►
From YouTube: The Data Enthusiast's Toolkit
Description
Unlock the power of data with Datasette! Join Simon Willison to explore this open source tool that's revolutionizing data exploration and publication at 1 pm ET on May 19th
A
A
A
A
How
can
we
get
involved
or
we
get
to
a
chance
to
learn
more
about
Open
Source
before
I
introduce
our
Our
Guest
I
do
want
to
like
point
out
just
people
saying
hello
feel
free
to
like
say
where
you're
calling
in
from
someone
said:
greetings
from
India,
so
hello,
everyone,
I'm
super,
excited
I.
Think
I
recognize
this
person's
name
from
resilient
coders,
which
is
a
coding,
bootcamp,
I
graduated
from
so
hi.
A
B
B
A
software
engineer
can
do
and
I've
also
I've
I've
ran
a
startup
for
a
few
years,
which
I
sold
to
Eventbrite
and
moved
out
to
California
I'm
from
England,
actually
and
so
I've
done
sort
of
like
architectural
work
on,
but
helping
build,
helping
scale
Eventbrite
and
things
like
that
too.
So
I
have
a
lot
of
different
interests.
B
A
B
So
the
idea
of
the
data
set
data
set
I,
call
it
an
open
source
multi-tool
for
exploring
and
Publishing
data.
Basically,
the
idea
is
you've
got
some
data,
so
you've
got
like
a
CSV
file
or
a
bunch
of
Json
or
something.
What
do
you
do
next,
like?
What's
the
next
step
to
do
something
with
that
data
and
I
started
this?
Originally,
it
was
inspired
by
work.
B
I
was
doing
at
newspapers,
so
I
worked
for
the
Guardian
newspaper
in
London
and
we
collected
so
much
data
about
the
world
because,
anytime,
you
see
an
infographic
in
the
newspaper.
You
see
a
map
or
a
chart.
Somebody
had
to
go
and
get
the
numbers
and
those
numbers
were
reported
right.
A
reporter
contacted
a
government
agency
got
the
numbers,
double
checked
them
like
there's
a
lot
of
work
that
goes
into
that
raw,
that
raw
information
and
I
met
the
journalist
at
the
guardian.
B
Who
was
the
sort
of
lead
on
on
doing
this
work,
a
chap
called
Simon
Rogers
and
he
had
hundreds
of
spreadsheets
on
his
computer,
just
hundreds
of
meticulous
spreadsheets
telling
about
the
world
and
they
lived
under
his
desk
on
like
a
big
desktop
computer
and
we
got
thinking
we
were
saying:
okay.
Well.
Is
there
any
reason
not
to
publish
this
data
like?
Could
we
start
publishing
the
data
behind
the
stories,
and
so
we
launched
this
thing
called
The
Guardian
data
block
where
the
idea
was
anytime,
we
put
a
story
out,
that's
backed
by
data.
B
We
also
publish
the
data,
so
you
can
go
to
the
data
blog
and
get
the
spreadsheet
of
numbers
that
we
use
to
report
a
story
on
the
economy
or
demographics
or
whatever
it
is,
and
at
the
time,
the
tool
we
used
for
publishing
the
data
was
Google
Sheets
because
it
works
it's
free.
Everyone
knows
how
to
use
it,
we
didn't
have
to
actually
buy
a
product
for
it
and
that
worked
out.
Okay,
but
I
always
felt
frustrated
by
its
limitations.
You
know
it's
it's
great
for
medium-sized
amounts
of
data.
B
B
Now,
because
what
verse
Outlets
you
do,
is
they
let
you
host
a
web
application
online
essentially
for
free,
provided
it
doesn't
have
a
database
like
they
let
you
do
stateless
applications,
which
is
great,
but
if
you
need
a
database
which
has
to
have
backups
and
all
sorts
of
complexity
like
that,
they
kind
of
leave
you
on
your
own.
You
have
to
go
and
spend
money
somewhere
and
I
realize
that
if
you're
publishing
data
you
don't
actually
need
a
full
database,
you
need
just
a
read-only
database
and
SQL
Lite.
B
The
the
embedded
database
that's
been
around
actually
for
20
years
as
well.
Just
gives
you
that
so
the
idea
I
came
up
with
is
okay.
