►
From YouTube: Weekly Sync 2020-05-15
Description
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#heading=h.uvm7zslc5152
A
A
So
the
auto
creation
for
the
definitions
of
spec,
slash,
sub
specs
stuff,
the
stuff
that
suits
doing
put
names
for
things,
and
then
we
need
to
support
a
secret
value
on
the
web
hook
that
we
just
did
the
demo
of,
and
we
might,
if
we,
if
we
end
up
with
extra
time
today,
we'll
just
run
through
the
demo
quickly
this
time
because
it
got
merged,
and
then
we
probably
want
to
support
the
this.
This
would
be
ideal
just
because
I
think
it
would
make
things
a
little
bit
simpler,
but
it's.
A
Okay,
so
basically
what
we'd
be
doing
is
having,
like
you
know,
you'd
put:
where
did
I
do
this,
oh
yeah,
so
we'd
say
like
oh,
you
know
like
you
could
put
in
something
like
you
know
the
Python
path
profile
and
then
calling
the
function
just
like
we
have
with
the
rest
of
the
amateur
points,
and
it
would,
you
know,
work
with
the
data
flow
create
command,
and
it
would
also
like
load
that
flat
that
function
as
an
operation.
When
we,
when
you
actually
run
the
data
flow,
and
so
then
you
could
write
something.
That's
that's.
A
You
know
you
could
have
a
little
Python
file,
that's
just
basically
one
function
and
with
the
stuff
that
we've
done
that
suit
Honshu
is
doing
with
the
auto
definition.
Creation,
you'd,
basically
just
be
able
to
to
have
functions
in
files.
You
wouldn't
even
have
to
decorate
them
with
up.
If
you
just
point
at
them,
you
could
say
if
you
aren't
decorated
without
decorated
with
op
and
then
use
it
as
an
operation,
so
that
would
sort
of
really
streamline
things
which
I
think
would
be
good.
A
We
need
to
expose
the
save
load
and
run
functions
from
the
high
leveled
up.
He
why,
in
the
know,
async
that
py
and
so
basically
that's
these,
and
so
some
of
these
we
don't
have
issues
for,
but
we
need
to
have
issues
for
them.
So
this
is
the
new
async
API
right
now
it's
just
got
the
machine
learning
ones
and
then,
but
the
high
level
itself
a.
A
We
should
have
everything
that's
in
high
level,
so
we
need
to
add
the
save
and
load
and
run
so
that
you
can
basically
just
for
people
who
aren't
using
async
io
already
for
those
people.
It
can
be,
as
you
guys
all
know
like
we.
Never
you
see
a
think
and
wait
for
the
first
time
there
you
go
I,
don't
know
what
that
is,
and
so
that
can
scare
people
off.
So
we
need
to
have
versions
of
things
that
don't
use
that
for
for
people
to
not
scare
them.
A
Let's
see
we
wanted.
I
would
love
to
see
this.
The
DEF
feature
to
feature
stuff
in
the
next
release.
That
would
be
really
sweet.
I
guess
that's
gonna,
be
I,
think
that
does
a
good
job
of
you
know,
making
things
easier
to
understand
and
also
easier
to
program,
because
we
don't
have
two
kinds
of
features
and
then
the
amnesty
image
resizing
now
that
we
have
the
data
flow
processing
source
that
would
be
really
cool.
I
know
suction
on
we
already
starting
work
on
that
put
names.
A
And
then
the
okay,
so
this
is
also
probably
sock
Shawn,
but
if
somebody
else
wants
to
jump
on
it,
the
DF
of
the
mail
edit
command,
there's
some
issues
up
we're
regarding
the
Edit
command,
documenting
it
and
making
it
work
like.
Basically,
we
could
just
take
the
Edit
command
and
make
it
so
that
it
you
give
it
a
source
and
you
give
it
a
data
flow,
or
maybe
we
just
give
it
one
source.
Basically
we're
just
gonna
say
you
know:
here's
the
Edit
commands
like
it
takes
a
source
and
it
just
calls
update
on
it.
A
After
it
it
goes
it
does.
You
know,
async
for
record
and
records
and
then
record
update
right
for
a
source,
not
update
record,
so
that
just
lists
everything
and
then
saves
it
back
right
and
if
you're
using
the
data
flow
source,
then
you'll
end
up
modifying
all
of
the
records
right
and
then
I'd
like
to
see
the
should
I
sub
flows,
stuff,
I,
think
yeah.
