►
From YouTube: W10 TEC Lab!: Tokenspice and cadCAD for Energy Web
Description
🙏 Thank you for watching! Hit 👍 and subscribe 🚩 to support this work
🌱Join the Community🌱
on Discord https://discord.gg/DDr5kYU
or say hello on Telegram http://t.me/CommonsStack
Join the conversation https://forum.tecommons.org/
Follow us on Twitter: http://twitter.com/CommonsStack
Learn more http://tecommons.org/
A
Yeah,
this
is
great
good
turnout,
so
hey
welcome
everyone,
who's
bright
and
early,
usually
people
trinkle
in
for
a
few
minutes,
but
so
this
is
a
special
occasion.
It's
our
first
guest
hosted
lab,
which
is
something
I
would
like.
I
would
like
the
majority
of
them
to
be
guest
hosted.
So
this
is
really
exciting
for
me
and
I
would
like
to
introduce
everyone
to
mark
who
has
been
active
here
in
the
community
for
quite
some
time.
A
We
were
classmates
in
the
october
value
flows,
ecosystems
course
that
was
led
by
angela
and
sebnam
and
michael,
and
so
we
got
to
learn
about
designing
these
stock
and
flow
diagrams
to
be
used
in
token
engineering
and
then
since
then
mark
and
I
both
of
us
went
through
the
ocean
protocol
study
group
that
was
hosted
here
in
the
token
engineering
community,
and
that
was
a
for
for
me
and
I
think,
mark
and
everyone.
That
was
a
really
good
experience.
A
I
learned
so
much
from
that
study
group
and
angela
pushed
me
to
give
a
presentation.
A
So
I
gave
a
presentation
on
sort
of
ai
in
token
engineering
and
specifically,
reinforcement
learning
and
token
engineering,
with
a
focus
on
the
the
token
spice,
simulator,
that's
put
out
by
trent
mcconaughey
at
ocean
protocol
and
so
mark
got
inspired
by
my
presentation
and
he
dragged
me
into
a
hackathon
and
he
got
inspired
to
see
if
we
could
use
these
reinforcement.
A
Learning
agents
for
energy
optimization
because
mark
has
a
background
in
that
and
he's
been
focusing
on
the
energy
web
protocol
and
there
was
a
hackathon
being
hosted
by
ocean
protocol
and
various
other
organizations
at
the
time,
and
so
we
we
were
investigating
token
spice
and
and
expanding
it
and
extending
it
to
add
these
data,
these
energy
web
agents,
who
can
publish
energy
data
and
stake
on
that
data
and
and
then
even
have
agents
that
make
some
predictions.
And
I
don't
want
to
give
away
too
much.
B
A
Giving
a
little
bit
of
background
and
context-
and
I
guess
some
inspiration
for
for
what
can
happen
in
the
community
here
you
know,
take
the
opportunities.
If
there's
study
groups
I
mean
all
of
you-
are
here
in
the
lab,
which
that's
exactly
what
I'm
talking
about,
because
they
it
things
just
seem
to
open
doors.
A
You
know
when
you,
when
you
enter
into
some
of
these
different
programs
that
are
available,
you'll
notice,
that
different
doors
open
up,
and
so
it's
been
a
nice
journey
with
mark
through
the
course
through
the
study
group
through
the
hackathon
and
now
he's
the
first
guest
on
the
on
the
tec
labs.
So
I'm
pretty.
A
A
I
always
kick
it
off,
so
I'm
going
to
share
my
screen
here,
just
so
that
I
can
track
and
people
can
follow
along,
but
how
we
always
start
is,
if
you
check
out
the
tec
labs
channel
here
in
the
server,
and
you
see
the
pinned
messages
and
go
ahead
and
open
up
the
notion,
workspace
and
I'll
just
get.
I
don't
want
to
take
up
too
much
time,
we're
almost
five
minutes
in,
but
I'll
just
remind
everyone
open
up
the
notion,
workspace
and
mark.
You
might
want
to
do
this
too.
A
There's
this
attendance
sheet
so
for
anyone
who's
new
to
the
labs,
find
your
way
to
the
to
the
workspace
and
the
attendance
sheet
and
what
I
always
do
is
I'm
just
going
to
insert
one
column
left.
I
guess
this
is
lab
10
and
then
just
punch
in
an
emoji
or
anything.
You
want
to
write
in
if
you're
new
to
the
labs
add
a
new
row
here.
A
Anyone
can
edit
this
file
so
go
ahead
and
you
can
fill
in
any
sort
of
data
that
you
wish
your
name
it's
nice
to
know
what
operating
system
people
are
running
and
and
some
contact
information
for
whatever
kind
of
socials
you
want
to
add
and
then
we'd
like
to
use
emojis.
A
A
So
I'm
going
to
find
my
where
am
I
here,
oh
yeah
and
so
yeah
awesome
mark's
in
here,
vitor,
that's
great,
and
usually
it
takes
a
few
minutes
for
people
to
trickle
in.
But
with
that
so
now
that
everyone
has
the
workspace
available.
B
B
Something
not
quite
right
here.
