►
From YouTube: 2020 09 14 Memory Team Meeting
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
All
right
welcome
to
this
monday's
version
of
memory
team
meeting.
Kick
it
off
we'll
go
through
social
styles.
I
shared
the
exercise
with
everybody
on
the
memory
team
channel.
I
don't
know
how
much
folks
want
to
chat
about
this,
but
posted
my
scores.
As
I
noted
below
I
bounced
between
analytical
and
amiable,
I
think
I
landed
on
analytical
a
little
bit
more,
but
most
time
I
was
within
about
point
one
of
being
dead
center.
A
B
I
had
like
the
score
so
they
were
assigned
to
like
a
certain
category.
So
I
don't
know
what
the
well
yeah
hang
on.
I
can.
I
can
paste
it
in
the
in
the
document
later
but
yeah
like
I
had.
I
had
a
similar
impression
as
I
think
it
was
alexi
that
maybe
it's
by
design,
but
it
was
very
specific
right,
yeah
the
question
so
so
I
guess
that
would
yeah
yeah.
D
C
Yeah
yeah,
I
mean
it
was
fun
to
take
this.
This
kind
of
fell
in
the
table
and
I
think
entertain
the
make
sense.
I
mean
it
fits
what
I
expected,
but
yeah
as
mattias
said
also
super
prompt
interpretation,
and
it
also
depends
on
the
context.
So
in
some
situation
you
could
be
analytical
and
you
should
be
analytical
and
some
you
should
be
so
on
and
I'm
pretty
much
sure
it's
more
like
a
mix,
but
it
could
raise
some
interesting
questions.
A
The
the
two
questions
that
probably
changed,
my
score,
mostly
from
amiable
to
analytical,
were
controlling
and
warm
versus
cool.
So
when
I
first
took
it,
I
took
the
controlling
question.
As
as
a
manager,
you
control
your
team,
your
micro,
managee
right
and
I
was
like
no.
B
A
Know
way
on
one
side,
but
when
I
actually
watched
the
video
about
some
of
the
questions
it
was
more
about
self-control.
Do
you
control
your
emotions,
or
are
you
very
expressive
in
your
emotions,
and
then
I
was
on
the
other
end
of
the
spectrum
and
then
yeah.
The
other
question
was
warm
or
cool
and
wow.
That's
open
for
all
kinds
of
interpretations.
C
No,
no,
I
just
noticed
it
seems
like
try
to
wait
and
and
like
soften
this,
like
interpretation
stuff
by
asking
repeated
questions,
as
nicola
mentioned,
no
sociological
trick
to
like
verify
and
maybe
correct,
but
I
don't
know.
I
also
noticed
that
some
question
could
like
drop
you
from
one
spectrum.
Another.
B
B
B
Like
one
thing,
I
also
wanted
to
add
is
one
thing
I
did
not
like
about
how
the
test
was
set
up
was
that
at
first
before
I
took
the
test,
I
read
through
the
whole
like
guide
and
the
social
site.
B
Styles
definition
on
our
in
our
hand,
the
handbook
section-
and
it
was
very
clearly
explained
like
what
the
quadrants
are,
and
I
think
like
you,
I
like
immediately
noticed
so
like
I
thought
I
had
some
traits
from
all
of
these,
but
I
immediately
noticed
oh
I'm
very
firmly
like
in
this
quadrant
and
not
the
other
quadrant,
and
I
think
this
kind
of
sets
yourself
up
kind
of
biases
yourself
already
before
you
even
take
the
test
to
like
give
certain
answers
that
I'm
almost
sure
of
so
I
don't
know
how
to
how
to
better
do
this
because,
like
on
the
other
hand,
you
need
kind
of
an
explanation
to
even
know
what
these
vague
questions
mean.
A
Yeah
I
like
the
the
sam
beckham
quiz
that
he
created
it's
pretty
nice,
because
you
can't
see
your
score
influencing
overall,
like
if
you
take
the
pdf,
you
can
see
the
total
kind
of
tally
building
up,
so
you
can
start
gaming.
