►
From YouTube: Jeromy Coffee Talks - Bitswap
Description
Jeromy is talking about bitswap while preparing some delicious coffee.
A
So
yeah
I'm
gonna
make
coffee
and
then
talk
about
bit
swap
at
same
time,
ish
I'm
going
to
be
doing
this
coffee
I
got
in
Spokane
last
weekend
for
last
week.
It's
Mexican
and
Mexican
and
Kenyan
blend
I
need
to
be
in
the
camera,
and
it's
pretty
good
but
grind
it
up
here
with
my
grinder
and
then
I'm
going
to
make
it
in
the
coffee
side,
then
I
need
to
go
grab
some
water.
So
one
second.
A
Alright,
so
there
went
my
chair,
it's
swap
for
those
who
don't
know
is
the
part
of
my
bffs
that
manages
like
sending
data
between
peers.
So
when
I
want
something,
I
tell
bit
swap
hey
I
want
this
thing
and
then
bits
talk,
goes
and
manages
asking
everybody
in
the
network.
For
that
thing,
it's
similar
to
how
bittorrent
does
it
in
that
it
will?
A
You
know,
broadcast
the
things
wants
and
ask
of
a
Pierce
floor
and
hope
it
gets
them,
but
it's
different
because
in
bittorrent
the
people
you're
talking
to
our
only
people
who
are
also
interested
in
the
same
file
Oh
an
IPF
esas
case
you're
talking
to
a
lot
of
people
who
may
or
may
not
have
little
pieces
of
the
file
or
may
not
even
care
at
all.
So
you
have
to
adapt
the
outer
just
a
little
on
that
regard,
to
get
things
to
work
right.
A
So,
on
the
coffee
front,
this
is
a
copy
siphon
for
anybody
who
hasn't
ever
seen,
one
before
it's
pretty
cool.
You
put
water
in
here
and
then
I'll
put
this
with
a
filter
in
it
in
the
top
and
then
I'll
heat
up
water.
In
the
bottom
it
boils
and
yeah
you'll
see
that
a
second,
but
to
boil
the
water
use
a
little
butane
burner
be
dead
and
go
for
butane
in
it.
A
And
I
makes
about
two
cups
of
coffee
which
isn't
really
enough,
especially
since
I
normally
do
this
for
both
me
and
my
roommate.
But
since
he's
not
here
right
now,
this
will
be
plenty
and
still
have
600,
milliliters
or
so
and
don't
know
if
you
can
see
the
cool
part
down
here,
but
a
fire,
and
then
I
just
put
this
guy
in
here
and
whip
that
the
boil.
So
while
we
were
at
the
boil,
let's
talk
about
where
this
swap
is
today
so
today,
bit
swap
has
an
extremely
naive
algorithm.
A
A
Because
that
really
gets
in
the
way
of
working
on
all
the
other
parts
of
the
system,
which
you
know
it's
good
for
us
right
now.
The
other
thing
is
that
it's
very
very
fast
for
the
simple
case
of
one
person
having
a
file
and
another
person
wanting
it
cuz,
you
know
you
don't
worry
about
waiting
or
doing
the
blocks
or
like
choking
peers
and
doing
all
the
BitTorrent
logic
behind
that.
A
So
it
works
right
now
works
pretty.
You
know,
surprisingly,
well
honestly,
for
how
simple
of
an
approach
it
is,
but
we
need
to
change
that
we're
starting
to
notice
that
there
are
lots
of
people
get
blocks
being
sent.
If
there
are
three
people
in
there
who
have
something
and
I
asked
for
it,
they're,
probably
all
three
going
to
send
it
to
me,
even
though
my
best
effort
is
to
as
soon
as
I
get
one
block
tell
the
others,
I
don't
want.
It.
A
They've
probably
already
started
sending
it
to
me
at
that
point,
which
wastes
a
lot
of
bandwidth,
and
you
know
it's
bad
for
the
network,
so
we
need
to
figure
out
well
either.
How
do
we
decide
which
blocks
to
send
or
how
do
we
decide
which
blocks
to
ask
or
on
a
fleet
bowl
so
deciding
who
dis
who
to
ask
for
what
is
a
hard
problem?
It's
you
know
been
solved
ish
in
bittorrent,
but,
as
I
mentioned
earlier,
our
use
case
is
little
different
from
bittorrent.