How
can
I
do
that.
B
I
think
there's
something
some
some
issue
with
rights
or
something
permissions.
Maybe.
B
B
B
B
B
B
A
B
B
A
So
in
the
meantime,
for
anyone
who
wants
to
start.
B
A
Opening
up
some
of
the
resources
on
the
background
information
here,
such
as
energy
web,
I've
posted
in
the
lab
agenda
on
notion,
I'm
just
starting
to
punch
in
some
resources
that
I
had
curated
myself
when
I
was
actively
working
on
this
project.
A
B
I
can
I
can't
manage
this
this
show.
Maybe
you
can
put
up
your
screen
with
the
dev
post,
submission
of
ours.
B
A
B
Okay,
take
a
pic
from
the
from
the
it's
I
think
somewhere
in
the
token
engineering
channel.
Oh
yeah
of
the
ocean
protocol
study
group
ocean
study
group.
I
think
there's
the.
A
A
Yeah
here
tokenized
power,
balancing.
B
Yeah,
let's
see
so
we'll
give
a
short
introduction
about
what
submission
is
all
about.
B
So
basically
power
is
being
produced
by
a
power
plant,
but
you
have
to
consume
it
immediately
because
there
are
no
economic,
viable
ways
of
storing
it.
Yeah
that
you
have
batteries,
of
course,
but
they
take
away
a
lot
of
loss,
and
so
basically,
our
power
grid
is
architected
around
the
idea
of
power,
consumers
and
power
producers
as
a
means
of
you
need
to
consume
the
electricity
as
soon
as
it
has
been
produced.
B
So,
basically
immediately.
So
you
can
imagine
you
need
all
kinds
of
parties
that
are
getting
hold
of
this
balance
of
power.
So
we
need
to
have
an
understanding
of
what
are
the
power
consumers
in
a
certain
area
and
what
are
the
power
producers
feeding
that
consumption
profile?
So
basically,
these
are.
B
We
call
that
the
energy
partners,
big
energy
companies,
are
doing
a
prediction
day
by
day
and
hour
by
hour
of
how
much
power
they
predict
is
going
to
be
consumed
in
that
hour,
and
I
need
to
fill
up
that
prediction
pattern
with
power
production
capacity,
so
they
need
to
scale
up,
maybe
gas,
fire
power
plants
or
coal
fire
power
plants.
These
are
not
easily
scalable
by
the
way,
the
the
last
ones
but
gas
fire
power
plants
are
easily
being
scaled
up
as
much
as
the
consumption
is
needed.
B
So
basically,
these
are,
quite
you
know,
fast,
regulating
fast,
controlling
power
power
architectures.
B
If
you
put
in
a
lot
of
renewable
energy
sources
like
wind
farms
or
solar
panels,
you
can
imagine
that
the
power
production
will
have
a
somewhat
zigzag
profile
according
to
let's
say,
weather
conditions
or
other
kinds
of
influences
like.
If
the
sun
is
shining,
you
get
production
of
solar
panels.
If
it's
not
you
don't
just
the
same
as
as
with
wind.
B
So
basically,
what
we're
trying
to
do
is
need
to.
We
need
to
have
some
sort
of
production
scheme
which
we
we
can't
predict
really
in
real
time.
That's
the
other
way
around.
Am
I
still
on
audio
or
not
yep.
We
hear
you
yeah,
okay,
yeah,
great,
all
right,
because
I
can't
see
your
screen
yet
sean
is
that
true.
A
B
B
This
was
the
main
idea
of
the
submission,
and
once
you
have
let's
say
a
better
production
of
of
your
a
better
prediction
of
the
power
producing
profile,
you
could
say
that
this
is
a
stakeholder
that
is
more
valuable
than
a
stake:
a
power
producing
device
that
is
not
so
good
at
predicting
the
production
profile
for
the
next
hour
or
so
so
this.
This
is
really
a
dynamic
kind
of
kind
of
thing.
B
So
what
I
try
to
what
we
try
to
do
is
to
have
energy
devices
registered
in
in
the
energy
web
foundation.
It's
called
the
energy
origin
toolkit.
B
And
if
you
can
pack
that
to
some
sort
ocean
data
market
pool-
maybe
you
can
play
around
with
a
staking
on
that
in
order
to
give
a
signal
to
the
power
producing
device.
If
it's
behaving
correctly
or
not,
what
do
I
mean
by
that?
If
it's
behaving
correctly,
it
means
that
this,
this
power
production
profile,
is
predicting
what
it
should
be,
and
this
is
not
something
that
is,
that
is
really
common
in
the
electric
electricity
sector,
because
you
have
those
gas-fired
power
plants
that
you
can
easily
control.
B
B
B
So,
basically,
what
we
try
to
accomplish
is
pack
a
power
producing
device,
a
renewable
power
producing
device
to
a
ocean
data
market
pool.
So
we
we
modeled
some
sort
of
them
in
the
token
spice
model.
