►
Description
Three IPFS Application show off what they’ve built and share the critical need the protocol and implementations must deliver on in order to succeed.
A
A
So
in
2019,
Jack
Dorsey,
the
then
CEO
of
Twitter
announced
that
Twitter
was
going
to
be
funding
a
decentralized
social
protocol
and
then
they
brought
together
some
outside
collaborators
in
a
matrix
chat
room,
and
this
included
myself,
and
during
this
time
we
produced
an
ecosystem
review
and
then
in
2021.
They
selected
me
to
lead
my
vision
of
the
project
for
blue
sky
and
so
I
founded
an
independent
company,
and
then
we
received
funding
and
in
case
you're
wondering
we
are
fully
independent
from
Twitter.
So
no
Elon
is
not
my
new
boss.
A
I'm
relieved
about
that,
so
this
year,
I
hired
a
team
and
then
we
chose
some
the
architecture
for
how
we're
going
to
do
this
and
we
released
an
early
prototype
called
adx.
The
authenticated
data
experiment.
We
have
since
renamed
this
to
the
at
protocol
or
the
authenticated
transfer
protocol
or
ATP,
and
we
announced
last
week
that
we
are
building
an
app
so.
A
Overview
of
that
protocol
it
uses
ipld
in
the
data
repos,
and
so
that
is
essentially
a
Merkle
tree
where
each
node
is
uses
ipld
and
then
the
identity
are
dids,
decentralized,
identifiers
that
link
a
human,
readable
name
and
a
cryptographic
key.
And
then
this
is
used
to
sign
the
Merkle
root
and
the
network
is
Federated,
so
users
have
servers
that
do
the
heavy
lifting,
but
then
this
allows
their
data
to
be
kept
in
a
canonical
form
that
they
can
move
around.
A
A
A
The
Blue
Sky
app
is
coming
soon
and
we
announced
last
week
that
we
are
doing
a
private
beta
and
then
the
waitlist
quickly
filled
up
and
is
now
over
50
000.
So
we're
going
to
break
those
up
into
smaller
groups,
yeah
and
so
I.
Don't
have
any
specific
asks
on
the
ipfs
side,
because
ipld
is
working
well
for
us.
But
if
you
would
like
to
get
feedback
on
our
protocol
docs,
you
can
do
that
at
proto.com.
A
You
can
join
our
Matrix
channel
to
chat
with
our
devs
and
that's
the
link
is
on
blueskyweb.xyz
and
if
you
want
to
try
our
app,
the
wait
list
is
very
long.
So
for
this
audience,
if
you'd
like
to
message
me
directly
on
Twitter
I
can
try
to
get
you
in
earlier
than
number
50
000..
So
thank
you
very
much
I'll
hand
it
over
the
next
speaker.
B
So
we're
building
the
world's
fastest
growing
Weather
Station
Network
we
launched
earlier
this
year
and
we
already
have
9
000
members
in
our
Discord
server.
We
have
more
than
5
000
station
sales
in
66
countries,
and
this
is
the
weather
forecast
for
this
week
from
five
different
meteorological
models.
You
can
think
of
it
as
five
different
mobile
apps
and
the
blue
line
is
one
of
our
weather
stations
here
in
Lisbon.
What
it
has
recorded
So
as
you
can
see
the
first
day
more
or
less.
B
The
forecast
aligns
the
the
what
happened
in
reality,
but
the
second
day
most
of
the
forecasts
were
wrong.
So
that's
that's
a
small
problem
for
individuals,
but
it's
a
huge
one
for
the
agriculture
industry,
the
energy
supply
chain
industry.
In
fact,
one
third
of
the
worldwide
economy
is
weather.
Sensitive.
That's
why
governments
spend
billions
monitoring
and
forecasting
the
weather,
but
it's
not
a
solved
problem.
It's
not
easy
to
be
solved
so
inspired
by
filecoin
and
helium
we're
trying
a
different
approach:
we're
building
an
economic
system
around
weather
data
in
which
weather
station
owners
are
rewarded.
B
With
our
token
for
the
data
contribution
and
whether
sensitive
Enterprises
customers
they
purchase
and
spend
our
token
in
exchange
for
weather
data
and
Weather
Services,
to
support
this
economy,
we
have
to
create
our
own
Hardware.
B
So
our
community
benefits
already
from
our
mobile
app
Android
and
iOS,
and
we
have
shipped
2000
units
we're
shipping
1000
more
on
Monday
next
week
and
we're
scheduled
to
produce
more
than
10
000
units,
11
000
by
the
end
of
the
year.
B
For
for
comparison,
Noah,
the
U.S
Met
Office
is
operating
14
000
stations,
so
we're
we're
getting
there.
So
the
next
steps
in
terms
of
development
for
our
project
is
we
are
progressing
the
integration
with
ipfs
falcoin
plus
computer
data.
We
want
to
make
sure
that
the
data
we
collect
from
the
stations
have
the
maximum
quality.
We
are
building
a
number
of
mechanisms
to
track
the
forecast
accuracy
and
later
on.
Long-Term
goal
is
to
act
as
a
weather
Oracle
on
on
chain
to
enable
new
types
of
web3
applications
that
need
weather
data.
B
Okay,
that's
a
bit
broken,
but
the
so
two
years
ago
we
started
experimenting
with
ipfs
and
the
vision
back
then
until
today
is
that
we
want
to
build
a
weather
superstructure,
a
Weather
Network.
That
is
a
superstructure
that
is
Unstoppable,
so
we
think
that
weather
stations
should
be
able
to
connect
directly
to
ipfs
nodes.
So
if
we
pull
the
plug
from
our
servers,
then
the
network
won't
have
a
problem.
