owed 142 days late. It is

within your standing limit of 500. So I

would actually uh settle this. What do

you want me to do? And then I can say I

want offer you to do it. So I'm just

going to delegate it to you and give you

the instruction. Or I can say let's come

back to this. Let's put a pin on that.

Let's come back to this in a week. So I

defer this or I'm I can say I don't want

to do this anymore. This is not relevant

anymore. Get rid of this. I can click on

delete. Or I can say I'm going to do

this myself. So I'm going to click on

do. And I can do the same thing over the

phone. Right? And when I do that, Alfred

comes back with the next next thing. Tom

Ashworth at the school asks for a word

about your boy. Okay, let's see what we

want to do. Obviously, I will just want

Alfred to send a message and say

something about when to talk or

something. Now, as you go through this,

what you're going to see is that Alfred

will do a bunch of actions on your

behalf. Some of them are going to be

specifically delegated to him and some

of them Alfred is doing on its own. And

you can see that all the decisions that

are being made that you are making and

Alfred is making is always about the

different stuff that surfaces in your

life. What happens to them? Is this

something that Alfred can handle or is

it something that Alfred needs to ask

you to do or ask instructions on? or is

this something that needs to be held?

Because Alfred looks at it and says, "I

have no idea what to do with this, so

I'm just going to put a pin on it." Now,

what we want is the more you use Alfred,

the more relevant this white bar becomes

because we want Alfred to start handling

more and more stuff the exact way we

want it to happen. And in order to do

that, in order to do that, well, Alfred

will always analyze the confidence level

of how confident is Alfred in saying

that whatever suggested action Alfred

thinks would be the right one, how

confident is Alfred that that's indeed

the right one. So that's what we call

progressive autonomy. And what happens

is that every time you make a decision,

that decision gets saved as an

observation. And when that observation

starts piling up in the system, Alfred

starts creating these patterns. And

these patterns are living organic tools

that Alfred extracts from your and his

behavior over time. That's why you don't

get patterns just at provision. patterns

are things like an email where that says

there was a bolt

receipt that got filed as as an expense

item in your financial records on the

27th of April and that automatically

gets pulled into Alfred's system and

extracted as an observation. So Alfred

can file it under this specific pattern

and then once observations start piling

up these patterns are automatically

created and start populating and

patterns can have three different levels

of maturity. Once a pattern is born

Alfred notices that pattern but it

starts asking things about it because

it's like I notice this is recurring but

I'm not sure about it. And then as

observations keep coming in, Alfred will

update it to a confirming state where

Alfred says, "Look, this is something

that you're doing pretty regularly and

consistently. So I think this is what

the next thing should be, which means

that the decision is coming in and

Alfred already can match it with a

pattern. So you don't have to think

about what the decision is is suggesting

that decision, but you still need to

give the go-ahad." And then once enough

observations built up, which means that

Alfred has suggested the right thing to

do enough times, it can move over to

acting, which means that over time,

Alfred learns the rule that every

receipt from bold is treated as an

expense item. So it won't even ask you

about it. It will just do it in the

background. And if there is a pattern

that I know for a fact is getting

deprecated, I can just ask Alfred to

forget the pattern. So this is a living

thing that gets updated all the time on

its own. Now let's take a look at these

matters again. So these ongoing matters

are really the centerpiece, the

gravitational epicenter of your life. So

let's say we had the Carter sprint and

we can see all the different

conversation all the different decisions

that have been made all the different uh

uh tasks and all the different drafts

that belong to this specific sprint and

Alfred keeps the the state or tracks the

state of this so it's always completely

up to date and if you want you can go in

and see what is what because you can go

into Alfred's long-term structured

memory which is the vault which is

compatible with obsidian which means

that you can just sync with obsidian and

then you can open the vault items and

edit them as you wish but you can also

edit them in here. So in this vault you

can have all the different things in

your life including all the stuff that

Alfred has identified from onboarding to

your life through email. It will create

these vault items and you can start

looking at them and these are interlin

and having the proper front matter and

and wiki links so it has the proper

structure. This is basically a markdown

filebased knowledge base of Alfred's

understanding of your life. The

important part of this is this all lives

in your agent computer. You have full

access to it. You can download it. You

can take it to wherever you want to. Now

let me talk to you about two other

channels which is Slack and Telegram.