B
I
think
we
maybe
sean
you
can
go
through
the
token
spice
model
a
bit
just
in
a
while,
because
what
we,
what
we
are
trying
to
accomplish
in
the
token
spice
model,
is
to
model
these
stakeholders,
which
are
shown
in
this
picture,
as
agents
simulating
a
kind
of
behavior
as
if
they
want
to
predict
the
power
production
profiles
for
the
next
hour
sean.
Can
you
maybe
give
us
a
light
introduction
on
the
token
spice
model.
A
Sure
so
I'm
thinking
there's
a
couple
ways.
I
can
do
that.
I'm
gonna,
first
of
all
make
sure
that
I
share
the
repository
with
everyone,
so
this
is
kind
of
neat
this
story
as
well.
We
have
token
spice
which
started
with
ocean
protocol
and
I'd.
I
noticed
there
were
a
few
things
missing
from
this,
particularly
it
didn't
work
on
python
3.8
because
it
was
using
this
static
type
checker
that
that
just
was
had
breaking
changes
on
python
3.8.
A
So
I
forked
this
onto
longtail
financial
and
I
fixed
a
lot
of
these
issues
and
kind
of
updated
some
of
the
tests
and
then
mark
went
ahead
and
forked
it
off
of
long
tail
financial.
So
I
always
like
this.
You
can
see
on
github
like
the
the
trail
of
events
and
how
something
is
sort
of
evolving
in
the
ecosystem.
A
So
let
me
share
that,
and
this
is
because
trent
gave
a
overview
of
token
spice
and
how
it's
being
used
for
ocean
protocol,
and
he
said
this
is
really
how
he
sees
it
being
used,
is
forked
and
it's
it
kind
of
highlights
the
difference
between
token
spice
and
cad
cad,
whereas
with
cad
cad.
We
have
this
very
open
general
system,
it's
like
a
general
simulator
that
could
simulate
anything,
whereas
token
spice
is
more
of
an
engine
that
you
adapt
to
a
use
case.
A
So,
by
default
token,
spice
is
very
adapted
to
ocean
protocol
to
modeling
the
ocean
protocol
and
the
ocean
dao
and
the
ocean
token,
and
how
it's
meant
to
be
used
is,
rather
than
being
a
general
simulator
device,
it's
more
specific
so
to
adapt
it
to
a
different
use
case.
You
actually
want
to
fork
it
and
then
change
those
specific
details
that
you
find
throughout
the
throughout
the
framework.
A
So
trent
was
really
happy
when
he
saw
that
this
is
getting
forked
and
used
for
different
purposes.
So
let
me
share
this
in
our
agenda
here.
A
I
think
I
want
to
start
I'm
going
to
start
fresh.
So
just
imagine
I'm
going
to
go
into
a
workspace
here,
I'm
going
to
go
into
the
tec
and
I
was
just
doing
this
beforehand
so,
but
I'm
going
to
start
fresh,
so
I'm
going
to
remove.
Let's
see
token
spice
should
be
here
token
spice.
Oh
maybe
I
wasn't
working
in
here.
A
B
B
Basically,
these
are
the
things
you
shouldn't.
You
shouldn't
skip
that
in
order
to
get
the
thing
running,
you
need
to
run
ganesh
and
you
need
to
deploy
the
contracts.
Otherwise
it
doesn't
work
and
you
need
to
have
a
of
course
a
token
spice
environment,
python,
virtual
environment,
so
I
tweaked
the
environment.gmo
file
a
bit.
So
you
need
to
take
a
look
on
that
because
I
added
some
extra
libraries
for
that.
So
take
a
look
at
at
the
environment
environment.yml
file
and
see
what
what
I
I
did.
B
A
Okay,
so
this
is
so
these
environment,
yaml
files
are
used
with
conda
and
you
can
see
that's
the
default
instructions
here.
I
I
generally
don't
use
conda
on
my
system.
I
just
use
vanilla,
pip,
and
so
what
I
just
did
there
is
I
I
just
when
I'm
working
on
this
locally.
I
just
make
that
make
sure
that
my
the
environment
file
is
sort
of
synchronized
with
the
requirements
file,
because
I'm
just
going
to
use
the
simple
requirements
file.
A
So
what
I'm
going
to
do
is
I'm
going
to
use
my
virtual
environment
manager,
this
virtual
fish
and
I'm
gonna
make
a
new.
Well,
I
made
one
earlier,
so
I'm
gonna
actually
delete
it.
Let's
see
token
spice,
okay
and
I'm
gonna,
make
a
new
virtual
environment
called
token
spice,
creating
and
let's
see
which
version
of
python.
I
have
3.8.6
okay.
A
A
And
how
I
I
like
to
use
this
ipython
interactive
python,
which
uses
it's
kind
of
the
same
as
jupiter
notebook,
but
just
right
here
in
our
terminal.
So
this
is
how
I
can
check.
So
I
can
go
ipython
now,
I'm
in
python
and
I
can
go
okay.
Import
cad,
cad
yep
seems
to
work
yeah
from
cad
cad
dot.
A
B
B
Dependencies-
yeah,
okay,
sure,
maybe
you
can-
we
can
go
to
the
directory
structure
of
the
of
the
repo
in
order
to
explain
some
changes
I
made
with.
