►
From YouTube: TE Academy: Gitcoin Model Research Group #1
Description
In this first session of the Token Engineering Academy's Gitcoin Model Research Group, TE Academy head Angela Kreitenweis and data scientist Marc Minnee overview the cadCAD repo.
Learn more about the Token Engineering Academy here:
https://tokenengineeringcommunity.github.io/website/docs/academy-welcome/
Join @gitcoin, @block_science, @tokengineering & QV/QF academics and enthusiasts to discuss the collusion hunting cadCAD model of the Gitcoin Grants ecosystem.
More information about this research initiative:
https://www.notion.so/blockscience/Gitcoin-Modelling-Co-Lab-TL-DR-Public-4e115525dbc0430289bb78d60ad79a04
Read more about Block Science: https://block.science/
Learn more about Token Engineering: https://tokenengineeringcommunity.github.io/website/
More about the Token Engineering Commons: https://medium.com/token-engineering-commons/kicking-off-the-token-engineering-commons-be6a253cba81
cadCAD is an open-source Python package that assists in the processes of designing, testing and validating complex systems through simulation.
Discover more at https://cadcad.org/
Join the cadCAD Community Discord: https://discord.gg/hs4dAN9
Telegram channel: t.me/cadCAD_org
Twitter: https://twitter.com/cadcad_org
Get up to speed on our forum: https://community.cadcad.org/
A
A
In
any
case,
thank
you
so
much,
and
I
I
guess
this
was
also
for
for
in
case
you
you
haven't
used
exci
in
case
you
haven't,
transferred,
die
or
houston
exchange.
Perhaps
this
was
also
a
good
crash
course
of
the
the
great
things
we
have
in
the
ecosystem
and
also
the
challenges
we
are
meeting.
I
talked
with
russell
today
they
launched
a
new
contract
and
you
can
imagine
all
these
transactions
sum
up
to
significant
amounts
once
you
want
to
make
any
changes
in
your
product,
not
to
mention
tiny
transactions.
A
C
A
A
A
So
note
your
questions
and
let's
dig
into
that,
so
that
we
all
have
a
good
understanding
of
the
kepler
model
and
the
data
available
and
then
second
after
well,
half
an
hour
45
minutes,
we'll
start
with
a
workshop,
and
this
is
about
discussing
potential
research
questions
we'll
try
to
nail
the
most
promising
one
and
form
teams.
If
you
wish,
as
mentioned,
you
can
work
individually,
you
can
work
in
teams.
You
can,
of
course,
work
time-wise
supporting
this
team
and
then
go
back
to
your
research
question.
A
A
A
E
I
prepared
some
a
bit
of
a
structure
for
myself
to
have
it
on
a
notion
page.
So
basically,
what
I
would
like
to
do
is
getting
started.
How
do
you
get
started
with
cat
cat?
Because
this
is
about
cat
model?
Of
course
the
git
coin
project
has
provided
us
or
maybe
the
token
engineering
project
I
don't
know
or
danilo
has
provided
us
with
a
github
thing.
E
So
if
you
want
to
get
started
just
begin
by
forking
that
git
coin
cat
cat
model,
because
we
need
to
get
our
hands
dirty
anyway,
so
might
as
well
fork
it,
and
maybe
you
have
something
good
out
of
it
and
then
you
can
do
a
pull
requests
on
danilo's
repo,
and
maybe
you
will
succeed
so
basically,
this
is
some
nice
perk.
You
can.
You
can
earn
after
this
course.
Maybe
so
basically
fork
the
get
going
cadcad
model
and
build
some
branches
on
it
and
get
on
with
it.
E
So
the
thing
I
noticed
is
that,
first
of
all,
you
don't
have
any
data
to
run
your
model
on,
so
I
you
need
to
get
the
run.
Simulation.Pi
file
started,
and
I
also
noticed
I
did
this.
This
afternoon,
if
you
and
it's
basically,
it's
really
written
very
well
in
the
in
the
in
the
files
themselves.
So
it's
a
well
documented
repo.
E
You
can
run
a
simulation
with
the
a
number
of
time
steps
all
that
took
me
a
really
long
time,
I'm
only
on
a
cpu,
but
I
aborted
that-
and
I
tried
it
with
6
000
time
steps,
basically
the
ones
that
was
shown
in
the
example
of
the
repo-
and
that
took
me
about
a
couple
of
minutes.