She
went
to
do
some
documentation
on
some
other
projects
and
so
would
be
plentiful
to
get
that
in
I'll.
Talk
to
him.
A
I
just
talked
to
him
and
saw
just
to
ping
him
and
see
see
if
that's
around
yesh
varshan
are
you
on
the
call
or,
let's
see
don't
see,
probably
not
I,
guess
so,
and
then
there's
this
thing
all
go
over
this
after
I.
Do
it,
but
basically
like
sort
of
security,
stuff
dependency,
analysis
and
listing
and
enumeration
and
versions
and
things
it's
it's
a
mess,
I'm
going
to
add
it
to.
A
Should
I
and
so
we're
going
to
hopefully
tame
some
of
the
mess
there
with
that
maybe
fix
the
automating
classification
demo,
because
Naima
brought
up
that
there's
issues
with
that
and
we
also
have
an
open
pull
request
already
that
it's
going
to
be
changing
things
there.
So
I
think
the
solution
is
really
just.
We
need
to
complete
this
pull
request.
A
A
B
B
A
B
A
A
Okay,
great
okay,
then
there's
an
open
question
on
Windows
support
and
I
know:
Nayeem
you're,
the
only
other
person
I,
think
I,
don't
know
if
anybody
else
I
think
now
you
have
Windows
and
I
sometimes
run
this
one
like
this.
So
but
you're
already
doing
the
def
feature
stuff,
so
I
might
just
poke
at
it.
A
A
A
Right
so,
like
probably
a
whole
tutorial
on
just
random
crap,
you
can
do
with
data
flows
and
ways
you
can
configure
them,
and
you
know,
like
here's,
how
you
write
a
yam
will
file
here's,
how
you
dump
it
to
a
JSON
file
and
Lola,
and
you
know:
here's
how
you
write
it
in
Python
and
just
just
a
bunch
of
examples
to
get
people
used
to
things.
Let's
see,
this
may
not
be
the
best.
A
C
A
A
D
A
A
A
A
I
mean
we
could
just
to
start
like
the
first,
the
first
obvious.
The
solution
of
this
is
like
no
feature
some
kind
of
flag.
That's
like
no
features
all
right
like
if
we
had
no
features
and
then
image
right
and
results
and
not
printing
any
feature
named
image
right,
and
so
basically
we
would
just
you
know,
do
record
or
we
would
we
would
make
it
remove
that
feature
data
and
then
guilt
that
that's
probably
like
the
easiest
way
to
do
this.
At
this
point,
I
don't
know
how
great
of
a
solution
that
is
going
forward.
A
A
A
D
A
A
A
A
A
A
So
I
think
that
I
was
looking
at
this
theater
take
sort
of
another
context,
but
it'd
be
nice
if
there
was
like
I
think
what
we
really
just
need
to
do
is
like
we
probably
need
to.
Let
me
just
write
it
in
here.
We
should
look
at
terminal
column.
A
A
A
A
C
A
Only
ask
you,
though,
so
only
ASCII,
no
utf-8,
because
that
just
always
problems
it
looks
great
cuz
problems,
laughing
I,
never
at
least
someone
can't
parse
it
okay.
You
know
they
should
be
person
to
Jason.
Oh
that's
beside
the
point.
All
right
so
is.
That
is
that
what
else
did
you
want
to
talk
about
here?
Shock
shock.
G
A
A
A
A
A
A
A
A
D
A
A
B
A
Classic
yeah,
those
are
the
best
priority
number
one.
Yeah
I
think
that
the
secret
on
the
web
hook
is
something
that
yeah.
That's
that
I
labeled
a
good
first
issue,
so
that's
sort
of
like
you've
got
more
experience
in
these
areas,
whereas
this
one,
hopefully
won't
take
somebody
who
doesn't
have
as
much
experience
that
long
so
I
think
it's
sort
of
like
somebody
else
can
hit
that
if
they
want-
and
if
you
hit
those
two-
that
the
best
course
of
action
yet.
A
C
B
B
A
A
A
A
B
A
A
B
A
So
you
you
return,
you
call
async
I,
go
create
task,
and
then
you
call
I
sing
ko
create
you
might
have
to
refactor
things
a
little
bit
so
that
it
runs
the
data
flow
and
obviously
you're
not
going
to
output.
Actually
I
mean
it
doesn't
really
matter.
If
you
you,
if
you
just
like,
throw
the
whole
thing
that
you
had
where
it
was
returning
some
web
response
into
a
different
function,
it
it
won't
matter
because
that
results
just
not
gonna.