The
system,
if
you
don't
want
to
land
in
one
or
the
other,
so
this
the
first
couple
times
I
took
it
was
on
the
pdf.
The
last
couple
times
was
on
sam
beckham's,
new
ui
that
he
created.
A
So
I
found
that
very
useful
so
and
I
added
some
more
resources
below
there's
a
handbook
entry
on
social
styles.
We've
got
a
lot
of
good
information,
there's
a
slide
deck
that
was
presented
to
the
managers
as
we
were
going
through
this
and
there's
a
there's.
A
lot
of
interesting
takes
on
there.
There's
one
slide
in
there
as
well.
That
states.
A
So
a
lot
of
good
information.
If
you
want
to
follow
up
and
read
more
about
it
and
then
overall,
if
you
found
this
type
of
quiz,
this
type
of
analysis
interesting.
I
posted
some
more
down
here
at
the
bottom
that
are
different
in
different
ways.
So,
if
you
like
reading
up
on
these
things,
there's
some
other
books
and
websites
that
I've
found
interesting
in
the
past.
So
any
other
topics
on
this.
A
All
right,
let's
jump
over
to
the
board,
so
I
wanted
to
run
through
the
board
real
quick.
Since
the
milestone
is
ending
this
week.
I
think
I
know
the
status
for
most,
but
let's
run
through
them
real
quick.
I
will
follow
up
with
chin
you
on
this.
One.
Matthias
you've
got
export
number
of
action,
cable
connections.
You
think
this
one's
gonna
close
out
this
week.
B
Yes,
that
is
done,
okay,
that
is
done.
We
actually,
we
actually
did
a
bit
more
than
this,
so
we
have
a
whole
bunch
of
action.
Cable
thread
pool
related,
metrics
yeah,
so
that
got
merged
this
morning.
B
So
the
thing
was
so
what
I
did
so
it
was
split
up
into
two,
mrs
and
the
first,
mr
I
actually
tested
in
omnibus.
We
can't
test
it
in
prod
because
we're
not
running
action,
cable
in
fraud,
so
I
would
close
it
out.
E
I
don't
remember
exactly
because,
like
when
we
were
working
on
a
bunch
of
these
documentations,
I
like
we
created
the
follow-ups
basically
for
for
everyone,
is
off
topic.
A
A
B
They
probably
shouldn't
have
my
face
on
it
like
this
is
really
about
so
so
yeah.
It's
like
the
feature
flag,
rollout
issue
for
the
dynamic
resizer
in
general.
It's
just
it
happens
in
so
many
stages.
I
mean
we
already
have
one
version
in
pride
behind
a
feature
flag
that
we
did
run
experiments
if
anything
was
alexei.
Working
on
this
right
now,
so
maybe
be
good
to
really
sign
that
yeah,
but
also
we're
unlikely
to
ship.
What's
currently
in
production.
C
B
Flag
issue
that
we
have.
B
D
Okay,
nice.
I
already
closed
the
mr,
because
it
is
unlikely
that
we
will
merge
this.
I
even
created
one
follow-up
issue
for
this:
to
introduce
this
new
duplication
strategy
first
sidekick,
but
unless
we
like
drop
out
this
dynamic
scaling,
we
will
probably
not
come
back
to
this
static
approach.
It's.
B
D
E
So
so
alexi
like
from
the
testing
on
100
avatars,
we
see
like
a
noticeable
difference
on
the
dynamics
carrying
being
on.
C
No,
I
I
didn't,
I
definitely
didn't
see
any
like
specifics
in
workhorse,
health
and
so
on,
because
I
think
the
request
rate
was
too
small
and
we
didn't
see
any
the
only
thing
that
we
noticed
is
on
such
small
image
selection.
We
have
this
client
what
materials
mentioned
kind
of
quadruple
like
time
growth
when
it
comes
to
the
image
size,
but
it's
not
confirmed
yet
because
we
have
a
small
selection
of
images,
so
I'd
prefer
to
run
it
on
a
like
full
gitlab
and
place
collection.