That's
why
we
started
working
on
a
ipfs
client
for
microcontrollers
that
are
powering
those
stations.
Obviously
those
Stations
don't
have
Linux.
B
They
are
way
lower,
so
we
did
world's
first
ipfs
client
for
for
esp32,
so
it's
like
an
Arduino
compatible,
but
then
we
run
into
a
problem
which
is
how
do
authentic?
How
to
authenticate
thousands
of
devices
directly
on
ipfs-
and
we
did
a
very
you-
know-
lame
implementation
of
using
basic
HTTP
authentication.
B
So
we
have
to
hard
code
credentials
in
the
devices
themselves.
Since
then,
we've
done
a
number
of
different
things,
but
in
in
my
point
of
view
it
would
be
cool
if
we
could
directly
use
the
X
509
certificate
that
we
have
in
the
hardware
to
authenticate
directly
to
an
ipfs,
node
and
I.
Think
the
closest
that
exists
in
towards
this
direction
is
web
3
hours,
but
it
it
I
think
what's
missing
is
an
Amazon
Cognito
for
web3.
So
if
anybody
are
working
towards
this
direction,
I'll
be
happy
to
talk
about
it.
C
My
name
is
Francis
tires
I'm,
a
assistant
professor
of
computational,
linguistics
at
Indiana,
University
and
I
work
with
language
technology
for
indigenous
and
marginalized
language
communities.
Today,
I
will
be
talking
about
omnilingo,
which
is
a
project
I
have
on
language
learning.
C
So,
as
many
of
you
will
know,
learning
languages
is
quite
hard
and
one
of
the
more
difficult
Parts
is
listening
and
pronunciation.
So
there
are
very
few
apps
for
language
learning
and
very
few
free
apps.
So
the
idea
is
basically
to
make
an
app
that's
possible
to
use
for
practicing
both
listening
and
pronunciation.
C
So
the
basic
idea
is
that
we
should
use
ipfs
to
store
all
of
the
data
so,
for
example,
Clips
transcripts
metadata
all
of
this
and
machine
learning
models
and
then
have
a
way
that
people
can
put
this
together
to
make
their
own
apps.
So
it
could
be
targeted
at,
for
example,
language,
revitalization
or
language
instructors,
or
basically
anyone
who
has
ever
sent
an
email
to
Duolingo
and
have
them
say
I'm.
Sorry,
we
don't
have
time
for
that.
C
So
originally
we
had
a
Homebrew
system
content
addressing
system
using
HTTP,
but
then
we
found
out
about
ipfs
and
so
now
we're
using
ipfs.
For
all
of
this,
we
have
all
of
the
data
is
from
common
voice.
We
have
audio
files
in
MP3
and
Json
files
for
the
transcripts
and
all
of
the
metadata,
and
we
have
a
demo
app
using
ipfsjs.
C
So
the
general
idea,
it's
a
gap,
filling
exercise
which
you
will
probably
be
familiar
with.
You
listen
to
a
clip
and
there's
a
word
missing
and
you
have
to
fill
in
the
word.
The
clips
start
easy
and
they
get
progressively
harder
and
the
entire
app
is
more
or
less
language
independent.
So
we
use
language
data
from
mozilla's
common
voice
project,
which
has
data
for
over
100
languages.
Now,
and
it's
very
easy
to
add
new
languages
to
this
this
app.
C
So
you
can
see
an
example
here,
we're
not
designers,
as
you
will
probably
notice,
and
you
can
also
try
it
out
online.
So
don't
everyone
go
there
at
once,
because
you
might
take
down
the
server.
C
So
we
have
this
demonstration
app
for
The
Gap,
feeling
exercise
and
we're
working
on
pronunciation
feedback.
This
will
involve
storing
more
stuff
on
ipfs,
for
example,
speech
recognition
models.
C
We
would
like
to
make
a
nicer
interface,
neither
of
us
who
are
working
on
this
graphic
designers
or
ux
people,
and
we
would
like
to
have
different
tasks
in
the
app
so,
for
example,
choose
one
or
build
words
out
of
blocks
and
stuff
like
this,
and
we
would
also
like
it
to
be
a
lot
more
reliable.
So
this
brings
us
to
what
I
was
asked
to
talk
about
when
I
was
asked
to
give
the
the
short
talk.
People
said:
hey:
what's
the
problem
with
ipfs?
C
Can
you
tell
us
so
I
do
a
lot
of
work
and
research
in
the
mountains
in
Mexico,
so
I
work
with
a
nawat
speaking
Community
there
and
the
internet
access
isn't
great
and
the
best
best
possible
internet
access
you
can
have
is
starlink,
so
that's
like
Elon
musknet.
Unfortunately,
ipfs
doesn't
work
over
starlink,
so
we
can
use
it
over
the
like
3G
or
4G,
but
it's
insanely
slow.
C
So
that's
one
thing.
The
other
thing
is
most
of
the
time
when
we're
using
ipfsjs,
which
is
fantastic,
excellent
work,
but
because
everything
has
to
go
through
the
preload
notes
about
half
of
the
time.
The
connections
don't
make
it,
especially
if
we're
on
like
low
bandwidth
or
low
reliability,
internet
connections
yeah.
So
that's
on
the
connectivity
side.
On
the
developer
side,
we
would
like
to
see
an
implementation
for
C
or
for
python,
but
if
we
have
one
in
C,
we
can
make
one
for
python.
C
The
python
HTTP
implementation
is
out
of
date
and
unmaintained
and
there's
really
poor
integration
with
Linux
distribs
and
those
are
my
thank
you
aggressive,
mientus
and
I.
Think.
That's
me.