What
happens
if
we
take
a
bunch
of
data?
Stick
it
in
a
sqlite
database
file,
which
is
just
a
file
on
disk,
and
then
we
package
that
with
an
application
and
we
ship
it
to
the
cell
or
Google
Cloud
run
or
Heroku,
or
one
of
these
very
inexpensive
hosting
providers
and
the
the
data
itself
just
becomes
a
another
asset.
B
It's
like
a
big
CSS
file
or
something
that's
bundled
into
your
application,
and
you
include
a
web
app
that
lets
people
interact
with
that
database.
So
that
was
the
initial
idea
for
data
set.
It
was
what
happens
if
I
build
a
simple
python
web
application
that
sits
on
top
of
a
read-only
database
and
lets
you
poke
at
it
and
lets
you
search
it
and
run
SQL
queries
against
it
and
browse
it,
and
all
of
that
kind
of
stuff
and
host
it
for
literally
cents
per
month,
like
the
cost
of
Hosting.
B
So
I
think
we
should
probably
jump
into
a
demo.
So
I
can
show
you
what
this
thing
looks
like
I
think
I've
got
screen,
sharing
set
up
yeah.
So
this
is
the
website
for
data
set
data.io.
This
is
actually
data
set
itself.
This
is
the
software
with
some
custom
templates,
but
that's
kind
of
misleading
I'll.
Show
you
what
it
an
example.
What
it
looks
like
so
here
is
a
database
that
I
created
of
the
training
data
behind
stable
diffusion,
the
image.
B
It
was
one
of
the
things
that's
so
interesting
about
these
models
is:
how
are
they
trained
like?
How
do
these
things
work
and
it
turns
out
stable
diffusion
is
trained
on
like
a
few
billion
images,
but
there
are
12
million
of
them
that
are
flagged
as
as
having
the
best
Aesthetics
and
have
the
highest
influence
on
stable
diffusion,
and
so
I
got
a
hold
of
database
of
these
12
million
images
and
I
stuck
in
data
set,
and
now
we
can
search.
We
can
say
things
like
Obama.
B
Cool
yeah:
this
is
data
set
it's
it's
a
table,
but
you
can
also
do
things
like
one
searches.
So
here
I've
searched
for
Barack
Obama
I've
searched
for
Obama.
There
are
9
000
images
that
went
into
this
training
set
of
Obama,
and
you
can
see
that
they're,
not
all
Obama
right,
like
that's,
not
Barack
Obama.
That
one
is
that
one
he
stood
there.
It's
kind
of
fascinating
how,
because
it's
just
going
on
the
captions
for
these
images,
so
it's
kind
of
a
miracle
that
these
things
are
so
good
at
rendering
Obama.
B
When
the
training
data
is
very
untidy,
yeah
one
of
the
features
of
data
set
that
I'm
really
I
use
a
lot
is
the
ability
to
facet
my
things.
So
here
we
can
see
so
we've
got
9000
results
for
the
bomber.
If
I
say
facet
by
domain,
ID
I
can
see
that
419
of
them
came
from
the
Daily
Mail
242
from
Fine
Art
America
233
from
Pinterest.
B
I
have
lots
and
lots
of
other
demos.
I'll
show
you
this
one,
because
it's
kind
of
fun.
This
is
a
demo
I
built
of
this
data
set
of
parks
in
California,
and
this
is
demonstrating
geospatial
data.
So
one
of
the
things
you
can
do
here
and
dataset
has
a
concept
of
plugins,
which
can
add
extra
features
and
I
built
a
plug-in
that
lets
you
draw
a
shape
on
a
map
and
it
will
then
run
a
query
for
just
things
within
that
shape
on
the
map.
B
So
now,
I
can
see
that,
within
that
shape
of
the
Bay
Area,
we've
got
2
600
parks,
wow
I'm,
going
to
search
for
mini
and
see
how
many
of
those
are
mini.
Parks
here
we
go.
This
is
every
mini
park
in
that
shape,
that
I've
drawn
and
really
this
is
a
demonstration
to
show
that
you
can
that
when
you've
got
plugins,
you
can
do
all
sorts
of
weird
and
wonderful
extra
features.