C
B
Yeah
so
yeah,
I
look
at
it
a
bit
so
from
the
benchmarks.
I
ran
locally
on
my
machine.
It
was
about
the
same
but
like,
as
we
already
saw
like
again
alexa.
I
don't
know
how
big
the
sample
sizes
were
in
fraud,
but
from
the
measurements.
D
B
So
I
don't
know
how
representative
this
is
of
all
of
gitlab,
because
we're
already
only
testing
a
small
slice
as
well
right,
but
I
mean
generally
look
good
like
outside
of
the
outliers
that
were
that
took
a
lot
of
time,
because
the
images
were
larger,
like
400,
pixels
or
so
and
which
also
looked
like
they
might
have
been
outliers
because,
like
I
said,
all
of
the
histogram
buckets
had
the
same
size,
it
looked
a
bit
suspicious
and
there
was
no
count
attached.
So
maybe
that
was
just
a
single.
It.
C
And
I
believe
400
was
also
my
like
queries
when
I
clicked
on
the
avatar,
because
nobody
it's
not
used
in
ui,
it's
upscaling,
it's
only
a
case
of
upscaling
queer
learning.
So
maybe
we
should
even
like
deny
it
this
kind
of
request,
but
we
don't
want
to
upscale.
To
be
honest,.
D
D
E
So
basically,
like
we
saw
some
some
increasing
timings
on
like
on
the.
If
the
resulting
image
size
was
about
400
pixels
right.
C
Carrying
I
would
not
even
I
would
not
even
take
400
is
a
good
example,
because
it's
upscale
in
our
case
we
store
all
images,
200
per
200,
and
then
we
decide
into
upscale,
so
it
probably
uses
a
different
algorithm
even
in
gm.
As
far
as
I
know,
so,
it's
not
someone
who
uses
strong
source
problem,
so
it
has.
C
B
Yeah,
that's
probably
like
something
we
should
did
we
erase
that
with.
I
think
we
did
right.
We
raised
this
with
tim,
because
this
is
not
even
specific
to
the
image
scaling.
Epic
right,
it's
just
weird
right
now
that
we
downscale
everything
on
uploads
to
200
pixels,
and
then
we
upscale
it
again
when
you
go
to
the
verified
page,
that
doesn't
really
make
sense.
B
That's
that's
not
going
to
be
great
for
high
def
high
res.
C
F
E
B
So
I
mean
like
that's:
I
mean
we
already
have
something
in
production:
it
uses
graphic
magic,
so
that
was
not
the
problem.
I
think
the
biggest
problem
right
now
is
that
the
and
the
currently
open
mr
and
they
use
the
same.
The
sandbox
approach
with
second
is
blocked
by
all
these
build
environments
that
we
need
to
compile
that
binary.
B
So
we
need
to
yeah.
So
let's
talk
about
this
because
that's
I
had
a
couple
points
here
anyway,.
E
B
E
I
I
I
want
to
like
start
and
verify
one
thing.
First,
before
we
get
into
building
it's
like
the
second
approach
and
like
our
reference
eoc,
like
security
validated
by
the
security
team
and
are
approved
at
this
point.
E
We
have
a
binary,
that's
like
static
linked
and
it
appears
to
be
working.
Is
it
like?
Did
we
get
like
that?
Some
additional
eyes
from
the
security
team
to
say
this
is
great.
You
should
continue
and
finish
that
work
or
yeah.
C
E
So
like
like,
we
have
pretty
complete
solution
at
this
moment
right
because,
like
from
previous
conversation,
I
understand
that
we
talk
about
specific
problems,
but
right
now
we
are
really
on
the
like
complete
solution
where
there
are
still
some
concerns
to
resolve.
E
So
I
think
maybe
it's
like
the
right
moment
to
get
like
the
let's
say:
semi-formal
approved
from
the
security
team,
that
this
is
what
they
had
in
their
mind,
and
they
are
fine
with
that
solution
and
we
don't
need
to
further
involve
them
in
the
all
the
follow-up
integration
work
because,
like
there
is
a
few
aspects
that
we
did
as
part
of
the
trade
which
is
separate
by
nare,
which
is
segcomp,
which
is
a
compiling
graphics,
magic
strip
version.