A
So
you
don't
have
to
like
write
different
code.
You
could
just
write
the
same
code
and
or
well
you
may
you
may
yeah
just
just
just
leave
it
the
way
it
is
yeah
and
yeah
I
mean
you'll,
see
how
it
works,
but
basically
just
call
create
task
and
it'll
run
it
in
the
background,
and
that
way
we
can
respond
immediately
to
get
help
with
a
status
of
okay.
So.
B
A
C
A
C
A
A
A
A
Okay,
now
we
know
who's,
calling
it
all
right.
So,
okay,
the
thing
I
think:
okay,
let's
just
go
into
so
so
everything
most
almost
everything.
That's
a
plug
in
derives
from
entry
point
right,
so
we've
basically
taken
feature
and
made
it
not
a
plug
in
so
we're
getting
rid
of
all
this
abode
stuff
from
picture,
and
so
what
we
want
to
do
is
so
yeah.
We
want
to
we're
trying
to
get
rid
of
anytime.
A
A
A
Convert
value,
config,
set
load,
okay
map
type
class,
ok,
must
have
set
tap
class,
equals
future
dot
load-
oh
yeah
cuz
in
May
MKR,
so
that
player
is
going
to
look
at
this
thing
and
it's
gonna
say
so.
This
is
when
we
have
a
config.
So,
as
is
like
part
of
this
that
unify
config
stuff
we've
been
talking
about.
Well,
we
now
have
something.
A
That's
like
like
you
know,
my
config
and
it's
got
F
is
feature
it's
looking
at
and
D
is
s
is
M
is
model,
so
when
it
looks
at
these
model,
is
a
plug-in
and
features
a
plug-in
right
now,
but
we're
making
feature
in
that
for
them,
so
so
when
it
goes
and
tries
to
parse
command-line
arguments
into
big
structures
like
this,
it's
looking
at
the
classes
that
it
sees
in
the
type.
So
it's
like.
A
Okay,
if
the
field
type
is
the
class
so
like
feature
and
model
or
classes
it's
coming
in,
and
it's
saying:
okay:
well,
does
it
have
a
load
method
like
does
it
have
you
know?
Does
it
does
it
have
a
dot
load
right?
And
in
that
case,
it's
setting
the
type
equal
to
the
type
dot
load
right.
It's
basically
setting
the
type
equal
to
the
load
method,
because
the
load
methods
a
class
method-
that's
going
to
return
a
class
so
because
what
art
parse
does
is,
if
you
say.
A
Type
equals
int.
All
it's
really
doing
is
just
taking
every
time
like
every
everything
that
you
pass
for
integers
right,
so
1,
2,
3
4,
and
it's
just
calling
it's
taking
the
string
value
here
and
passing
it
as
the
only
argument
to
whatever
you
put
in
type
so
int
right,
so
2
we're
using
that
trickiness
to
just
call
pass
to
have
type
be
something
dot
load
because
then
the
end
result
is
we're
going
to
end
up
with
a
loaded
version
of
that
type
or
well.
A
The
loaded
version
is
basically
we
pass
a
string
value
of
the
represent,
like
we
pass
a
string
representation
of
the
class
that
we
want
to
load.
For
example,
model
like
we
might
pass,
psych
a
tell
man.
The
load
method
comes
into
entry
point
like
and
well.
This
is
not.
This
is
the
feature
load
method
right,
but
imagine
it's
the
well.
Let's
not
just
imagine,
let's
open
it
up,
diff
until
entry
point,
so
it
comes
in
here
and
it's
like
okay.
Well,
what
is
what
am
I
loading
right?
A
C
A
A
Here,
where
we
have
like
what
was
the
plug-in
name,
so
what
was
the
thing
passed
to
load
and
then
what's
the
config?
Well,
what's
the
config
for
it
and
so
that
stuff?
Basically,
when
a
class,
we
now
have
the
class
because
we
called
load.
So
we
have
the
instance
of
the
class.
Now
we
want
to
create
in.
A
We
want
to
create
an
instance.
Sorry,
we
have
the
class
itself
now
we
want
to
create
an
instance
of
the
class,
and
so
basically
it
the
class
itself
looks
the
class.
Has
this
method
called
dot
config?
And
it
looks
in
that
extra
config
thing,
which
is
this
giant
dictionary
that
has
all
the
configuration
parameters
parsh
from
the
command
line
here
and
it
grabs
out
all
the
values
that
it
needs
for
each
of
the
fields
within
its
config
so
like.