B
On
top
of
this
another
thing
you
can
do
with
type
set,
which
I
hints
that
earlier
you
can
use
custom
templates
to
turn
a
data
set
instance
into
a
full
website.
So
here
what
I've
done
is
I've
taken
the
archives,
the
San
Francisco
microscopical
society,
which
is
a
150
year
old
Society
of
microscope
enthusiasts
based
in
San,
Francisco
and
I.
Also
they
they
scanned
all
of
their
journal
entries
going
back
150
years,
I've
run
those
through
OCR
and
now
I've
got
this
dataset
instance,
which
can
actually
run
searches.
B
So
if
I
search
for
say
Isaac
Newton,
it
will
show
me.
Oh
it's
crazy.
Listen
here,
we
go.
It'll,
show
me
pages
in
their
archives.
That
are
that.
Mention
is
at
Houston,
and
this
is
actually
just
my
data
set
software
running,
but
with
a
oh
I'm,
trying
to
remember
I'm,
trying
to
remember
where
the
database
of
this
is.
B
Maybe
it's
slash
data.
Let's
check
this
oops,
okay,
it's
called
sfms,
and
so,
if
I
go
here,
this
is
just
data
set
again
it's
a
sqlite
database
with
a
bunch
of
documents.
Each
document
document
has
a
bunch
of
pages
and
then
the
rest
of
the
site
is
custom
templates
built.
On
top
of
this,
that
give
you
this
interface
for
searching
through
journal
entries
about
microscopes,
so
yeah,
it's
I'm
interested
in
a
lot
of
things
and
one
of
the
reasons
I'm
so
excited
about
data
set
as
a
project.
B
B
Going
one
of
my
hobbies
is
I
love,
exploring
tiny
museums,
so
whenever
I
go
to
a
new
city,
I,
look
on
Google,
Maps
and
museums
and
I
scroll
past
all
the
big
ones
and
I
try
and
find
the
little
ones,
because
the
little
ones
are
always
more
fun
because
it
doesn't
matter
what
they're
about.
If
they're
small
enough
the
person
who
runs
the
museum
is
probably
there.
So
then
you
get
to
go
and
have
a
conversation
with
somebody
about
their
little
tiny
Museum,
and
so
this
website
right
here,
is
how
many
is
it?
B
It's
111,
little
museums
that
I've
been
to
and
for
each
of
them,
I've
got
like
photographs
and
notes
about
what
I
saw
there
like
here's,
the
Golden
State
Model
Railroad
Museum,
which
has
photos
of
the
Railway
and
all
of
that
sort
of
stuff.
Super
super
rewarding
hobby.
But
again
this
website
is
data
set.
If
I
go
to
oops.
B
If
I
go
to
slash
browse
here,
we
go,
we
can
see
that
there's
a
database
table
called
museums
and
it's
got
all
of
those
museums
in
this
is
a
plug-in
I
wrote
which
kicks
in
if
it
finds
a
latitude
and
longitude
column
and
draws
them
all
on
a
map.
So
now
I
can
zoom
in
and
see.
Oh
here
are
the
like:
wow
there's
27
I
visited
in
the
Bay.
Area
I
mean
I
live
here.
So
that's
not
you
supplies,
but
then
it
also
lets
you
build
features
like
I've
got
a
use.
B
That
was
a
great
one.
There's
an
aviation
museum
nearby,
all
of
that
sort
of
stuff-
and
all
of
this
is
completely
everything
I've
shown
you
today
is
open
source.
If
you
go
to
Simon,
W,
slash
museums,
this
is
the
GitHub
repository
that
shows
that
it
has
all
of
the
source
code
for
my
museums
and
all
of
the
data
and
all
of
that
kind
of
thing,
but
yeah.
So
it's
there
were
a
lot
of
moving
past.
B
This,
oh
I've
got
actually
one
more
relevant
demo,
so
data
set
is
a
python
web
application,
which
means
that,
in
order
to
use
it,
you
have
to
be
able
to
install
Python
and
use
Python
to
install
data
set
and
run
terminal
commands
on
your
computer
and
it's
if
you're
a
python
developer.
It's
completely
fine.