E
B
Yeah
we
can
do
that
so
just
to
be
clear.
They
are
aware,
like
two
security
team
members,
have
been
reviewed
this
mr
multiple
times,
so
there
was
a
lot
of
back
and
forth
with
them
on
like
how
we
can
cut
down
on
the
system
calls
and
they
made
a
specific
list
of
syscalls
that
they
consider
problematic,
not
all
of
which
we
were
able
to
get
rid
of.
B
But
I
think
the
last
thing
I
remember
I
think
was
anthony
saying-
was
that
these
mem
copy
system
calls
are
not
ideal,
but
if
we
yeah,
but
if
it
because
it's
the
sandbox
binary,
it
shouldn't
be
a
big
problem.
B
But
we
can
ask
for
like
a
formal
approval
again,
then
there's
this
other
issue
of
the
strict
checks
that
we
need
to
redo
and
that's
something
that's
nicola
currently
works
on,
because
with
the
gm
binary,
we
got
that
for
free
with,
like
special
prefix
syntax,
that
you
can
use
when
you
pass
the
file
name.
So
we
have
to
build
this
manually
now
through
the
api
integration,
so
that
is,
that
is
a
separate,
mr
though
I
think
we
can
just
like
we
shipped
the
other.
B
I
would
just
like
to
paralyze
this
stuff,
because
it's
already
taking
much
longer
than
we
thought
it
would
be
so
that
we
can
look
at
all
these
different
aspects
in
parallel
and
not
wait
for
like
one
team
to
come
back
with
an
okay
before
we
even
start
the
remaining
work
because
for
the
whole
build
environment
stuff,
it
is
a
lot
of
work
still.
So
I
would
prefer
to
work
on
this
now.
B
Right,
and
on
top
of
that,
there
is
another,
mr
currently
in
maintainer
review,
which
extends
the
current
solution.
That's
in
production,
with
a
number
of
these
things
like
this
magic
bite
check,
for
instance,
because
they
had
already
been
open.
There's
a
couple
other
improvements
in
there
like
passing
down
the
content
type
from
rails
into
workhorse,
which
is
something
that
we
will
have
to
look
at
anyway,
and
a
bunch
of
unit
tests
added
and
stuff.
B
B
But
yeah
like
so
okay
I
mean
I
can,
should
we
do
that
in
the
initial
security
review
issue
like
to
ask
if
we
can
get
like
a
thumbs
up
thumbs
down
on
the
current
approach?
It's
just
the
mr
is
still
like
very
much
a
work
in
progress.
E
What
can
change
is
like
exactly
where
we
get
dependencies
from,
but
it
seems
to
me
that
we're
gonna
stick
with
the
with
this
baseline,
which
is
like
static,
compile,
segm
and
the
graphics
magic
library
implemented
in
cgo
right.
So
what
can
really
change
is
like
maybe
some
specific
functions
being
heroes
or
like
dependencies
where
we
take
them
from,
but
it's
not
going
to
change
the
security
aspect.
B
Except
for
the
we're,
not
I'm
not
100,
confident
that
we
have
really
honed
in
on
the
final
list
of
system
calls
we
need
to
allow,
because,
with
this
super
restricted
set
that
I
last
was
experimenting
with
when
I
tested
this
at
scale
and
like
nikolas
saw
something
similar.
We
saw
occasional
crashes
when
the
application
was
trying
to
allocate
memory,
and
this
might
may
or
may
not
be
related
to
us,
preventing
this
map
call
from
happening.
B
C
B
B
That
said,
this
whole
like
build
environment
stuff,
since
that,
so
this
all
fell
out
of
this
issue,
just
that
one
from
the
top
here
and
sure.
Oh
sorry,
actually
no
there's
like
another
one
that
should
have
my
face
on
which
has
the,
which
is
the
the
sandboxing
issue.