Five
or
something
on
my
next
book
so
be
sure
to
be
or
be
wary
of
the
of
the
number
of
times
that
you
want
to
include
in
your
simulation.
E
Basically,
what
this
is
doing
is
getting
the
raw
data,
the
contribution
data
of
the
of
the
round-
seven,
I
guess,
get
coin
grants
into
your
system
and
then
the
fun
begins
you
can.
Then
you
can
open
your
notebooks
and
first
start
with
the
dynamic
network
notebook
and
later
on
with
the
graph
community's
notebook
yeah
before
we
take
a
look
on
that
on
that
thing,.
E
E
E
The
next
important
file,
I
would
say,
is
the
partial
state
update
block.
This
is
basically
the
dynamics
of
the
simulation
here.
You
state
what
you
want
to
achieve
during
the
simulation
and
in
what
order
this
would
be,
and,
of
course,
the
how
you
do
that-
or
this
is
basically
in
the
partial
state
of
that
block.
You
state
what
you
are
doing
and
in
the
functions
that
are
mentioned.
There
is,
of
course,
the
implementation
of
what
you're
doing
the.
How
and
this
is
basically
put
in
the
system.file.
E
E
E
If
you
understand
a
bit
about
cadcad,
it's
all
about
policies
and
mechanics
and
policies
are
basically
what
you
want
to
achieve
and
the
mechanics
are.
Basically,
your
state
update
functions.
So
in
the
policy
you
determine
what
you
want
to
achieve
in
the
state
update.
You
actually
execute
that
that
part
and
update
the
state
accordingly,
so
p-type
functional
policies.
S-Type
function
are
the
real,
the
actual
state
of
that
function
in
order
to
get
your
simulation
moving
forward
and
updating
the
state.
E
Well,
this
is
an
example,
of
course,
of
the
parcel
stat
update
block.
I
think
it's
from
the
dynamic
network
thing
and
what
you
can
see
here
is
well,
it's
neatly,
labeled
append
new
edges
to
the
network.
So
basically,
this
is
what
you
want
to
achieve,
and
this
is
what
kind
of
state
variable
is
going
to
be
updated
after
you
execute
this
policy
and
it's
quite
clear:
you
have
to
append
contribution
the
other
things
are
more.
Like
a
what
I
mean,
I
would
say
a
statistics
like
a
simulation
around
here.
E
You
do
the
actual
quadratic
match
and
that's
being
split
into
two
actually
two
blocks,
one:
the
yeah.
What
I
would
say,
the
full
monty
of
quadratic
funding
together
with
that
pairwise
comparison
and
something
an
algorithm.
I
I
try
to
get
a
grasp
on,
but
I
I
find
it
quite
difficult
difficult
to
to
understand,
but
there
are
lots
of
knowledgeable
types
here
in
the
call
that
can
educate
me
on
that
stuff
and
you
have
a
blog,
simple,
coder
function
that
is
without
a
pairwise.
E
E
What
you
do
in,
for
instance,
here
in
the
new
contribution
policy.
What
you
are
saying
is:
okay,
what
time
step
am
I
in,
and
I
want
to
add
a
new
contribution.
I
get
it
from
the
contribution
sequence
parameter
in
that
time,
step
and
I
add
it
to
a
new
contribution
variable
this
one
is
fed
into
the
state
update
function
as
append
contribution.
E
You
have
to
copy
you
deep
copy,
the
previous
state.
I
had
some
difficulties
when,
when
you
have
large
objects
hanging
around
basically,
the
cat
cat
model
is
about
more
like
a
functional
programming
style
style,
I
presume,
but
then
again
I
know
the
python
is
style.
So
this
is
just
my
imagination.
Maybe-
and
here
you
attend,
actually
you
append
the
new
contribution
from
the
policy
input
you
get
here
into
the
state
here
so
here.
Basically,
this
is
the
state
of
that
function
for
the
contributions
all
right.
E
E
E
You
can
see
here
the
the
files
are
mentioned,
so
it's
really
worthwhile
to
take
to
to
really
read
the
notebook
beforehand
for
before
you
go
start
hacking
around.