If
you're,
not
the
barrier
to
entry
is
quite
substantial
in
getting
this
thing
running,
which
I
find
really
frustrating,
because
I
want
people
to
be
able
to
use
my
software
yeah.
B
So
last
year,
I
started
getting
interested
in
webassembly
and
being
able
to
like
run
things
run
more
things
in
the
browser
and
I
realized
that
with
webassembly
you
can
actually
run
python
applications
entirely
in
the
browser
as
well,
which
is
kind
of
an
amazing
trick.
So
I
built
this
thing
called
data
set
light
and
actually
for
this
demo
I'm
going
to
open
up
the
dev
tools.
So
you
can
see
what
it's
doing.
Let's
do
the
network
tab
so
when
I
hit
this
page.
B
B
The
python
interpreter
is
now
running
in
my
browser
and
then
it
runs
pip
install
and
it
installs
data
set
itself
and
all
of
its
dependencies
and
then
it
launches
at
the
browser.
So
this
right
here
is
my
python
web
application,
except
if
you
look
at
the
URLs,
it's
a
slash
like
hash,
slash
fixtures.
This
is
data
set
running
entirely
in
the
browser,
which
is
honestly,
it
kind
of
amazes
me
that
this
is
even
possible.
B
B
You
can
actually
put
these
in
gists
as
well.
So
if
I
look
at
by
the
way
that
someone
called
train.json
here,
we
go.
Here's
some
here's,
some
more
training
data
from
a
from
a
large
language
model.
This
is
from
the
staff
code
model,
so
I
can
copy
and
paste
the
URLs
on
that.
Yes
and
I.
Can
click
load
Json
paste
that
in
it
will
now
reload
that
application,
but
one
of
the
things
it
will
do
is
it
will
hit
it
should
be
in
there?
B
Now
I
can
see
in
this
case
it's
exact.
This
is
this
is
code
from
GitHub
that
was
used
to
train
the
starcode
language
model,
which
can
then
write
code
and
things
and
a
couple
things
in
a
read-only
database.
You
can
run
your
own
SQL
queries,
so
in
this
case
I
can
say
you
know
what
I
just
want
the
Sha
and
the
language,
and
let's
do
the
content
from
this
table
here
and
hit
run
SQL
and
it'll.
B
B
A
B
B
So
this
is
so
I'm,
basically
on
a
constant
quest
to
make
this
easier
to
run
because
I
want
journalists
to
use
this,
and
journalists
aren't
necessarily
python
programmers.
So
right
now,
I've
got
this
sets
itself.
You
can
install
desktop
that
this
web
app.
This
desktop
application
dates
that
light
which
you
can
use
in
a
web.
Browser
and
I've
been
working
on
data
set
Cloud,
which
is
a
paid
hosted,
SAS
version.
A
A
There
was
a
lot
of
information,
but
I'm
like
geeking
out
about
all
of
it.
I'll
highlight
some
comments
of
people
also
geeking
out
about
it,
and
then
I'll
highlight
some
questions
that
they
had.
So
someone
said
this
is
so
interesting.
This
is
amazing.
Oh
my
gosh,
this
is
fascinating.
Someone
said
you're
a
more
experienced
person,
it
seems
I,
don't
know
if
you
missed
it,
but
he
did
say
he
has
20
years
of
Open
Source
experience
and
he
like
created
Django,
but
then
in
terms
of
questions.
A
I
actually
had
a
similar
question
to
this
person.
They
said
in
terms
of
making
data
set
more
accessible.
What
are
your
thoughts
on
piodide
versus
depth,
containers
and
code
spaces?
I
just
happened
to
find
Simon's
death
container.json,
while
working
on
a
sequel
like
DC
today,
I
actually
was
wondering
that
too
I
was
like
what
about
our
code.
Spaces
enabled
yes,
okay,.
B
So
I
have
a
tutorial
on
the
dataset
website
and
they
said
I
O,
slash
tutorials,
which
talks
about
using
all
code
spaces,
because
code
spaces
is
just
perfect
for
this
kind
of
thing.
My
favorite
thing
about
code
spaces
is
I
like
running
tutorials
and
workshops,
and
the
worst
thing
about
a
workshop
is
the
first
half
hour
when
everyone
in
the
room
has
to
get
a
python
development
about
working
and
it's
a
nightmare.