B
Oh
yeah,
this
one
down
there
image
scaling
minimize
security
and
pick
up
scalar.
I
I
would
really
suggest
at
this
point-
and
I
can
at
least
create
templates
for
this,
but
some
help
would
be
appreciated.
They
will
create
stories
for
all
of
these
different
things
that
need
to
be
fixed
for
this,
mr
to
be
mergeable,
because
we
can't
even
we
can't
even
ship
this
by
feature
flight
currently,
because
if
we
were
to
merge
it
now,
everything
would
fail
like
the
gdk
would
fail.
B
Gck
would
fail,
except
for
workhorse
cci
like
all
the
builds
would
fail
so
and
I'm
currently
looking
into
omnibus
support,
but
it
kind
of
ballooned
a
little
bit,
because
we
had
no
idea
up
front
like
like
how
much
work
this
would
be
so.
But
I
think
now.
E
Can
we
can
we
use
your
branch
and
basically
use
your
branch
to
be
all
possible
rights
that
we
have
to
support
to
actually
know
what?
If
what
is
failing
and
how
we
could
fix
that
because,
like
you
mentioned
on
omnibus-
and
I
know
that
like
for
omnibus,
we
can
have
our
own
branch
of
the
omnibus
that
uses
different
branches
of
the
guitar
portholes.
We
can
do
the
same
for
the
cng
gck,
as
you
mentioned,
like
it's.
Probably
the
easiest
to
support
and
probably
like
that.
E
The
hardest
like
to
support
is
a
gdk,
but
maybe
we
could
get
someone
from
the
gdk
team
to
help
out
how
to
bundle
this
defensively.
B
I've
spent
like
two
days
on
this
already,
and
it's
like
not
really
related
anymore
to
this
particular
issue.
It
was
just
kind
of
fell
out
of
it,
so
I
don't
like.
C
C
Let's
see,
we
could
think
we
could
think
after
this
meeting
and
create
this
items,
if
you
want
sure.
A
C
B
E
B
C
B
So
no,
no,
I
I
think
this
is
so
useful,
so
I
mean
the
first
half
of
this,
I
would
say,
is
done,
which
was
to
make
sure
that
environment.
So
I
think
what
this
is
useful
for.
Oh
you
mean
like
to
actually
okay,
so
you
mean
to
grab
like
things
like
performance
numbers
and
resource
consumption.
C
C
C
Idea
was
to
like
find
performance
indicators
and
so
on,
like
from
this
test
and
see
if
there
will
be
any
issues
with
like
production
like
traffic,
but
it
seems
like
we
are
replicating
this
with
the
current
gm.
What
is
merged
so
I
don't
know.
Maybe
we
could
just
change
the
priority
for
now
and
not
concentrate
on
it
because
yeah
there.
C
B
B
Right
now
is
at
least
unless
you,
unless
you
have
access
to
a
production
box,
you
can't
actually
see
what's
going
on
on
that
machine
right
in
terms
of
the
process
he's
running
and
stuff,
and
I
guess
like
one
idea
had
there
was,
if
you
run
it
something
like
a
silk
test,
you
could
see,
I
would
creating
zombie
processes
that
never
terminate
you
know
because
gm
gets
stuck
or
something
like
that.
So
so
this
is
something
that
we
would
have
been
able
to
do
with
that
environment,
but
yeah
like
I
guess.
B
That's
mostly,
you
would
do
that
before
you
deploy
production,
so
maybe
we'll
pass
that
stage
right
now.
B
C
C
B
C
I
think
in
general
it's
a
great
design
decision.
I
mean
for
such
you
know
for
such
architectural
feature
to
have
this
kind
of
sandbox
with
production
like
traffic,
pointed
to
it
and
replay
it
and
see
any
potential
issues,
but
yeah
these.
D
B
The
test
environment
to
have
have
like
sufficient
data
that
it's
meaningful
to
test
right.
Because
if
you.
D
B
D
B
So
so
it
wasn't
even
testable.