So
this
is
these
are
all
the
cells
I
well.
I
would
like
to
have
a
short
overview
of
what
is
actually
in
the
after
the
simulation,
because
I
already
ran
the
simulation.
This
is
something
which
might
be
confusing.
E
E
E
Something
is
not
here,
you
can
see
you
have
an
overview
of
the
quadratic
total
funding
during
those
time
steps,
and
you
can
see
it's
growing
mostly
linearly
with
time
steps.
So,
basically,
this
is
the
funding
that
is
built
up
during
the
during
the
events
of
the
round,
so
actually
what
I
would
like
yeah
this
is
taking
maybe
too
much
time
but
just
play
around
with
it
add
some
cells.
E
E
E
C
E
E
E
C
F
F
E
G
Jay
just
shoot
me
a
message
anytime,
it
will
work.
I
mean
there's
a
lot
of
many
ways
to
skin
a
cat
as
they
say
so
it
you
will
get
it
working,
but
you're
gonna
have
to
play
around
with
it
and
just
don't
be
afraid
to
reach
out
to
really
anyone
on
the
call
here
and-
and
we
can
drop
into
these
voice
chats
anytime
as
well.
If
you
just
need
to
screen
share
something.
F
E
H
H
A
C
B
A
B
A
B
C
B
C
C
H
C
B
B
I
B
B
C
A
Oh
sorry,
and
on
a
side
note,
bitcoin
rants
has
two
elements
in
the
interface
that
might
might
cause
a
bias
towards
a
similar
behavior.
A
is
the
card
checkout
and
the
option
to
share
your
card
once
you
have
contributed,
and
since
it's
a
lot
of
clicks,
just
imagine
you
have
more
than
one
grant
you'd
like
to
support.
A
If
you
do
it,
you
start
from
scratch.
You
have
a
lot
of
clicks
and
then
collect
your
card
and
then
check
out.
While,
if
you
build
on
a
card,
that's
already
there
from
your
friend
somebody
you
like
or
where
you
think.
Okay,
this
person
really
has
expertise.
I'd
like
to
donate
to
the
same
grants,
then
you
would
just
use
this
link
and
the
second
is
collections
that
are
also
collections.
We
had,
for
example,
the
token
engineering
collection
where
we
presented
token
engineering
grants
worth
supporting.
B
B
C
B
C
C
E
The
pull
request,
and
basically
that's
danilo,
he's
in
charge
of
selecting
the
best
propositions,
maybe
well,
no,
there's
not
no
there's
not
a
real
real.
I
think,
if
it's
good
enough
or
not,
it
has
to
add
something
to
the
current
model,
something
that
that
hasn't
been
tried
before
or
maybe
even
a
bug
that
you
discovered
could
also
be
a
pool.
You
can
also
open
a
pool
request
for
that.
It's
just
you
know
the
community
way
of
making
something
better.
I
would.
C
B
B
A
C
A
A
H
C
D
Presentation
right,
okay,
perfect!
So,
like
I
basically
just
made
a
short
presentation
on
the
on
the
bitcoin
grinds
data.
So
I
I
did
it.
I
split
it
into
two
parts.
The
the
first
thing
is
overview
of
the
of
the
round,
seven
and
round
eight,
and
that's
going
to
repeat
some
of
the
things
that
that
basically
kevin
presented
already.
D
D
So,
basically
like
every
round
kind
of
like
increase
the
the
the
the
contribution
amount
and
also
mostly
it's
also
the
matching
pool-
and
here
you
can
see
basically
that
that
it
increased
from
round
six.
It
increased
150
percent
and
then
also
the
there
were
a
lot
of
new
projects
and
also
the
new
category
structure.
D
D
So
when
it
comes
to
the
product
topic,
this
is
like
one,
for
example,
that
in
the
round
seven,
the
increase
was
pretty
driven
by
d5
projects
which
basically
were
were
yeah
building
their
reputation.
Somehow,
because,
because
in
in
some
ways
they
they
yeah,
they
might
not
have
been
totally
accepted
by
the
whole
hearing,
community
and
so
on,
and
also
there
was
another
interesting
thing,
which
is
the
collection
feature,
which
is
something
that
that
was
mentioned
before,
and
I
think
in
some
way
also
basically
impact
the
data.