A
B
So
basically,
what
I
did
is
just
you
know,
copied
the
whole
token
spice
repo
it's
in
there
and
I
added
a
directory
cat
cat
in
the
directory
cat
cat.
I
put
everything
which
is
needed
to
run
a
capcat
simulation
around
the
token
spice
agents.
B
So
maybe,
if
you
can
look
at
the
token
spice
is
actually
a
simulation
environment
without
the
perks
of
cat
cat.
I
will
I
would,
I
would
say
so.
If
you
look
at
at
how
it's
it's
ordered,
it's
all
around
agents,
agents
doing
stuff
and,
as
you
can
look
into
the
well,
maybe
you
can
go
just
into
the
agents
yeah.
The
test,
yeah.
A
B
Great
you
can
look
at
it,
but
basically
these
are
the
agents
which
resemble
the
stakeholders
in
your
system
and
these,
of
course,
the
token
spice
agents
are
all
around
ocean
protocol
and
the
data
tokens
and
the
data
token
pools.
You
can
see,
of
course,
the
data
ecosystem
agents.
You
can
see
mentor
agents,
burner,
agents,
pool
agents.
B
This
is
this
is
an
interesting
one,
because
a
pool
agent
is
basically
the
the
the
agent
that
is
delivering
the
pack
to
the
energy
web
power
devices.
B
So,
basically,
as
trent
said
in
one
of
his
his
medium
blocks,
if
you
stake
on
some
sort
of
pool
it's
a
signal
of
curation,
it's
a
signal
that
this
data
token
pool
is
is
a
is
a
nice
thing
to
have
is
a
thing
that
has
value
or
something
like
that.
B
So
what
I
did
I
just
copied,
the
the
earlier
publisher
agent
and
the
staker
speculator
agents
into
a
ew
x
agent,
as
I
can
say,
to
have
a
a
difference
between
the
normal
token
spice
agents,
which
are
focused,
of
course,
on
the
ocean
market,
the
data
token
stuff
and
the
energy
web
agents
which
are
going
to
to
mimic
the
the
behavior
of
the
energy
producing
devices
and
the
stakeholders
around
that
which
sean
showed
in
the
in
the
earlier
picture
of
the
death
post,
submission
yeah,
yeah
exactly
that
that
stuff.
B
So
I
think
we
need
to
go
into
the
to
the
cat
cat
directory
now
and
look
at
the
agents
I
put
in
there
and
they
are
not
in
an
agent
directory,
because
cat
cat
is
basically
structured
around
a
different
kind
of
basically
a
different
kind
of
structure
than
token
spy.
Token
spice
is
built
from
the
ground
up.
Cat
cat
is,
of
course,
some
sort
of
a
simulation
environment.
I
would
say
which
has
all
sorts
of
nuts
and
bolts
already
stuck
until
it.
B
So
I
basically,
I
built
a
directory
simulation
underscore
abm
that
stands
for
a
simulation
based
on
agent-based
modeling.
That's
where
the
abm
stands,
for
you
have
also.
You
can
also
have
a
directory
a
simulation
around
differential
equations,
which
is
a
certain
cat
cat
thing
of
of
looking
at
at
a
things.
A
differential
specification
simulation
is
about
mathematical
specifications
in
how
let's
say
how
you
can
model
a
system.
B
B
You
should
recognize
certain
structural
things
about
cat
cat
and
basically
they're
around.
I
would
say
two
basic
files.
B
First
of
all,
you
start
with
the
state
variables.pi
file,
and
maybe
you
can
look
into
that
one
sean
and
what
I
did
there
is
basically
in
the
state.
Fair
bills,
you
you
define
the
structure
or
you
define
the
the
model,
the
model
variables
which
you
can
simulate
on
so
and
basically,
your
your
model
is
about
the
set
of
stakeholders
in
your
system,
so
the
set
of
agents
in
your
system.
So
if
you
scroll
down
this
this
list,
you
can
see
all
kinds
of
agents.
B
I
am
I'm
going
to
to
address
and,
as
you
can
see,
there
are
a
lot
of
original
token
spice
agents
as
well
just
to
get
it
running
because
you
know
there's
a
whole
ecosystem
of
agents
needed
to
start
up
before
you
can
do
a
simulation.
B
I
need
to
flesh
that
out
even
more
because
cat
cat
doesn't
need
these
all,
but
I
haven't
got
time
to
figure
that
out
yet,
but
basically,
what
you
can
see
here
is
that
I,
I
add,
some
publisher
agents,
an
ew
publisher
agent
and
an
ew
staker
agent.
Now,
what
does
that
mean?
An
ew
publisher
agent
is
basically
a
fork
of
the
publisher
agent
in
token
spice.
This
is
the
agent
who
is
responsible
for
publishing
data
token
pools.
B
So
if
you
go
to
ocean
market,
you
can
publish
a
data,
token
pool
yourself-
and
this
is
the
agent
basically
responsible
for
doing
that
and
if
you
look
at
it
from
the
perspective
of
the
energy
web,
this
energy
web
publisher
agent
is
a
stakeholder
which
has
been
represented
by
some
sort
of
power,
energy
producing
device.