A
If
we
were
trying
to
configure
Mike
and
my
config,
it
would
look
for
F
and
it
would
look
for
m,
and
so
when
it
does
that,
then
it
does
the
same
process
all
the
way
down
on
each
of
these,
and
so
it
looks
for
their
plugin
in
their
config.
And
why
was
I
talking
about
this?
I'm
sorry,
basically
for
a
feature.
We
just
want
to
get
rid
of
dot
load
and
then
it
won't
do
this
right.
So
if
we
remove,
if
we
remove
the
load
method,
then
it
won't
go
in
this
block
here.
E
A
Test
merge
we're
also
trying
to
mock
the
load.
Okay,
also
trying
to
mock
the
load,
also
trying
to
mock
the
load.
Also
try
to
mock
the
load,
I
feel
like
I
made
these
older,
like
so
classroom
the
same
thing
so
hopefully
I
go
one
place.
We
have
to
get
rid
of
the
mock
a
call
or
is
that
best.
Seo
lie
okay
and,
let's
see.
A
A
A
A
All
right
there
we
go
so,
let's
see
what
we
did
to
get
here
all
right,
so
we
got
rid
of
load.
We
removed
the
references
to
okay.
Well
thanks
for
auto.
For
me,
this
is
just
that's
just
Auto,
for
this
is
just
Auto
for
a
minute.
We've
got
rid
of
the
mock
called
a
load,
and
then
this
got
auto
format
it
as
well,
and
we
got
rid
of
the
more
mocking
craft.
Oh.
A
A
Okay,
yeah,
sorry
that
was
kind
of
fast,
but
basically
right.
What
we're
doing
is
is
features
no
longer
a
plug-in
right
and
I
think
the
best
example
of
that
is
here
right.
So
in
test
CLI,
where
we
wanted
to
take
out.
Let's
just
do
a
terrible
yeah,
so
we
want
it.
We
had
to
remove
this
call
to
mock
right,
because
we
need
to
remove
the
feature
download
method
and
we
had
been
replacing
it
with
this.
This
fake
loader,
so.
A
F
A
A
A
Fit
end
up
with
a
valid
feature,
if
always
well,
with
that
parser
helpful,
helper,
stuff
and
number
is
required.
Not
a
list.
Okay,
well,
where's
that
oh
well
I
tried
to
print
it
and
it
go
mad
okay.
So
now
we
need
to
help.
We
probably
need
to
update
this
too.
So,
let's
just
do
percent
all
right,
but
hey
it
looks
like
it
did.
Parse
it
successfully.
So
because
now
I
mean
now
yeah.
What
we
did
was
we
just
switch
cuz
load.
A
It
basically
well
we'll
find
out
if
it
matters
I,
guess
like
people,
people
will
yell.
You
know
like
that.
That's
sort
of
down
the
road
right
like
because
people
will
probably
get
used
to
just
specifying
just
the
name,
just
the
name,
just
the
name
and
also
they'll-
have
a
data
type.
That's
a
float
and
they're
like.
Why
is
my
float
getting
rounded?
So
actually,
we
should
probably
know
what
float.
A
C
A
A
We've
replaced
feature,
we've
replaced
that
feature
with
feature,
but
we
still
need
to
do
some
more
work
here.
Well,
actually,
let's
go
through
and
change.
Every
time
before
we
merge
this.
Let's
change
every
time
type
it.
We
should
really
get
rid
of
DT
type
in
length
as
functions
and
make
them
just
properties.
So
we.
A
Just
just
nuke
these
methods
here
and
and
then
call
that
good
and
the
other
thing.
Oh,
the
last
thing
is
that
we
should
have.
We
should
do
something
here
to
say.
If
so,
if
we,
if
we
hit
this
from
that
likely,
the
the
D
type
is
going
to
be
a
string,
and
so
we're
going
to
need
to
do
a,
if
is
instance,
D
type
stir
and
then
do
to
convert
helper
there,
because
this,
like
you
know,
cuz
D
type,
will
get
expanded
here.
A
But
it's
nice
to
it's,
not
gonna,
be
it's
gonna,
be
a
string
still
if
it
comes
from
from
dict,
rather
than
just
from
the
name
getting
split.
So
that's
still
going
to
be
great,
so
we
can
just
do
this
so
I'll
post
this
diff
and
then
yeah.