H
D
D
Increased
and
then
the
insight
this
is
mostly
again
product
stuff,
so
first
for
some,
for
example,
for
example
in
the
in
the
in
the
analysis
by
either
vitalik
or
the
or
the
bitcoin
team,
for
example,
they
they
mentioned
a
couple
of
topics
that
they
were
looking
into.
For
example,
the
pro
quo
is,
for
example,
one
where
the
the
teams
were
were
giving
out
certain
rewards
for
the
people
that
that
contributed
to
to
the
bitcoin
grant
so,
for
example,
the
determination
whether
some
something
is
probably
good
or
not.
D
D
I
will
not
go
deeply
different
to
this
and
then
also
this
is
the
the
insight
in
terms
of
the
the
cohort
analysis
over
time
into
not
sure
how
familiar
familiar
you
are
with
this
analysis,
but
you
can
see
that,
let's
say
over
time
about
20
to
30
percent
of
of
of
contributors
also
contribute
in
in
the
in
the
next
round
and
and
so
forth.
Right.
So,
for
example,
let's
say
this
is
basically.
D
This
is
around
three:
it's
all
contributors
that
started
in
the
round
three
contributed
in
round
two,
and
then
we
can
see
four
forty,
four
percent
of
the
users
that
contributed
in
round
two
also
contributed
in
round
four
and
that
basically,
like
usually
goes
down
right.
That's
kind
of
like
the
analysis,
they've
designed.
C
D
C
D
I
did
I
I
just
took
the
data
from
the
github
repo
and
just
got
it
into
my
data
frame
and
something
that
I
didn't
notice
right
away.
That
like
in
quite
a
few
contributions,
the
the
amounts
are
replicated
across
the
same
contributor
and
that's
also
something
that,
like.
I
noticed
that
the
same
amount
across
multiple
grants
in
the
same
contributor
might
actually
mean
that
this
is
basically
a
collection
right.
Like.
D
But
that's
something
that
let's
say
it
might
mean
that
people
can
can
put
money
in
the
same
collection
and
that
might
be
interpreted
as
as
the
same,
let's
say,
copying
or
collusion
and
so
on,
which
is
obviously
a
false
positive
in
this
case.
And
then,
if
you
go
further,
this
is
basically
just
the
basic
stats
and
we
can
see
that
we
have
17k
time
steps
and
then
the
the
interesting
part,
of
course,
is
the
amount
because
the
flag.
C
D
The
amount
is
the
interesting
part
right,
so
we
can
see
that
the
the
mean
is
23
and
the
median
is
2.7,
which
basically
means
yeah
that
it's
quite
skewed.
We
can
see
that
also
from
the
center
deviation
and
then
also
that
only
25
percent
of
of
of
individual
grants
are
basically
above
five
dollars
right.
D
This
is
basically
the
75th
percentile
and
then,
when
it
comes
to
the
the
contributions
of
the
contribution
amount,
the
total
is
407k,
and
here
I
basically
made
a
comment
comparing
that
to
what
was
reported
in
the
round
overview,
and
this
is
of
course,
because
we
have
partial
data.
So
so
we
cannot
expect
all
of
those
and
then
also
the
the
unique
contributors
is
lower
4.7
k
compared
to
5
000,
which
was
mentioned
in
the
in
the
round
overview
and.
D
But
yeah,
so
the
grants
are,
are
very
low,
let's
say
lower
than
10
10
dollars
and
that's
something
that
I
basically
also
calculated
here
as
comparing
to
all
the
grants.
So
this
is
basically
just
limiting
all
only
the
grants
that
were
below
100
dollars,
because
if
I
go
further
than
it
sold
is
q,
then
it
doesn't
give
me
any
any
signal.
D
C
C
D
Like
almost
90
percent
of
grants,
the
the
grand
contributions
are
lower
than
ten
dollars
and
then
I
basically
just
split
the
the
the
contribution
amount.
Basically,
the
coin
top
contributors
by
total
contribution
amount,
and
we
can
see
that
that
we
have
like
two
kind
of
whales
right
that
have
contributed.
Let's
say
one
one
thousand
100
000
and
the
other
one,
let's
say:
70
000,
so
that's
170
000,
which
is
like
not
half
of
the
total
amount
that
we
have
in
the
data
set,
but
close
to
that
right.