So
basically,
I
can
say
I'm
the
owner
of
this
solar
power
plant
or
of
this
wind
farm.
B
B
This
is
the
link
between
energy
web
on
the
on
the
one
hand
and
ocean
market.
On
the
other
hand,
now
you
also
have
an
ew
staker
agents.
Basically,
these
are
all
kinds
of
speculators
who
are
thinking.
Well,
I
see
that
we
have
some
data
token
pools
of
these
wind
farms.
I
I'm
going
to
stake
on
it
because
I
believe
that
this
this
data
token
pool
this
power
producing
device
is
behaving
according
to
plan
and
I'm
going
to
stake
on
it.
B
And
if
I
don't
believe
that
he's
producing
according
to
plan
or
if
I
believe
his
his
behavior
is
bad,
it's
all
about
behaviors,
and
then
I
unstake
on
it.
So-
and
this
is
the
neat
thing
about
simulation,
if
you
go
into
maybe
sean,
you
can
go
into
the
implementation
of
the
ew
staker
agent
in
a
parts
agent,
mobile
parts
agent.
B
Yeah
you
can
see
if
trent
modeled
this
this
quite
you
know,
coarse
grained,
I
would
say
I
implemented
it
slightly.
Fine,
finer
in
a
finer
grain,
basically,
every
step
the
simulation
takes
every
time
step.
We
are
going
to
do
a
speculative
action
and
if
you
go
into
the
speculation
it's
going
to
see
if
there
are
any
pools
to
stake
on
and
might-
and
this
is
where
the
where
the
the
the
nice
thing
is
going
to
happen,
we
need
to
have
pools
to
stack
on
and
if
I,
if
I
look
at.
B
I
see
yeah
there's
something
wrong.
I
think
with
this
implementation.
Maybe
that's
why
it
doesn't
work
at
your
side
and
sean,
because
something
is
fishy
here,
but
okay,
no
problem,
it's
about
about
staking
on
pools
and
what
I
do
here.
If
you
scroll
down
a
bit.
B
Yeah
fishy
is
not
quite
right.
It's
not
it's
smelly.
A
A
Anyone
see
it,
I
don't
see
it
but
I'll
put
a
note
I'll
put
a
little
yeah.
B
Here
this
is,
this
is
thing
I
I
I
removed.
I
think
yeah
debug,
that
yeah
exactly.
B
So
what
I
did
is:
okay,
let's
look
at
if
I
hold
some
sort
of
balancer
pool
tokens
in
that
pool,
and
once
I
have
that
I
randomly
stake
on
it
once
in
a
while.
B
So
if,
in
the
half
of
the
in
half
of
the
the
time
steps,
I
will
stake
on
it
and
on
the
other
half
I
don't,
and
basically
I
think
this
is
this
is
the
still
the
the
old
simulation
of
token
spice,
which
has
not
been
completely
put
into
the
into
the
git.
I
think
sean.
B
Maybe
this
is
the
problem
because
I
really
did
did
did
make
some
some
really
changes
here
in
order
to
have
some
more
dynamic
staking
behavior,
and
this
is
the
the
stuff
we
really
need
to
take
care
of,
because
staker
agents
have,
of
course,
some
sort
of
behavior
behind
them.
As
I
said,
if
they
believe
this
power
device,
behavior
is
good,
they
stake
on
it
if
they
believe
it's
bad,
they
unstack
on
it.
A
Couple
cool
things
to
see
so,
first
of
all,
this
is
a
practice
that
trent
came
up
with
this
idea
of
labeling
magic
numbers,
and
if,
for
anyone
who
has
a
background
in
software
engineering,
you
might
be
familiar
with
this
idea
that
you
should
never
have
magic
numbers
floating
around
your
code.
You
know
like
just
a
number:
that's
not
labeled!
It's
not
given
a
variable
name,
but
trent
is
kind
of
bending
this
principle
and
saying
well
actually
to
reduce
abstractions
and
reduce
complexity.
A
He's
not
going
to
generalize
everything
out
to
like
an
interface,
which
is
what
you
would
normally
do.
You'd
pull
all
these
numbers
out
into
a
single
interface,
given
the
give
them
all
names
as
variables,
and
then
you
can
tweak
them
from
there
and
he
said
to
reduce
abstractions
and
reduce
complexities.
He's
not
going
to
do
that.
He
leaves
all
the
numbers
inside
the
engine
itself.
So
this
is
like
a
difference.
You
could
see
from
cad
cad
cad
cad
you're
gonna,
get
make
all
your
configurations
and
then
throw
it
into
the
engine
with
token
spice
it.
A
A
So
if
you
wanted
to
search
for
all
the
places
of
the
engine
that
you
could
configure,
you
could
search
for
all
the
instances
of
this
label,
and
so
this
is
what
mark
is
saying
right
now
we
have
a
very
simple
staking
policy
on
this
agent,
where
half
the
time
it's
going
to
unstake
10
percent
of
its
stake
and
let's
just
break
this
down
a
bit.