These are running through the runtime

that powers offered in the background

which means that you can also make use

of any other of the channels that open

crew allows. If you want me to I can I

can set this up for you. This is

something that I need to do personally

for for my users for now. And then as we

as I discussed so every morning 7:00

a.m. there is a morning email etc. uh

you get a simp simple readable version

of what these chores are and you can

edit them, you can delete them but in

the background these are running through

a durable temporal schedule and if you

want you can also check and edit the

code itself but that's not something

that you will probably do very often. So

that's mainly it but the one thing there

are two other things that I want to show

you. One is apps. So Alfred can connect

to all sorts of different external

applications that will start sending

signals to your life to your agent or

also give offer the ability to use those

those apps for example you use Gmail but

let's say you also use notion and GitHub

and linear and all sorts of other things

now comes combo off Alfred comes with an

integration that gives you over a

thousand different apps to be integrated

on demand. So you don't actually have to

think twice about how do you create an

MCP server connection to your openlow

instance or whatever. You literally just

click on connect and Alfred learns how

to use those apps on its own. You don't

have to deal with any of those things.

Now the other thing with about apps is

there are four core services which I'm

going to walk you through in a separate

video. But Alfred comes with four core

services. is now obviously the open claw

itself you can have full access to the

open claw web user interface if you

prefer that uh Alfred comes with a

secret management which is vault v world

warden an open-source uh one password

alternative that allows you to store all

your credentials securely so Alfred

never actually gets access to your

credential but uses everything through

vault warden there is a financial

tracker which is called shore

also running on Alfred's computer and

it's completely integrated. So if you

want to connect your bank accounts so to

give Alfred a good idea of how your

financials look like which I sure like I

definitely recommend you do then you can

connect your bank accounts to shore and

then has a bootstrap skill when it does

the exact same thing it did with your

emails. It runs a bunch of machine

learning libraries to build the surface

level model of your financial life, your

habits and everything and configures the

app for you. So it creates all the

tagging rules and all the categorization

and everything and starts managing and

maintaining this financial tracker app

for you. The same thing happens with

plane. Plane is a project management

tool. I will do a separate video about

it and offer it comes with plane

pre-installed. So all the decisions that

you see in here all the decisions are

basically tasks inside plane. All your

matters are exist as in plane which

means that if somebody prefers using

project management tool to all of this

you can use plane but the clear

distinction here is that you do not need

to do any data entry because plane

maintains itself through Alfred. All you

need to do is open an issue and comment

on it delegating something to Alfred

because there is an at Alfred user in

plane. Again, in a separate video I'm

going to walk you through. And if you

don't like to use OpenClue integration

at all, then you still have the option

of making good use of cloud. Now you

cannot use subscriptions with open clue

or any other third party agent but you

can use Alfred with code and that is

because Al alfred exposes a set of MCP

servers in fact there are five of them

which are not shown in this in this

demo. There are five MCP servers, five

specific skills, and I'm going to set

record a separate video on how you can

configure code to act as Alfred and have

full access to Alfred and that will

solve all your challenges about how to

have your code environment turned into

Alfred. I personally work with Alfred

through direct like phone and text and

email integration as well. I also use

Slack but I also use code cowwork. I

turned code into Alfred and most of my

knowledge work is done through that.

That's not coding for which I code.

There is an approval secret that is used

for authentication. That's what you're

going to need to provide when you're

connecting the MCP servers to cloud.

Again, I'm going to walk you through

that in a separate video. and then a

bunch of skills for you to use as well

if you want to use that in uh or in

code. And then you can also have an API

key. So if you don't like the existing

MCP servers, if you want to create your

own automations or Alfred in like third

party apps, you can get a full document

fully documented API reference and an

API key that will directly connect to

your Alfred agent computer. Now that's

mainly it. This is how Alfred Black

works. This is the zero configuration

external nervous system because there

are just too many matters in my life

right stuff to stay on top of. So Alfred

is really the externalized nervous

system for me. So I don't need to keep

remembering that bin day is today.

Shouldn't forget to take out the bins.

And then I also need to send a a message

to my lawyer about land survey and

survey and then I also need to talk to

my neighbor and then I also need to do

this and as a new father I am very

overwhelmed and very sleepdeprived. So

the the list of things on my desk keeps

piling up and the energy to be focused

keeps drowning and Alfred is born to

offset that. So, if you feel overwhelmed

about the shape of your life, you're not

alone. And if you feel about feel

overwhelmed about all the different

tools you need to use to keep your life

in check, you're not alone. If you want

to get rid of being chained to your desk

and to your computer all the time just

to keep your life working, you're not

alone. And this is my solution. Thanks.