A
A
Well
now
so
they
asked
me
for
a
block,
but
I
really
want
to
set
it
to
them
and
once
that
changes,
the
whole
game
gets
fun
because
well
now
I
have
to
figure
out
okay,
how
many
people
do
I
have
to
ask
to
guarantee
that
at
least
one
of
them
is
going
to
send
me
a
block
in
a
certain
amount
of
time
and
I'm
not
going
to
get
too
much
duplicate
data?
And
that's
that's
something
that
we
want
to
have
you
know.
A
Maybe
an
agent
set
up
for
maybe
a
smart
agent
that
can
tweak
these
values
and
adjust.
You
know
for
you,
probably
at
first
it'll
be
manual.
You
know:
okay,
here's
all
your
input
values,
here's
ABCDE!
You
can
tweak
them
until
you
find
a
good
value,
but
you
know
more
advanced.
More
advanced
code
is
going
to
just
do
that
for
you
and
that's
good,
because
as
the
network
grows
and
files
become
more
or
less
popular
and
you
know
different
aspects
of
geolocation
versus
you
know,
network
churn
and
where
you
are
and
your
bandwidth
versus
their
bandwidth.
A
Those
factors
are
gonna
change
like
really
rapidly
and
changing
the
manually
is
isn't
going
to
be
super
feasible.
So,
like
the
number
of
people
you're
requesting
to
talk
from
versus,
you
know
how
many
you
sent
to
gonna
change.
Also,
I,
don't
know
if
you
can
see
this
but
the
water's
starting
to
come
up
here.
It's
kind
of
like
really
cold
in
my
apartment,
so
this
is
taking
longer
to
boil
than
it
probably
should.
A
A
The
actual
mechanisms
are
it's
not
an
actual,
like
request,
I,
don't
send
you
request
and
wait
for
it.
You
send
me
things
back
kind
of
like
a
traditional
RPC,
it's
more
of
a
it's
like
a
market.
Exactly
how
one
puts
in
the
paper
is
it's.
You
know,
I,
basically
make
it
known
that
I
want
these
things.
Yeah
I'm
standing
here
with
a
sign
saying:
hey,
so
what
I
want
and
people
can
walk
by
and
if
they
have
the
blocks,
they
can
say.
A
Oh
here's
this
book
and
it's
completely
you
know,
choice
based
as
opposed
to.
You
know
me
directly
targeting
you
and
waiting
for
you
to
return
that
which
I
think
is
actually
pretty
cool
as
far
as
protocols
go,
because
that's
that
whole
you
know
register
your
interest
is
more
like
a
pub
sub.
Almost
now,
if
it's
all
kind
of
like
a
pub
sub
I,
don't
I,
don't
see
one
disagreeing
song
and
I'm
gonna
go
with
that.
B
So
so,
just
to
see
I'm
Brandon
Iraq
example,
we
have
any
specific
strategies
when
we
are
searching
for
a
while
looking
for
a
walk
but
all
of
our
beers,
and
that
we
contacted
don't
let
it
fall
and
actually
the
bucket
we
are
looking,
is
a
folder
and
is
a
UNIX
FS
folder.
So
it
has
like
a
bunch
of
wings
and
I
already
know
which
blocks
like
oh
I
know
it's
not
what
I
will
want
to
veg
from
them.
B
A
So
good
question:
we
have
an
optimization
right
now,
where,
when
I
try
to
read
like
a
folder
or
a
file
that
has
sub
locks
I,
we
do
a
bit
swap
request
for
multiple
blocks
at
the
same
time,
and
then
we
initially
just
look
for
the
person
who
has
the
root
of
whatever
a
request
is
and
then
so
we
just
find.
We
do
one
search
for
that
and
then
after
that
search
is
done.
A
We
send
that
person
all
of
our
want
list
and
that
works
out
pretty
well
in
most
cases
because
say
we're
requesting
a
file.
Yeah
say
we're
requesting
a
file
generally.
If
somebody
has
the
you
know,
top
metadata
block
of
a
file
they're
also
going
to
have
the
rest
of
the
file,
which
saves
us
a
lot
of
time
and
then
bits
wobble.
Also,
if
it
doesn't,
you
know
it
doesn't
get
what
it
needs
within
a
certain
time
out.
It
will
stop
and
it
will
go
actually
request
each
individual
one
again.