D
D
The
number
one
is
is
the
most
prevalent
and
that's
basically
3.2
k,
which
is
compared
to
the
to
the
fourth
four
four
thousand,
seven
hundred
and
twenty
one,
it's
71.
So
let's
say
71
percent
of
of
users
contributed
to
one
grand
and
then
we
see
like
a
huge
drop
after
right.
So
that's
like
another
signal
that
that
we
get
that
it's
not
like
a
normal
distribution
at
all,
it's
like
as
usual,
so
then
top
ground
by
total
contribution
amount.
We
can
see
that
yeah.
D
D
Just
plotting
the
top
grants,
so
let's
top
10
grand
and
then
I
basically
just
checked
whether
whether
people
that
contributed
to
these
grants
basically
contributed
to
very
similar
to
very
similar
brands
right.
So
it's
so
somewhat
like
pairwise
correlation,
but
but
a
bit
different,
like
yeah,
it's
simpler.
C
C
D
D
There
would
be
much
signal
so,
but
here
like
we
can
already
see
certain
patterns
right,
for
example,
here
we
can
see
that
the
eu
crypto
initiative
is
it's
not
very
correlated
or
even
negatively
correlated
with
others,
but
but
some
other.
It's
like,
for
example,
they
get
quite
opens
for
a
support
fund.
It's
it's
quite
highly
correlated
or
or
the
this
one.
The.
D
C
D
H
D
D
Individually,
I
can
see
that
the
the
the
one
which
is
the
top
one,
which
is
four
nine
one-
this
one-
is
zero
c.
I
can
see
that
all
these,
this
one
has
more
grants
like
that
they
invested
in,
but
this
one
is.
We
can
see
that
there
is
a
deals
overlap
between
let's
say
the
interjs
is
here
then
the
hard
hats
deploy
like.
D
D
D
I
I
I
wasn't
really
sure
what
the
flag
meant
like.
I
basically
just
saw
that
all
the
values
are
zero,
so
I
assume
that
that
is
not
worth
exploring
into,
and
that's
it
so
and
and-
and
the
last
thing
that
I
looked
at
is-
is
the
the
total
contribution
amount
versus
uni
grant
contributions,
which
is,
let's
say
the
intuition?
Is
that
the
more
unique
contributors,
the
more
total
contribution
amounts?
D
D
Let's
say
that
there
are
certain
outliers,
like
here,
a
bank
list
that
you
mentioned
before,
that
that
doesn't
have
let's
say
a
huge
contribution
amount,
but
has
a
lot
of
unique
contributors
and
then,
on
the
other
hand,
we
have
these
two
outliers,
which
are
the
the
hard
hat
and
the
coin
center,
which
I
basically
don't
have
that
many
unique
contributors
relative
to
their
contribution
amounts
and
that
this
is
maybe
more
penalized.
Let's
say
based
on
the
quadratic
funding
mechanism
right
and
and
this
one,
for
example
like
they
like.
D
I
imagine
like
I
don't
know
like
how
much
much
funds
they
got
but,
like
I
mentioned
that
this
got
quite
a
lot,
considering
that
there
were
a
lot
of
unique
contributors
and
then
yeah.
The
last
thing
is
just
checking
the
the
top
grants
by
average
contribution
for
contributor,
and
this
is
yeah.
Most
of
them
are
basically
here
because
it's
let's
say
like
one
or
two
contributors
and
then
like
a
huge
contribution
amount.
D
C
C
D
D
That's
that's
not
the
case,
because
it's
just
like
the
timestamp
just
looks
like
just
number
of
rows,
and
so
it's
not
very
informative
in
terms
of
also
how
this
evolves
over
time.
It's
just
like
it's
just
arbitrary,
just
using
the
rows,
and
that's
it
right.
So
it's
so
this
would
be
another
thing
that
would
be
useful
like
actual
time
timestamps,
but
maybe
that's
just
not
good
for
for
yeah,
for
that
you
can
basically
recognize
or
who,
who
made
the
contributions
and
so
on
right.
E
A
D
A
A
B
B
B
C
C
B
A
Not
at
the
moment,
I
guess
you
can
at
any
time
reach
out
to
yan
in
case
you
have
questions
and
play
around
with
the
notebook
as
well.