Yeah.
E
D
A
A
Okay,
binary
match.
B
Yeah,
so
this
one
is
like
interesting,
because
so
there
is
an
openmr
that
is
waiting
for
our
maintainer.
Well,
that
is
in
maintainer
review.
There's
some
back
and
forth
there
by
the
way,
lexi
I'm
looking
into
this
right
now.
I
need
to
change
the
way
we
log
again
in
as
well,
because
we
weren't
using
the
structured
logging
everywhere
yeah.
B
It
doesn't
depend
on
second,
but
not
like
how
we
log
in
the
app
and
stuff.
So
this
is
all
also
because
we're
still
a
bit
unfamiliar
with
workhorse
and
stuff
just
takes
time
and
there's
a
lot
of
back
and
forth
between
maintainers
and
then
nikola
is
working
on
an
mr
against
my
new,
mr
that
address
that
concern
again
in
that
variant
of
the.
B
A
B
E
C
B
B
D
Right,
okay,
so
for
the
last
one
we
created
two,
mrs,
like
two
pocs
to
check
those
golang
libraries
instead
of
graphic
magic.
We
run
the
performance
tests
against
them
and
at
least
the
mini
minimum
required
system
calls
for
both
of
them,
and
I
I
posted
the
results
in
in
the
document
for
this
meeting,
so
you
guys
can
check.
So
I
guess
we
can
close
it
since
the
pocs
are
completed.
I
already
closed
the
mrs.
D
C
D
B
Because,
like
all
things
being
equal,
all
other
things
being
equal
which
they
seem
to
be
because
this
is
not
faster,
it
does
not
appear
to
be
performing
fewer
syscalls,
so
it's
also
definitely
requires
to
be
second,
so
all
other
things
being
equal.
I
I
personally
see
no
reason
to
use
any
of
these
libraries
because
they
they
look
like
single
person
like
site
projects
and
then
I'd
rather
go
with
gm.
B
I
know
it's
ugly,
but
it's
you
know
it's
been
around
for
like
20
years
or
something
so
have
a
bit
more
confidence
in
the
longevity
of
that.
A
Okay,
do
you
want
to
close
that
out
later
on
nicola
yeah.
D
A
A
Are
there
any
release
post
items?
We
should
add
to
this
release
post
for.
A
Mind
everything
so
I
will
add
some
issues
for
the
next
milestone.
We've
been
talking
about
lessons
learned
and
I
think
everybody
I've
talked
to
said.
We've
learned
some
great
lessons
from
image
scaling
how
we
should
actually
document
them
as
part
of
our
okr
and
something
we've
talked
about
for
a
while.
A
So
I
will
add
those
issues
for
13.5
and
I
might
even
suggest
folks
for
different
areas
within
those
issues.
So
nothing
else
there,
some
telemetry
recovered
future
flags.
Anything
else
you
want
to
add.
B
B
To
say,
like
I
think
we
should
you
know,
it
was
really
great
to
see
like
coaching
pulled
that
off,
because
I
mean
he
worked
on
this
all
by
himself.
This
milestone,
I
think,
was
really
cool,
so
he
did
the
console
integration.
So
I'm
really
excited
like
to
ship
this
and
see
see,
what's
going
to
happen
because
you
know
we
should,
we
should
now
be
able
to
fall
back
to
console
if
it
if
it
is,
if
it
exists,
we
have
a
bunch
of
extra
cool
stuff
in
there
like.
B
We
now
perform
readiness
checks
against
prometheus
to
to
make
sure
that
the
requests
will
actually
have
a
chance
of
succeeding
like
when
we
before
we
query
it,
and
he
also
created
fixes
to
deal
with
what
we
think
might
be
yeah.
This
whole
ssl
related
issue
like
if
you
put
prometheus
behind
ssl
proxy,
that
currently
we
just
fail
opening
a
connection
so
that
that's
still
in
review
but
like
yeah,
we
should
be
able
to
ship
it
in
34.,
so
there
should
be
some
cool
fixes
in
there.