It's
an
app
because
somebody
there
are
so
many
ways.
B
It's
incredible
and
so
I've
switched
all
of
my
workspot
shops
in
the
past
year
over
to
using
Code
spaces,
because,
firstly,
every
other
room
is
on
the
equal
playing
field.
They
click
a
link
and
they
get
an
environment
and
then
the
best
part
is
if
they
break
their
environment,
they
can
get
a
new
one.
So
you
you
can't
break
it,
which
I
get
that
to
me
is
the
that's.
The
the
perfect
thing
about
the
thing
for
development
environments
is
just
completely
removing
that
ability
to
break
things.
B
The
one
catch
with
code
spaces
is
that
you
can
run
data
set
in
it
and
it
works
for
you.
But
you
can't
share
that
with
other
people
that
code
spaces
only
allows
you
with
like
log
in
with
your
answer
to
view
that
application.
So,
honestly,
that's
the
only
annotation,
so
it's
a
rate
for
personal
usage
and
for
learning
how
to
use
it.
Things
like
that.
The
one
thing
it
doesn't
solve
is
is
publishing
to
other
people,
and
there
are
lots
of
ways
to
publish
dataset
has
a
built-in
command
on
the
command
line
version.
B
Where
you
can
say,
data
set,
publish
the
cell
name
of
sqlite
database.db
and
you
hit
enter
and
it
will
package
it
up
and
it
will
publish
it
to
the
cell
or
to
Cloud
run
or
to
Roku
or
to
fly
and
that's
a
plug-in
thing
as
well.
So
you
can
build
plugins
for
extra
publishing
platforms,
and
that
was
the
one
of
the
original
ideas
was
okay.
What
if
you
could
have
an
application
that
could
deploy
itself
that
can
that
knows
how
to
itself
up
and
put
it
online?
B
So
people
don't
have
to
learn
how
to
deploy
to
their
cell
without
going,
and
that's
a
really
important
feature.
You
can
do
that
from
code
spaces.
So
I
think
around
the
code.
Spaces
got
myself
to
a
point
where
it
works
and
I
published.
Two
Versailles
faces:
that's
kind
of
cool.
You
know
that
that
totally
works
but
yeah,
so
Dev
containers
I've
not
got
my
head
around
yet
I've
been
trying
to
understand
I.
As
far
as
I
can
tell
that
it's
it's
another.
It's
it's
built
on
top
of
webassembly.
B
So
it's
it's
a
much
more
sort
of
Rich
webassembly
based
environment
I
want
to
learn
more
about
those
and
figure
out.
If
they
can
work.
Higher
dyed
is
extraordinary,
like
data
set,
light
took
me
a
weekend
to
build,
and
that's
because
iodide
just
sold
all
of
these
problems
for
you.
It's
astonishingly,
well
engineered
piece
of
software
and
I
I
love
that
and
I
also
love
that
it's
it's
another
version
of
a
sort
of
safe
environment.
You
know,
if
you
install
software
on
your
computer,
you
might
break
your
computer.
It's
running
in
iodide.
B
A
B
B
And
I
think
that's
really
exciting.
I
think
the
thing
that
angers
me
most
about
being
a
software
developer
is
how
hard
this
stuff
is
like
the
learning
curve.
I'm
just
getting
started
with
high
school
or
JavaScript
is
so
steams
and
the
number
of
people
who
don't
make
it
up
have
three
months
of
misery
and
hence
never
become
programmers,
and
if
we've
managed
to
get
them
past
our
three
months,
they'd
be
doing
amazing
things.
Right
now
is
that
so
anything
we
can
do
to
reduce
that
friction.
I
think
is
super
valuable.
A
Yeah
I
completely
agree
with
you,
oh
I'm,
also
echoing
a
little
bit
but
I
completely
agree
what
you
want
everything
you
said
from
you
when
you,
when
you
brought
up
the
code
spaces
and
we're
teaching
workshops
that
made
me
really
excited,
because
when
GitHub
first
started
talking
about
code
faces-
and
they
were
like-
oh
it's
like
good
for
like
if
you
you
want
to
reduce
your
onboarding
time
or
whatever,
but
the
main
thing
I
saw
was
I
was
like.