A
A
A
On
you
don't
want
to,
like
you
know,
stir
vigorously
at
first,
because
you
want
it
to
what's
known
as
a
bloom
where
the
coffee
beans
get
heated
up
and
they
release
all
the
carbon
dioxide
in
there
and
the
carbon
dioxide's
like
a
lot
of
a
kind
of
bitter,
your
type
flavors.
So
you
want
that
to
sit
on
top
and
just
leave
to
stir
it
that's
going
to
get
inside
the
coffee
and
it's
going
to
add
a
little
extra
bitter
flavor.
A
So
I
stir
this
now
wait
for
about
a
minute
and
10
seconds,
give
or
take
a
few
seconds,
and
so,
while
this
is
going,
the
water's
still
boiling
in
the
bottom
and
every
once
and
while
I
get
a
little
air
bubble
that
gets
pushed
up
through
here,
but
the
most
part
water
stays
up
here.
Eyre
stays
in
the
bottom
kind
of
weird,
but
yeah,
that's
physics.
So.
A
A
And,
like
my
favorite
part,
because
it's
the
only
reason
the
coffee
is
coming
back
down
is
because
the
air
in
here
is
cooling
and
pulling
the
coffee
through
the
really
tight
filter
in
here
and
I'd.
Actually,
so
the
part
that
confuses
me
is
this
phone,
the
phones
because
they're
somehow
air
in
the
tube
in
the
middle,
but
I,
don't
know
how
there's
air
in
the
tube
in
the
middle,
because
there's
just
water
above
it
and
that
that
doesn't
make
too
much
sense
to
me.
A
B
A
B
A
The
provides
buffer
is
so
it
swap
has
kind
of
two
operations
right,
you
have,
you
can
ask
for
blocks
and
a
new
sending
blocks.
So
we've
kind
of
the
provides
buffer
is
an
interesting
thing
because
it
doesn't
quite
belong
there.
It's
when
we
add
a
new
block,
we
add
it
through
bit
swap
so
we
tell
bit
swap
to
add
the
block
and
what
it
does
is.
It
will
grab
the
key
of
that
block
and
then
it
will
write
the
block
to
disk.
You
know
through
your
data
store
and
then
with
the
key.
A
It's
going
to
do
a
DHT
provides
the
network
and
that
buffer
is
when
we're
adding
tons
of
blocks.
We
can
add
a
block
much
faster
than
we
can
actually
provide
it
to
the
network,
so
that
buffer
is
us,
just
buffering
up
blocks
to
be
provided,
and
it's
kind
of
misleading
right
now,
because
what
we
do
is
for
every
block.
We
spawn
off
a
worker
thread
to
consume
that
and
provide
it.
A
But
yeah,
so
that's
what
that
does
is
that's,
basically
us
adding
blocks,
and
it's
buffering
provides
a
reason
for
that
us
adding
blocks
through
bit
swap
is
so
that
we
can
make
kind
of
like
a
single
interface
or
block.
So
you
can
say,
get
a
block
and
put
a
block
right,
and
you
can
also
check
if
you
have
a
block
and
that
interface
is
just
done
through.
You
know,
what's
called
the
block
service,
which
is
a
wrapper
around
bit
swap
so
we
can
have
this
nice
interface
that
feels
like
it's
local.
A
We
can
also
have
the
same
interface
for
just
local
blocks,
but
it
you
know
if
you
ask
for
block,
it's
gonna,
get
whether
it's
on
your
disk
or
whether
it's
on
somebody
else's
disk,
and
it
makes
it
really
cool
because
you
know
just
feel
like
you
have
everything
right
there,
no
matter
what
it
is.
Yeah.
B
A
Okay,
so
I
thought
about
adding
a
want
command
and
it
probably
probably
should
just
for
the
API
to
feel
complete,
but
this
swab
want
is
just
a
BFS
block
get
or
a
PFS
cat
a
given
hash.
It
does
basically
the
same
thing:
minus
the
spitting,
the
data
out
so
I
mean
yeah.
We
probably
should
add
a
want
command.
That'd,
be
pretty
simple.
If
you
agree
want.
A
B
Was
just
thinking
if
I
should
bring
as
a
question
if
there
is
any
discussion
or
like
next
features
but
we'd
swap
when
it
comes
from
well
tit
for
tat
and
like
a
bit
swap
but
like
uses
other
bookstores
to
like
just
being
the
two
IPA
fest
like
use,
an
external
power
storage
somewhere?