A
Is
this
idea
of
using
staking
as
a
curation
mechanism,
because
every
data
set
on
the
ocean
marketplace
has
its
very
own
bonding
curve
with
its
very
own
data
token
that
can
be
minted
as
people
come
and
stake
and,
as
you
note,
the
bonding
curve,
as
as
you
stake
more
and
mint
more
of
these
tokens,
the
price
moves
up
along
that
bonding
curve.
So
this
is
what
creates
this
curation
market.
A
Is
that
if
you
find
something
brand
new,
maybe
a
brand
new
data
set
that
no
one
has
staked
on
then
the
initial
minting
of
those
data
tokens
is
very
cheap
and,
as
more
people
meant
more
data
tokens.
That
price
goes
up.
So
if
you
find
a
data
set
that
hasn't
been
discovered.
Yet
you
can
stake
on
it
and
just
unstake
later
to
make
a
profit
being
one
of
the
first
signalers
of
a
quality
curation,
and
so
this
is
the
mechanism
that
mark
is
using
and
applying
to
to
data
producers
and
data
optimizers.
A
Actually,
rather
so
we
have
the
data
producers
with
their
solar
farm,
and
then
we
have
a
data,
optimizer
who's,
actually
making
predictions
and
outputting
predictions
into
ocean
ocean
market
on
on
the
future,
consistency
and
outcomes
of
these
power
producers
and
then
the
the
stakers
are
curating
the
optimizers
because
they
are
checking
the
accuracy
of
these
optimizers
and
staking
on
the
very
accurate
ones.
So
we
create
this
reward
incentive
for
optimizers
to
very
accurately
predict
what
the
energy
outputs
are
going
to
be
so
that
we
now
have
these
forecasts
and
we
can.
A
The
grid
can
can
use
these
forecasts
for
optimization
of
the
of
the
energy
flow.
So
this
is
a
really
neat
neat
mechanism-
that's
happening
here.
B
B
If
this
power
producing
device
is
behaving
correctly-
and
this
is
where
the
staker
agents
come
into
hand
come
in
handy
because
once
staker
agents
are
going
to
stake
on
it,
then
you
have
some
sort
of
a
signal
that
okay,
this
is
a
power
producing
device
which
is
behaving
correctly,
but
as
sean
mentioned,
you
can
use
that
also
the
other
way
around.
So
instead
of
having
staker
agents
already
staking
on
on
these
power
production
pools,
I
would
say
you
have
optim.
B
You
can
have
optimizer
agents
in
place
that
are
optimizing
these
pools
and
staking
on
these
pools
in
order
to
have
some
sort
of
signal
of
okay.
This
is
this.
One
is
going
to
predict
correctly
and
you
can
have
both
so
both
staking
on
just
you
know
your
normal
production
profile
data
sets
and
you
can
stake
on
the
prediction
of
optimizer
agents.
Data
sets.
So
you
have
a
really
neat
influencers
network.
I
would
say
around
publishers,
stakers
and
optimizers.
B
We
need
to
figure
that
out
yet,
but
well,
let's
say
if
we
can
and
if
we
can
figure
that
out,
but
actually
that's
something
for
the
for
the
next
session,
because
once
you
get
once
you
want
to
to
have
reinforcement,
lurging
a
learning
agent
in
place,
which
we
had
in
mind
for
the
optimizer
agent.
You
need
to
have
some
data
and
some
sort
of
reward
function
in
order
to
have
them
behave,
behaving
correctly
right,
sean.
A
Exactly
yeah,
and
so
an
optimizer
basic
or
a
reinforcement,
learning
agent
essentially
has
this
capacity
to
observe
an
environment
and
make
a
decision
so
it'd
be
really
interesting
to
implement
this,
and
this
was
the
content
of
my
presentation
that
I
gave
in
the
ocean
protocol
study
group
is.
We
could
have
these
reinforcement,
learning
agents
as
optimizer
agents
that
are
observing
the
data,
sets?
Oh,
no,
not
optimizer
agents,
but
rather
staker
agents.
A
Well,
there's
multiple
ways:
you
could
implement
machine
learning
in
here
I
mean
you
could
just
use
simple
machine
learning
like
forecasting
to
sort
of
take
in
for
the
optimizer
agents.
They
could
be
observing
the
production
capacities
of
the
of
the
power
producers
and
essentially
the
patterns
in
those
production
patterns,
and
they
could
be
making
forecasts
and
the
staker
agents
could
implement
reinforcement,
learning
where,
based
on
the,
what
they're,
observing
from
the
optimizer
agents
they
could
make
have
an
action
set
that
they
optimize
over
on
which
optimizers
they
want
to
stake
on.
A
So
I
mean
it's
pretty
neat.
In
fact,
every
agent
in
this
system
could
could
be
modeled
with
reinforcement,
learning
the
beautiful
thing
about
reinforcement
learning.
It's
a
nice
generalization
trick
where
all
you
need
is
an
environment
and
an
action
set.