Oh
my
gosh.
A
This
is
going
to
make
teaching
workshops
easier
for
me
because,
like
I,
I,
run
or
I
hope
to
run
a
coding
boot
camp
for
a
woman
who
were
learning
to
code
and
like
I,
relate
with
you,
especially
in
a
remote
like
setting.
It
was
really
hard
to
be
like
make
sure
you're
on
this
repo
make
sure
you're
doing
this
so
like
just
having
them
open
it
up.
A
All
in
one
is
great,
I,
don't
know
much
about
piod
or
webassembly,
but
it
sounds
amazing
and
I
agree
with
you
on
wrapping
your
head
around
the
dev
container
stuff
like
I
am
I,
am
now
like
diving
in
to
like
start
I've,
been
building
out
Dev
containers
just
to
teach
myself
about
like
what
are
the
idiosyncrasies.
What
are
the
like
limitations
and
all
that
we
have
another
question
around
pioda
I,
don't
fully
understand,
but
they
said
yay
pile
died.
Are
you
doing
everything
in
the
web
worker.
B
Yes,
I
am
I,
wrote
this
up
if
you
I,
think
date
set
light,
there's
a
link
to
the
GitHub
repository
which
links
to
some
blog
entries
where
I
write
about
how
all
of
it
works
but
yeah,
it's
it's
done
in
a
web
worker.
It's
a
few
hundred
lines
of
JavaScript.
It's
the
whole
thing
like
it's,
astonishingly
lightweight
in
terms
of
the
the
code
that's
used
to
get
this
working
like
it
really
surprised
me
how
easy
that
was
I've
switched
to
a
different
microphone
by
the
way.
So
let
me
know
no.
B
A
Sometimes
I
have
a
I,
have
a
Blue
Yeti
too
sometimes
I
don't
know
if
now,
but
it
gives
me
problems
too,
like
sometimes
it's
a
little
echoey.
Another
question:
I
didn't
fully
read
it
yet:
hey
Scott
I
think
they
mean
Simon,
hey
Simon.
What
have
you
done?
What
you've
done
here?
Super
cool
and
your
dedication
is
great
to
see.
I
was
just
wondering
as
someone
that's
used,
shiny
r,
a
couple
of
times
for
similar
web
apps
to
share
with
specific
people.
What
would
the
advantages
of
switching
over
to
data
set
be
I.
B
B
Once
you
get
past
that
it,
you
need
a
pretty
expensive
like
server
to
to
serve
something
up,
that's
larger
than
two
gigabytes,
so
my
my
sort
of
rule
of
thumb
is
if
it's
too
big
for,
if
it's
not
like,
if
it's
more
than
a
few
more
than
a
few
hundred
rows,
it
can
be
good.
Data
set
light
is
free
because
it's
running
on
GitHub
Pages.
B
Actually,
so,
if
you
just
want
to
share
data
up
to
sort
of
10
000
rows,
you
can
stick
a
CSV
file
and
adjust
and
you
can
use
data
set
light
and
there's
no
installation
and
it
won't
cost
you
anything
so
I
feel
like
that's
sort
of
almost
a
no-brainer
for
a
whole
bunch
of
data
sharing
things.
But
yeah
I
can't
really
talk
to
differences
between
data
set
and
shiny
are
unfortunately.
A
A
B
A
B
A
Cool
yeah
I
have
a
blog
post
on
it,
so
I
can
I.
Can
I
can
share
that
with
you,
someone
asked:
is
it
possible
to?
Oh,
that's,
not,
maybe
necessarily
related,
sorry
I
think
they
they
also
had
asked
earlier,
which
is
why
I
highlighted
them.
They
wanted
to
know
like
when
you
were
showing
those
pictures
of
Obama
and
stable
diffusion
pictures.
Can,
someone
like
add
original
photos
to
the
database.
B
Don't
know
in
my
case
so
that
demo
there
that's
the
the
fixed
set
of
training
data
that
was
used
for
stable
diffusion.
So
this
is
an
interesting
thing
with
dataset
generally
is
originally.
It
was
all
about
read-only
data,
so
you
can
publish
data,
but
if
you
want
people
to
modify
that
data,
you
should
use
a
different
tool.