And
if
there
is
any
indication
right
now-
and
this
is
the
video
that
we
go
potty
and
where
those
discussions
so
that
are
really
enjoying
that
Petra.
A
Sure
do
we
could
talk
about
that
right
now,
but
wait
so
there's
a
few
like
future
things
with
this
slop,
like
different
categories,
think
about
the
one
you
mentioned
is
a
different
like
essentially
transport
for
Viswa,
so
like
right
now
bit
sloppy
uses
I
PFS,
you
know
to
get
blocks
essentially,
but
you
could
also
theorized
a
bit
swap
that
just
used
HTTP
like
for
every
time.
I
wanted
to
go
fetch
a
block
it,
which
is
query
some
HTTP
endpoint
for
the
block
and
if
it
gets
that
it
gets
it.
A
A
A
Well,
if
we
had
a
way
to
request
every
other
block,
you
know
that
becomes
a
lot
easier
to
mitigate
the
duplicate
blocks,
you're
receiving,
and
so
that's
going
to
be,
it's
going
to
be
fun
and
a
little
tricky
to
write
be
just
because
of
the
way
I
PFS
itself
works
not
necessarily
bit
swap,
but
with
I
PFS.
It
becomes
tricky
because
hashes
are
put
around
academia
XOR
table,
so
you
know
things
might
not
necessarily
be
close.
A
So
asking
you
know
one
person
for
all
of
these
hashes,
even
though
they're
not
like
likely
to
have
them
it's
a
little
weird,
but
we're
already
making
assumptions
that
are
similar
to
that
anyways
with
the
you
know,
request
just
one
person
for
all
of
these
in
the
tree.
So
it's
probably
a
good
a
good
thing
to
make
anything
narrow
on.
C
If
you
do,
if
you
order
all
of
the
nodes
in
this
graph,
send
me
every
other
note
or
maybe
like
send
me
every
like
mod
five
of
the
nodes
in
the
graph,
which
allows
you
to
express
the
same
selector
to
multiple
peers
and
then
ask
for
different
sub
selections
of
the
graph,
which
makes
them
send
you
different
types
of
nodes
to
you
like.
Send
you
basically
like
a
set
of
interleaved
nodes,
and
that
becomes
like
a
really
powerful
and
expressive
replication
model,
because
you
can
simultaneously
give
out
these
queries
with
these
different
constraints.
B
A
C
C
Oh
so,
in
this
one
example,
so
this
is
not
what
the
language
would
look
like,
but
suppose
these
are
the
kinds
of
things
we
would
want
to
be
able
to
do
right.
So
in
the
first
line,
there's
like
here,
dollar
ref
would
be
replaced
by
specific
cash
or
something
or
you
might
even
consider,
having
variables
/
star
would
say,
like
hey
get
everything
underneath
this
one
hash,
/
star
star,
would
be
kind
of
like
in
UNIX,
get
everything
underneath
it,
but
then
you
can
start
doing.
C
You
could
select
specific
nodes
underneath
the
graph,
even
if
you
don't
know
the
hashes-
and
in
that
case,
on
the
third
case,
where
you
say:
ref
/
parents,
life,
parent,
/,
transaction,
/,
10.
In
that
case
the
other
node
is
meant
to
send
you
a
ref,
then
ref,
current,
then
ref,
parent
parent,
then
ref
parent
transaction
and
then
the
last
one
with
the
10.
C
In
that
way,
you
can
verify
every
note
along
the
way
to
get
the
last
one
and
then
maybe
with
different
ways
of
expressing
like
whether
you
want
the
whole
path
or
you
only
won
the
last
node.
In
the
fourth
example,
there
I
have
the
idea
of
being
able
to
express
it
kind
of
like
a
as
Jeremy
Oh,
saying,
a
regex,
where
you
can
that
one
with
a
trunk
/
grow.
C
So
I
can
get
the
the
first
few
chunks
of
data
that
represent
the
movie
and
immediately
start
streaming.
Those
that's
how
you
might
get
like
really
fast
streaming
over
a
graph,
but
this
could
be
any
kind
of
graph
right,
like
their
you're
no
longer
tied
to
the
layout
of
the
graph.
You
can
specify
a
query
that
that
optimizes,
this
kind
of
behavior,
the
last
one
is
where
you
might
consider
like
embedding
expressions
within
another
one
right,
so
you
could
do
you
could
select.