So
you
it
gives
you
all
these
intelligence,
algorithms
for
agents
that
are
going
to
observe
something
and
then
have
some
set
of
actions
that
they're
going
to
choose
to
to
do
so.
B
Yeah
right
great
sean,
yeah,
excellent,
you're
right
everybody
can
be
a
reinforcement,
learning
agent,
of
course.
Maybe
you
can
give
me,
give
his
some
sort
of
demo,
because
I
I
don't
think
this
wrapper
will
work
as
I
can
see
it
right
now,
but
I
I
do
I
sent
you
an
a
movie
of
a
real
simulation
perfect.
Then
maybe
you
can
show
that
up
on
the
screen
right
now,
so.
B
A
Excellent
okay,
so
you
just
sent
me
a
link.
I
guess:
where
was
that.
B
B
Yeah
yeah,
okay,
so
basically
what
I'm
I'm
doing
right
now
is:
okay,
can
you
stop
it
right?
There
yeah.
B
B
I'm
not
sure,
I'm
sure
it's
it's
it's
going
to
fill,
but
let's,
let's
yeah,
let's
go
from
from
there
stop.
So
this
is
basically
the
command
which
is
has
been
put
into
the
readme
file.
B
So
you
can
try
that
yourself
once
you
have
to
set
up
the
environment
correctly
so
in
in
the
token
spies
it's
run
underscore
one,
and
this
this
one
is
just
you
know
a
a
fork
of
that
and
but
then
it's
it's
starting
up
the
cat
cat
simulation
and
don't
mind
the
arguments
they're
just
writing
stuff
to
a
output
test
directory.
I
guess
in
the
indie
repo.
B
A
B
B
A
These
are
the
agents
that
you
you've
taken
these
from
the
token
spice
engine.
Is
that
right,
so
that
these
agents
are
actually
have
the
capacity
to
interact
with
the
evm
ganache.
A
Is
amazing,
so
you've,
you've,
you're,
running
analog
mixed
signal?
Simulator
is
what
trent
would
call
it
we're
using
cad
cad?
So
it's
evm
in
the
loop
is
the
idea
exactly.
B
Amazing,
exactly
yeah,
so
what
you
see
is
here:
if
you
build
the
token
spice
environment,
you
can
all
only
see
info
ticks,
but
now
you
can
see
that
our
pools
are
being
published.
Now,
that's
the
people
statement
and
they
are
really
interacting
with
ganache
now,
so
they
are
going
to
do
that
and
what
I
did
is
I
put
some
energy
web
publisher,
their
unstake
and
their
stake
randomly
on
these
pools.
B
So
basically,
what
I'm
trying
to
accomplish
here
is
to
have
a
log
of
pools
which
are
being
staked
upon
which
are
basically
input
or
for
the
space
for
the
reinforcement,
learning
agents,
but
I'm
not
sure
sean.
You
have
to
help
me
out
on
this
here.
If
I'm
doing
that
correctly
of
is,
is
it
really
the
actions
and
the
observation
space?
We
need.
A
I
I
think
I
had
pushed
a
simple
example
when
we
were
doing
the
hackathon
that
was
sort
of
observing,
maybe
a
random
we'll
have
to.
Maybe
we.
If
we
have
time
we
can
dig
around
in
that
or
maybe
next
week,
but
let's
yeah.
A
B
So,
as
you
can
see,
there
are
randomly
staked
upon,
but
once
in
a
while,
I
think
every
three
every
eight
hours
and
new
pools
are
being
published,
and
these
are
data
token
pools.
You
can
stake
on.
Of
course,.
B
B
Yeah
yeah,
exactly
okay,
so
basically
this
is
this
was
how
far
I
got
until
now.
A
few
words
again
on
on
the
migration
of
token
spice
to
cat
cat.
B
I
really
need
to
flesh
out
a
lot
of
a
lot
of
more
stuff
because,
as
I
as
I
experienced,
cat
cat
is
not
really
a
font
of
object.
Oriented
big
objects
floating
around,
so
I
needed
to
flash
out
some
sort
of
object-oriented
code
into
functional
code
and
then
later
on,
I
I
made
it
objects
again
but
bear
with
me
because
I'm
not
a
pythonista
in
this
matter,
so
I
just
you
know,
fooled
around
a
bit:
I'm
not.
A
Sure,
just
a
quick
aside,
if
you
want
to
see
some
object-oriented
cad-cad
code,
vitor
and
santiago,
who
are
both
in
this
call,
along
with
random
shinichi,
actually.
A
I
don't
see
andrew
random
shinichi
is
called,
but
they
worked
on
this
object-oriented
model.
B
Yeah
yeah,
I
I
noticed
that,
but
you
know
I'm
as
I
say,
I'm
not
a
pythonista,
so
I
hacked
around
a
bit
in
order
to
get
it
running
and
focus
on
the
actually
on
the
cat
cat
simulation
getting
it
to
run
according
to
my
wishes.
So
as
you
as
you
can
imagine,
oh
maybe
sean
it's.
It's
a
good
idea
to
go
into
the
polymax
directory
of
the
cadcad
model
parts.