If
you
can
go
through
and
remove
these
methods
and
let
the
comments
so
basically,
just
like
you
did
for
name
right,
I
just
want
to
have
them.
As
you
know,
D
type
and
length
like
that.
C
A
That'll
be
good
and
obviously
there's
going
to
be
all
the
modules
will
need
changing.
So,
like
mostly
it's
probably
just
going
to
be
the
the
HTTP
service.
It's
probably
gonna,
be
one.
The
tensorflow
metal
psychic
models
scratch
model.
There
should
be
minimal
references
to
things
in
there.
It
should
hopefully
just
be
changing.
Def
feature
to
feature
I
would
think
at
this
point.
So
but
but
that's
that's
the
rest
of
probably
what
needs
to
be
done.
Let's
just
get
grep
and
see
it
grow
def
feature.
A
A
C
A
A
A
A
A
Chrome
needs
to
get
to
start
reading
in
response,
so
you
know
HTTP
status,
okay,
new
line
and
then
echo
will
print
another
new
line.
And
then
you
have
two
new
lines,
which
means
the
HTTP
headers
are
done,
and
then
you
can
just
say
any
command
and
pipe
it
through
netcat,
listening
at
some
port,
that's
accessible
via
Web
browser
and
Al's
and
it'll
just
dump
the
file
to
you
so
that
that
was
a
fun.
That's
a
fun
thing
I
find
has
to
be
really
fun.
A
D
A
B
B
A
B
A
A
D
A
A
F
A
Good
that
grape
that
you
finally
got
that
we
got
that
example
going
through
that's
this
pine
cone
is
how
much
you
did
the
first
thing
and
then
that
other
guy
jumped
on
did
half
the
example.
I
did
have
the
example
after
that,
and
then
you
finally
got
it
merged.
So
that
was
great.
So
now
we
finally
have,
let's
see
if
that's
a
it's
should
be
live
by
now.
What
are
you,
what
are
you
up
to
next
year?
You
jumping
on
that
applaud
ml,
or
what's
that
amount,
not
the
dowel,
the
dowel
one
yeah.
F
A
A
All
right,
okay,
well
and
the
issue
was
basic:
okay,
so
yeah.
Let
me
bring
this
up,
so
we
can
view
it
while
we're
talking
about
this.
A
All
right
so
well,
we
ran
into
here
was,
as
you
guys
are
probably
and
this
the
point
is
like.
Let
let
us
know
if
you
ran
into
this
yourself,
but
when
you
specify
commands
for
a
model,
you
have
to
specify
all
the
same
arguments
for
each
invocation
rate
between
train
accuracy.
A
Because
we
do
hashing
on
on
some
of
the
some
of
the
arguments
right
to
make
it
so
that
people
don't
have
to
specify
a
new
directory
for
each
model.
It
just
ends
up
in
their
cache
right
and
so
I
think
that
this.
This
is
at
times
a
very
helpful
feature,
because
we
don't
have
to
tell
people
like.
Oh,
you
know,
specify
a
directory
for
this
model.
We
just
give
one
to
them
and
it
prevents
them.
A
B
A
A
So,
let's
see
this
is
going
to
be
interesting
because
some
of
our
test
cases
and
stuff-
just
like
you
know
some
of
our
example
usages
and
things
just
like-
don't
have
a
directory
and
I
guess.
Maybe
we
could
just
throw
one
in
temp
yeah.
Let's
just
throw
one
attempt
where
we
don't
have
one
like:
whenever
there's
an
example,
we
can
just
have
a
temp
of
directory.
A
A
A
All
right,
great
I'm,
glad
that
this
came
up,
because
this
is
this
has
been
I.
Think,
like
the
other
thing
that
I've
noticed,
that
is
a
waste
of
like
kid
has
wasted.
A
lot
of
my
time
is
that
we
only
train
on
features
that
we
find,
and
so
with
like.
This
is
another
thing
like
if
you
capitalize
a
feature
name
and
you
didn't
capitalize
a
feature
name
and
actually
in
the
feature
data,
then
it
says
no
records
to
train
on
that
one
I'm,
not
so
sure
about
how
we
should
handle.
A
What
do
you
guys
think
about
that?
One?
Is
that
something
like
I?
Don't
really
have
a
better
solution
for
that
this
one.
It
has
a
sort
of
an
obvious
solution,
but
does
anybody?
Is
everybody
run
into
that
other
one
as
well?