C
You
can
first
do
a
selection
on
one
like
graph
and
pull
out
of
value
from
that
graph
and
then
use
that
value
for
another
query,
and
so,
if
the
other
node
has
rough
one
ref
to
then
it
can
resolve
this
query
entirely
and
return
to
you
all
of
the
nodes
that
you
need
to
verify
it
for
yourself
and
do
so
in
one
like
without
having
to
do
these.
These
are
duties
like
what
we're
trying
to
avoid
here.
C
At
the
end
of
the
day,
is
these
huge
gaps
where
like
if
I
were
to
request
every
single
one
of
these
nodes
iteratively,
I
would
have
first
had
to
say:
hey
give
me
a
rough
one.
Once
I
get
rough,
one
I
figure
out
the
hash
for
the
next
one
and
say:
hey,
give
me
the
next
one
and
and
so
on,
and
that
turns
out
to
be
really
really
slow.
And
so,
if
you
can
express
this
in
a
way
of
the
other,
node
can
send
me
the
whole
thing
and
that's
much
better.
C
So
you
can
imagine,
saying:
hey
I
want
to
cash
a
sub
selection
of
the
graph,
because
I'm
gonna
go
offline
and
want
to
cache
like
some
selection
of
a
graph
for
like
offline
reading
or
whatever,
and
you
can
select
what
that
that
selection
looks.
Like
with
a
notation
like
this
cool,
so
anyway,
no
not
clear
yet
what
the
actual
language
will
look
like,
but
I
said
the
sort
of
things
that
were
that
we
want
out
of
it.
Yeah.
A
I
really
like
so
like
just
looking
at
this
listening
to
you
talk
about.
It
makes
me
think
that
we
can
generalize
it
this
just
not
be
like
just
for
bit
swap
we
can
make
this
just
be
the
I
PFS
path,
format
itself
and
you
know,
take
the
path
package
that
we
have
and
put
all
of
this
logic
there.
So
that
way,
we
have
just
one
really
cool
way
of
defining
queries
over
to
rekal
bags,
which
is
really
what
we're
doing.
C
A
Would
be
pretty
cool?
Also,
you
were
talking
about
like
index
selection,
I,
think
it'd
be
really
cool.
We
probably
we
can.
We
can
avoid
like
index
selection,
pretty
much
completely
when
we
move
to
forcing
named
links
and
then,
like
all
of
our
files,
all
the
chunks
are
just
numbers.
So
that
way
you
can
actually
the
name
of
the
link
is
literally
just
the
block
index
and
that
way
yeah.
A
C
A
C
To
think
what
are
the
features?
I
was
going
to
talk
about
I'm,
pretty
excited
about
like
hooks
for
other
strategies
and
like
being
able
to
express
agents
that
download
other
things,
and
this
is
many
things
like
I
think
Jeremy.
You
might
want
to
talk
about
like
different
kind
of
bots,
that
you
might
write
one
of
the
things
that
I
find
interesting
about.
C
C
So
you
could
turn
around
entirely.
How
CD
ends
are
done
today
by
hooking
up
protocols
like
these,
because
you
can
create
other
kinds
of
behaviors
like
collect
on
delivery
through
cryptocurrency
and
if
you
make
it
agnostic
to
the
kind
of
cryptocurrency
and
create
a
protocol
that
allows
any
to
our
class
notes
to
negotiate
the
protocol
and
say
hey
like
do
you
use
Bitcoin,
or
do
you
use
the
theory
amor?
Do
you
use
like
some
other
set
of
protocols,
then
delete
like?
C
C
Yep
and
there's
also
things
like
you
know
how
quickly
you
can
send
it
to
me,
are
like
bandwidth,
taking
into
account
right,
so
I
might
be
willing
to
count.
Give
you
a
higher
share
of
credits,
I
guess
if
you
can
get
the
content
to
me
faster
than
everybody
else,
so
that's
where
we
can
unlock
a
whole
bunch
of
research
that
has
been
dormant
in
this
area
for
a
long
time.
C
It
then
becomes
really
easy
to
play
around
with
a
kind
of
stuff,
because
when
those
papers
came
out,
they
didn't
really
move
the
needle
very
much
because
BitTorrent
was
very
closed,
though
it
was
an
open
source
protocol.