Thank
you
again,
dan
for
sharing
this
great
work,
and
now
I
think
it's
it's
time
to
work
on
our
research
questions.
So,
first
before
we
start
I'd
like
you
to
check
out,
if
you
can
use
lucidchart,
you
might
be
aware
of
that
tool.
A
A
A
A
A
A
J
A
A
Brainstorm,
the
most
exciting
research
questions
throughout
this
presentation
today
the
sessions
we've
had
and
it
doesn't
matter
if
you
feel
comfortable
with
working
on
this
research
question
yourself
or
alone.
It's
just
some
of
the
aspects
you
really
think
are
worth
exploring
and
that
we
should
take
a
look
at
in
this
research
group
and
just
pick
a
note.
A
A
post-it
and
one
post
tip
per
research
question
just
take
one
and
put
it
to
the
write
on
it.
First,
everyone
is
removing
it
from
the
board.
But,
finally,
your
research
question
should
be
back
on
the
blue
board
and,
let's
collect
the
most
exciting
research
questions,
most
valuable
research
questions
you
see
in
this
project
on
this
blue
board,
I'd,
say
five
minutes
type
down
and
create
as
many
posts
as
you
can
just
don't
think
at
the
moment.
A
A
A
C
C
J
J
A
A
A
A
And
see
what
you
think
is
worth
exploring
in
this
research
group
now
I
think
we
can
all
look
at
one
screen
so
I'll
try
to
share
but
feel
free
to
add
notes.
While
we
are
discussing-
and
let's
start
with,
can
we
build
a
model
of
known
methods
of
collusion
and
fraud
to
see
how
to
subvert
this?
Who
has
put
this
note,
who
has
added
this
question.
C
C
A
A
J
G
So
I
have
one
that
might
be
kind
of
related
just
directly
to
the
right
skipping
over
the
big
one
in
the
middle
can
we
use
reinforcement,
learning
in
our
simulations
to
identify
collusion
patterns.
I
think
we
might
be
able
to
use
ai
to
identify
collusion
patterns.
This
is
common
in,
like
in
in
ai
research.
There's
often
video
games
are
used
as
the
research
bed
and
sometimes
brand
new
sort
of
exploits
in
these
video
games
are
discovered
by
the
ai
systems.
G
So
I
wonder
if
we
could
get
the
objective
function,
right
sort
of
have
this,
like
collusion,
objective
function,
kind
of
like
the
optimality
gap
that
we
see
in
the
research
so
far
and
have
some
aia
systems
that
can
find
these
collusion
strategies,
and
then
we
could
like
index
them
to
build
a
model
of
known
methods.
That
would
be
pretty
interesting.
A
G
Sure
I
kind
of
I
have
about
three
ideas
that
I
kind
of
rambled
on
in
in
various
sticky
notes.
So
over
there's
like
two
columns
where
that
ai
one
come
from.
There's
I
think
seven
more
so
one
of
the
main
ones
I'm
interested
in
is
is
like.
Can
we
scrape?
Can
we
collect
additional
data
sets
so,
for
example,
for
each
grant
that
was
rewarded?
G
Maybe
we
could
even
manually
we
could
just
find
like
a
website
for
each
grant.
I
don't
know
that
might
be
a
lot
depending
on
the
grants.
Maybe
we
could
do
that
automatically,
but
then
I
was
thinking
if
we
could
connect
to
say
twitter
data.
If
these
grants
are
associated
with
some
sort
of
twitter
account
or
github
it's
in
the
top
right
there,
the
metadata
can.
G
So
yeah,
I
think,
that'd
be
interesting
to
and
then
I
kind
of
go
on.
I
think
the
most
interesting
one
for
me
is
twitter,
it'd,
be
interesting
to
sort
of
model,
twitter
networks
and
then
compare
that
if
we
could
do
some
sort
of
comparison,
analysis
between
twitter
networks
and
the
git
coin,
grant
networks
or
any
sort
of
direction
this
that
this
sort
of
metadata
approach
could
go
and
then
the
other
main
topic
I
was
thinking
of
is
like
optimizing,
just
general
optimization
of
the
git
coin
grant
system.