And
does
anybody
have
an
idea
of
how
to
solve
it?
Because
we
don't
have
any
ideas
we'll
have
to
leave
it,
but
it
would
be
sort
of
good
to
know
what
I
assume.
This
is
also
a
common
problem.
We.
D
A
F
E
A
That
would
probably
be
a
little
bit
quicker
than
we
don't
have
to
do
the
if
we
can
hook
into
the
other.
One
is
a
little
bit
more
complicated,
but
I
like
that
too
either
way,
you'll
see
right
away.
Okay,
so
let's
also
put
this
on
the
list
of
things
that
needs
to
happen,
because
these
two
things
are
really
freaking.
Annoying
and
I've
spent
like
a
ridiculous
amount
of
time,
debugging
things
that
are
just
my
styling
mistakes,
but
that's
like
goes
for
everything
so
yeah.
Okay,
so
let's
say:
let's
open
an
issue
for
also.
A
See
issue
40.2
explained
release.
This
is
gonna,
be
a
really
solid
release.
We're
gonna,
I,
think
I
think
this
is
there's
gonna,
be
one
that
finally
makes
things
very
usable
for
everyone
else.
There's
always
been
like
one
thing
or
another.
That's
like
hard
understand,
I.
Think
now
we
finally
have
like
what
are
all
the
things
that
we
have.
We
have
like.
A
A
A
A
A
A
A
Specified
blank
but
the
options
that
the
options
were
like
and
there's
gonna
be
some
more
modification
in
there,
because
that
so,
let's
see,
there's
I
mean
yo.
Whoever
does
this
will
see
that
there's
guy
their
second
few
tweaks
in
there
so
I'm
going
to
go
through
and
make
issues
are
the
ones
that
we
don't
have
issues
and
then.
A
A
A
A
B
A
A
A
So
I
think
maybe
we
could
just
add,
like
a
seed
parameter
that
functions
similarly
to
the
inputs,
parameter
great,
so
I'm
just
trying
to
think
about
what
we've
got
and
what
we
will
have
here,
because
I
like
that
idea
that
the
thing
is
that
if
we
do
pull
in
and
then
something
then
we
need
to
have
some
kind
of
like
we
we
need
to.
We
can't
be
sure,
like
how
each
operation
spec
works
right,
so
we
probably
have
to
have
each
output
operation
implement
some
kind
of
parser
to
say
like
okay.
A
A
A
A
A
A
Oh,
this
will
be
sweet
actually
with
this
you
can
literally
just
create
a
data
flow
because
you
can
add
to
the
seed
inputs.
Well,
I
guess
you
could
do
this
before
but,
like
you
know,
you
could
add
to
the
seed
inputs
for
get
output.
I
guess
you
could
do
this
before
by
like
I'm.
Just
thinking
like
you
can
do,
data
flow
create
and
pipe
it
straight
to
dataflow,
run
I.
B
A
A
Oh,
let's.
Let's
make
an
issue
to
track
about
that.
So
in
so
arguments
need
a
way
to
modify,
modify
input
flow
of
operation
when
doing
dataflow
dot
or
data
flow
create,
because
that
that
is
also
something
that
like
we
should
have
one.
You
know
when
you
create
a
data
flow,
you
should
be
like:
okay,
here's
the
operations,
here's
some
seed
inputs
and
here's
any
modifications
to
the
flow
and
then
that
way
to
end
up
with
a
data
flow
that
works
right
away.
So
that's
another
thing
that
we
should
do
at
some
point.
B
A
Yeah
I'm
gonna
take
a
look
at
that.
That's
something
that's
on
my
docket
of
things
to
do
soon.
Still
so
yeah,
that's
basically
the
next!
That's
the
next
thing
that
I'm
looking
at
I
have
to
so
just
to
give
you
guys
an
idea
of
what
you're
gonna
see
out
of
me.
Soon,
I
am
going
to
I'm
going
to
be
throwing
up
some
of
this
stuff
about
dependencies
into
should
I
I
might
go
and
finish
ashes
stuff
off.
A
We
should
I,
because
I
want
to
get
that
stuff
into
the
next
release
and
then
the
other
thing
I'm
going
to
do
is
I'm
going
to
do
the
unified
config
stuff
to
make
sure
that
we
know
what
we're
doing
there.
Is
that
that's
sort
of
my
action
items
for
the
next
release
and
then,
if
I
finish
those
than
some
other
things
but
yeah,
that's
what
I'm
looking
at
right
now.