A
B
Basically,
these
are
the
the
the
policies
and
the
mechanisms
of
the
cut-cut
structure.
So
if
you
go
into
accounting,
for
instance,
this
is
where
I
I
account
for
all
the
all
the
the
things
happening
on
the
on
the
side
of
the
staking,
so
the
amount
of
stake
is
taking,
and
always
so,
basically,
I'm
on
policy.
This
is
a
policy
to
update
the
state
and
consumption
top
pi
logistics
of
pi.
These
are
all
kinds
of
stuff
that
this
is
really
neat
about
cat
cat.
B
You
can
really
have
a
fine
grained
idea
of
how
you
structure
your
policies,
so
basically
each
agent
could
have
a
policy
for
themselves
and
then
structured
in
this
in
this
way.
So
this
is
all
wired
up
in
the
of
course,
in
the
partial
state
update
block,
that's
something
left
and
then
scroll
down
a
bit
in
the
other.
Now
it's
over
in
the
directory
structure,
yeah
a
bit
bit
more,
underneath
partial
state
of
that
block
there
yeah!
B
B
I
think
that
that's
it
for
now
sean
as
for
me,
is
concerned.
A
This
is
incredible:
mark
you've,
really
you've
taken
the
whole
token
spice
engine
and
and
shoot
and
put
it
into
cad
cad
simulator,
which
is
just
a.
I
think,
that's
an
amazing
feat.
One
thing
that
I
I
want
to
highlight
so
I
think
mark
has
highlighted
a
lot
of
the
advantages
of
I
don't
know
if
I
should
be
so
opinionated,
but
might
say
advantages
of
cad
cad
over
token
spice.
Having
this
very
clean
interface,
the
partial
state
update
blocks
where
we
get
to
wire
up
exactly
what
we
want
to
change
over
time.
A
What
is
the
state?
You
know
it's
contained
in
a
very
clean
standard
way.
What
are
the
policy
functions?
What
are
the
state
update
functions,
so
I
want
to
highlight
one
other
sort
of
on
the
other
side
of
that
token
spice.
What
something
that
token
spice
is
really
focused
on
is
this
idea
of
hierarchical
verification
and
essentially
what
that
means
is.
It
has
a
very
comprehensive
test
suite.
So
once
you
run
when
you're
looking
at
token
spice
and
you
open
up
the
repository
you're
gonna,
you
wanna
go
through
the
instructions,
it's
gonna.
A
Have
you
run
the
ganache
blockchain
and
compile
the
contracts
and
then
the
main
the
sort
of
highlight
of
the
repository
is
to
run
these
tests,
and
I
think
I
can
do
that
in
a
few
minutes.
I
know
that
I
had
just
let's,
let's
see
I'll
I'll,
see
if
in
four
minutes
I
can
give
everyone
a
full
demo
of
sort
of
just
running
token
spice
from
scratch
and
getting
to
the
point
where
you're
running
the
tests.
A
So
first
you
gotta
grab
token
spice
and
then
you
also
gotta
grab
contracts,
ocean
protocol
contracts.
A
A
A
And
you'll
notice
a
lot
of
warnings
and
errors.
This
might
eat
up
the
time,
but
let's
see
what
we
can
get
to
and
kind
of
what
I'm
trying
to
showcase
here
is.
This
looks
like
a
lot,
but
it
ends
up
being
pretty
smooth
and
simple,
and
it's
nice
to
be
comfortable
running
a
local
evm,
because
even
with
mark's
adaptation
of
the
cad
cad
modeling
because
he's
borrowing
those
agent
implementations
from
ocean
protocol,
they
have
this
web
3
interaction
capacity
built
in
so
we
can
run
real
simulations
on
the
evm.
A
A
So
let's
deploy
our
contracts
so
we're
going
to
see
a
bunch
of
transactions
going
through
on
our
local
evm
as
these
contracts
get
compiled
and.
A
And
then
this
is,
I
love
this
part
when
you
see
all
green,
so
this
is.
This
is
running
all
the
test,
simulations
that
are
built
into
token
spice
for
all
the
different
agents
and
yeah,
essentially,
all
the
agents
so.
A
And
I'm
a
little
bit
over
time,
so
I'll
let
this
keep
running,
but
I
want
to
really
thank
mark
for
being
the
first
guest
host
on
on
the
lab
here.
I
think
this
was
amazing
and
mark
it's
really
incredible
work
that
you've
adapted
token
spice
into
cad
cat.
I
think
that's
really
cool
and
there's
so
much
value
here
to
be
explored,
so
I'm
really
looking
forward
to
the
next
session
mark's
doing
it
to
a
series
of
two
sessions,
so
he'll
be
back
next
week.
A
I
hope
that
everyone
tunes
in-
and
I
hope
everyone
found
this
as
fascinating
as
and
exciting
as
I
did-
and
it
was
incredible.
Thank
you
so
much.
B
Guys
and
see
you
next
week,
hopefully.
A
A
B
B
Sean
I
have
to
go
as
well,
and
maybe
we
can
grab
some
time
next
week
for
a
quick
call
about
reinforcement,
observation,
space,
happy.