It
was
very
close
and
lock
down
ecosystem
where
it
was
actually
free,
difficult
to
change.
C
So
then
have
them
be
able
to
run
on
some
subset
of
nodes
that
want
to
do
this,
where
you
might
say,
hey,
like
I,
might
be
willing
to
engage
in
a
in
a
protocol
as
long
as
it's
signed
by
a
certain
set
of
people
and
like
you
may
not
even
have
to
have
the
code
before
hand,
you
can
then
get
the
code
and
then
engage
in
that
in
that
strategy.
Yeah.
A
That's
really
the
the
idea
of
like
live
updating
code.
You
know
as
long
as
it's
you
know
it's!
Okay!
If
this
person
sends
me
code,
I'm
gonna,
run
it
and
then
being
able
to
you
know,
do
a
couple
different
things
like
recover
from
failures:
they
ship
you
code
that
doesn't
work
for
you
being
able
to
not
break
and
just
reverts.
The
old
stuff
I
reminds
me
of
some
paper.
I
was
reading
about
ibm's.
I
think
it
was
ibm's
mainframe
stuff.
B
Yeah,
if
you
want
to
look
at
that,
look
at
laying
DM,
which
does
that
on
a
vm
level
you
can
so
holding
systems
can
be
patched
live
because
the
way
it
was
designed
as
downtime
wasn't
allowed,
because
otherwise
nobody
would
have
a
telephone.
I
was
Kanaan
cool.
It's
so
good!
It's
pretty
amazing!
If
you
just
inject
your
code
and
just
all
right,
let's
upgrade
yep
and
you
can
just
do
like
upgrades.
A
A
You
cool,
though
there
was
one
other
thing
on
top
of
a
bit
swap
is
so
once
you
get
to
the
point
where
you
know
it's
generalized
requests
and
sending
mechanism,
you
can
start
scripting
and
you
can
start
making
a
is
that
just
sit
on
the
network
and
do
data
trading
they
can
listen
in
for.
What's
you
know
in
high
demand
and
then
get
that
themselves
and
we
host
it
and
then
you
get
it
becomes
this
place
where
you
can
start
writing
box
just
to
try
and
maximize
their.
A
You
know
value
to
the
network
and
I
think
when
we
get
to
that
point,
it's
going
to
be
really
fun
because
I've
seen
other
bak
communities
before
you
know
like
program
or
fight
type
things
where
you
like
script,
a
robot
and
a
fight
in
some
virtual
arena.
Things
like
that
that
are
actually
beneficial
to
everybody.
If
we
can
get
you
know
some
sort
of
competition
or
challenges
set
up
to
build
a
better
bit
swap
AI
I,
think
it'd
be
really
cool
things
happening.
You
know.
B
A
C
C
B
B
Very
efficient
and
but
yeah
I
think
that
this
really
good
and
like
we
can
definitely
do
for,
like
all
the
other
parts
of
like
a
events
and
also
like
after
posting,
this
12
yr
c,
we
can
see
other
people
I
interested
in
being
the
ones.
I
dreaming
questions
like
specific
questions
about
that
specific
one
hour
difference,
so
that
we,
after
the
presentation
of
that
component,
you
get
like
a
very
like.
What's
the
main
points
of
the
community
when
they're
trying
to
understand
part
I
like
it
we're.
A
B
C
Before
last
sign
of
I
had
one
working
and
that's
what
which
I
guess
it's
it's
a
just
like
improving,
perf,
think
like
one
of
the
things
that
will
help
I
PFS
/
/,
a
lot
will
be
getting
the
dike
pattern.
Half
note
that
we
explained
before
in
and
a
few
other
like
improvements
to
the
top.
So
that's
worth
noting
for
people
out
there
that
more
work
on
this
web
stuff
will
will
be
really
useful
to
make
I
because
much
faster
and
optimized.
C
And
it
has
a
lot
of
room
to
grow
and
get
better,
and
a
particular
like
one
of
the
pretty
interesting
branches
and
forth
is
building
simulators
for
understanding
how
this
hope
is
doing
things
and
how
well
it's
working
and
yeah
modeling
the
behavior
there.
So
both
Karthik
and
I
on
have
been
looking
into
that
to
some
degree
and
that'll
be
probably
like
energy.
That
would
be
an
interesting
chat
to
have
on
its
own.
Just
like
simulating
and
modeling
that
network
yep
that'd.