G
So
I
was
thinking
like
what
kind
of
measurements-
and
I
think,
this
one's
the
most
abstract,
but
like
it's
just
down
below,
how
can
we
measure
optimal
funding
for
the
ecosystem,
such
as
maximizing
impact,
innovation
or
well-being,
and
then
below
that?
What
other
metrics
can
we
use
to
measure
the
health
of
the
ecosystem?
G
And
I
would
like
to
just
read
sort
of
the
mission
statement
of
get
coin
and
see
really
what
are
they
trying
to
do?
What
are
they
trying
to
maximize?
I
think
it's
funding
for
public
goods,
something
along
those
lines,
but
I
haven't
read
their
sort
of
mission
statement.
So
it'd
be
interesting
to
see
what
they
are
exactly
trying
to
achieve
and
if
there's
any
sort
of
like
branches,
we
could
go
off
of
this
for
optimizing
different,
different
impact
measurements
and
then
yeah.
So
those
are
the
three
ideas
I
think
I
was
like.
G
G
Yeah,
oh
yeah,
the
one
in
the
top
left.
This
is
about
matrix
representations
with
the
research
that
danielle
has
been
doing,
basically
just
any
any
network
data
structures
that
we're
using.
I
know
it
would
be
cool.
We've
talked
about
this
a
few
times
to
translate
the
network
algorithms
to
like
vector
and
matrix
representations
for
computational
efficiency.
So
I
find
that
one
quite
interesting
too.
J
J
J
Bright
id
and
all
these
things,
but
it's
still
such
a
high
priority.
So
that
makes
me
wonder
why.
So
that's
one
of
my
questions.
My
second
question
is
something
I've
been
kind
of
wondering
about,
though
you
know
there's
so
lot
to
think
about,
but
maybe
using
like
quadratic
funding
for
rewarding
data
stakers,
so.
C
J
J
C
A
C
A
C
E
I'm
really
interested
interested
in
the
dynamic
behavior
during
a
round.
So
what
do
I
mean
by?
That
is
that
if
you
can,
if
you
can
see
how
contributors
are
staking
on
grants
over
time,
maybe
you
can
you
discover
you
can
discover
some
sort
of
behavior
collusion
wide
or
not
in
order
to
get
some
meaning
about
from
what
is
happening
during
overtime?
E
C
A
E
E
E
C
E
C
E
E
A
A
E
C
J
J
I
F
C
J
J
So
you
know
how
how
does
the
time
over
people
allow
it
to
vote
or
actually
find
a
project
impact
their
decisions
in
that
sense
and
the
other?
The
other
question
I
guess
I
was
just
thinking
about-
was
just
different
kinds
of
functions.
Besides,
like
quadratic
equations,
how
do
those
look
like
and-
and
I
guess
I
dropped
parallel
to
sort
of
activation
functions
within
you
know
deep
learning
or
other
machine
learning
algorithms?
I
was
curious
to
see
you
know
how
would
these
equations
suggest
impact
the
decisions
of
people?
J
So
if
you
look
at
like
the
sigmoid
function
has
like
an
upwards
curve
to
a
certain
extent
and
that
it
flattens
out,
so
I
was
wondering
if,
for
example,
we
apply
this
to
the
way
people
fund
projects,
if
that
would
have
any
impact
in
a
sense
and
the
other
topic.
I
was
reading
more
about
bonding
curves
and
it
seems
like
there's
a
corollary
between
just
the
way
that
quadratic
voting
or
funding
works
and
how
bonding
curves
are
also
modeled.
A
A
A
J
F
Right
so
I
think
it
was
sean
I
believe
was
talking
about
collecting
metadata,
and
I
find
that
really
interesting,
especially
given
that
this
date
is
on
the
blockchain.
So
I've
written
a
piece
of
software
called
trueblocks,
which
lets
me
extract
pretty
much
any
data
that
you
could
possibly
want
from
the
blockchain.
F
So
I've
already
done
a
lot
of
work.
Extract
bitcoin
has
always
been
sort
of
my
test
case
for
data,
so
I've
already
done
a
huge
amount
of
extraction
of
metadata
about
the
bitcoin
grants,
and
while
I
recognize
that
it's
important
to
keep
that
this
is
sensitive
data,
it's
also
on
the
blockchain,
which
means
that
it's
totally
transparent.
F
So
I
want
to
think
about
how
open
this
data
actually
is
and
what
we
can
actually
see
in
this
data
that
maybe
I
think
almost
everybody
doesn't
realize
how
open
it
is,
because
I
can
see
things
that
I'm
certain
people
don't
want
me
to
see
so
I'm
kind
of
on
a
fence
of
being
sensitive
to
people's
privacy.
I
certainly
think
that's
extremely
important,
but
I
think
that
the
data
is
much
more
exposed
than
it.
Anybody
really
realizes.
C
F
C
F
A
H
C
A
H
C
D
Yeah,
I
can
basically
go
in
that
sense
yeah,
so
my
both
are
basically
like
the.
How
can
we
model
the
metoo
effect
in
git
congruent
and
that's
the
first
one
and
the
second
one
is:
how
can
we
quantify
the
situation
effect
on
grand
contribute
contribution,
behavior?
So
those
are
the
two
questions
and
yeah.
I
I
don't
have
a
clear
idea
how
to
answer
them
like,
like
the
first
one.
D
The
mental
effect
like,
as
far
as
I
know
is
the
is
like
the
the
chapter
five
from
the
network
science
group,
that
that
is,
is
part
of
the
of
the
reading
material,
that's
kind
of,
but
I
don't
know
how
deep.
Basically,
anyone
went
into
this
kind
of
research
and
also
how
to
relate
that
to
the
actual
data.
D
C
A
C
I
A
A
I
Okay,
cool
mine
are
the
two
in
the
top
right,
the
purple
ones,
and
they
seem
related
to
maybe
like
two
topics
where
they
talked
about
so
the
first
top
right
one.
What's
the
efficacy
of
pairwise
method
for
combating
collusion,
so
danilo
was
kind
of
saying
that
he
hasn't
really
looked
at
that
a
ton
and
then
like?
Is
there
a
more
extensible
general
method?
I
So
I
think
that
one
there
was
a
group
over
here
of
I
don't
know
where
it
was.
There's
a
group
off
to
the
left
of
like
security
based
questions
and
then
the
second
one
is.
Can
we
develop
a
performance
metric
for
the
fairness
of
quadratic
funding,
and
this
goes
back
to
sort
of
like
the
mission
statement
question
of
like
what
are
what
is
get
coin,
trying
to
optimize
but
it'd
be
cool
to
know
like?
I
I
A
C
A
Cool
then,
for
the
last
couple
of
minutes
in
this
course
I'd
like
to
you
to
take
this
red
sticky
note
put
your
name
on
it
and
then
just
stick
it
to
the
question
you
find
most
interesting.
Of
course
you
can
now
cluster
questions
and
then,
let's
find
out
if
we
have
teams,
if
you'd
like
to
work
individually,
just
pick
whatever
you'd
like
to
go
first
with
perhaps
eight
weeks
are
enough
to
work
on
more
than
one
question,
but
if
you
would
need
to
choose,
you
have
to
choose
that.
A
A
A
A
A
A
A
A
You
can
find
this
in
the
schedule
tomorrow
february
18th
at
6,
00
or
12
eastern
time,
bitcoin
model
collab
the
live
coding
session
and
two
and
a
half
now
three
and
a
half
hours
later,
the
academic
session
and
this
academic
session
now
is
the
chance
to
maybe
sharpen
your
research
question
a
little
bit,
define
it
and
then
introduce
it
and
get
feedback
from
bitcoin
and
block
signs
as
well
and
also
danilo
hasn't
had
the
chance
to
make
any
comments.
But
certainly
he
has
comments.
A
C
A
Yeah,
okay
cool,
so
you
can
use
this
as
a
template,
so
copy
it
and
now
add
music
for
your
for
formulating
your
own
research
question
and
then
also
start
with
taking
down
your
road
map,
so
make
make
a
first
overview
on
okay.
What
is
this
question
all
about?
What
are
what's
the
target
the
objective
here
and
then
you
might
want
to
add
first
definitions,
but
that's
not
mandatory
in
any
case,
try
to
formalize
your
hypothesis
and
then
also
add
some
more
details
on
the
methodology
you'd
like
to
apply
here
right.