That's
changed
over
the
past
few
years,
I've
started
adding
plugins
that
can
let
people
upload
upload
additional
CSV
data
or
edit
or
run
like
SQL
queries
that
update
things
and
so
forth.
B
So
over
time
data
it
becoming
more
of
an
interactive
application,
but
that
the
hosting
becomes
more
expensive
then
because
all
of
these
cheap
hosting
providers,
the
one
thing
they
don't
give.
You
is
a
disk
drive
that
you
can
write
to
you
know
Heroku
and
Versa
and
Cloud
Runner
will
read.
Only
fly.io
does
give
you
this
disk
volumes
and
so
I've.
That's
sort
of
my
my
hosting
choice.
Right
now
for
data
instances,
they're
going
to
write
to
a
databases
fly,
but
you
also
still
have
to
think
about
backups
and
all
of
that,
so
it's
it's.
A
A
B
Okay,
so
my
number
one
motivation
is
completely
selfish
right
in
that
I
never
want
to
have
to
say
solve
the
same
problem
twice
like
in
the
years
before
open
source
I
would
find
myself
I'd
build
like
a
little
like
article,
editing
interface
for
a
company
and
then
I
go
and
work
for
someone
else
and
I
build
the
same
blasted
thing
again
and
again
and
again
and
all
over
the
world.
Everyone
else
was
doing
this
solving
the
same
problems.
B
I
will
never
have
to
write
that
code
ever
again
for
the
rest
of
my
career,
like
no
matter
where
I'm
working
in
the
future
I
can
always
go
back
and
say
well,
15
years
ago,
I
solved
this
stupid
problem
and
I'm
just
going
to
reuse
that,
and
so,
as
a
result,
in
working
on
open
source
myself,
it's
basically
a
way
of
guaranteeing
that
I
won't
have
to
duplicate
my
own
effort,
which
is
I.
Think
a
perfectly.
B
It's
a
purely
selfish
way
of
looking
at
it,
but
it
totally
works
for
me
and
then,
when
everyone
else
joins
in
as
well,
the
the
pro
we
like
I
feel
like
open
source
has
won,
and
it
really
has
one
you
know,
because
we
really
we
don't
have
to
constantly
reinvent
the
wheel
all
the
time.
There
was
a
time
like
15
years
ago.
A
lot
of
companies
was
resisted
open
source.
There
were
so
many
companies
who
would
say
no.
We
do
not
do
that.
B
There
you're
not
allowed
to
use
open
source
software
within
the
bounds
of
this
about
about
10
years
ago,
I
feel
like
that
kind
of
kind
of
died
off
because
of
npm
and
JavaScript
like.
If
you
want
to
build
a
web
page,
you
can't
do
it
with
a
no
open
source
policy.
It
just
doesn't
work
anymore,
but
it
took
a
long
time
to
get
there.
You
know
Microsoft
themselves
used
to
be
used
to
like
be
anti-open
Source.
That's
completely
changed
like
Microsoft.
B
It's
the
perfect
project
for
me,
because
I
have
very
wide
ranging
interests
and
every
single
interest
I
have
in
the
world
can
be
boiled
down
to.
Can
you
build
a
database
of
that
like
and
can
you
can
you
do
something
with
the
data
so
having
a
sort
of
core
project
in
the
middle?
That's
that's.
Data
oriented
is
really
useful.
The
other
thing
I
sort
of
touched
on
this
earlier
data
set
has
this
concept
of
plugins,
so
you
can
install
plugins
that
add
new
features
and
those
plugins
I've
written
about
100
of
them
and
I.
B
Think
there's
120
total
now,
so
the
people
in
the
community
have
been
writing
them
as
well.
The
thing
I
love
about
those
is
I
can
have
a
stupid
idea
for
a
feature
and
I
can
build
it
with
no
risk
to
the
project
itself
like
I
can
play
with
with
things
that
I
think
are
dumb
and
because
I
do
them
as
plugins.
There's
no
harm
cause
like
I,
don't
end
up
having
to
having
code
in
data
set
core,
which
relates
to
some
stupid
feature.
B
That
was
just
an
experiment
for
a
weekend,
but
as
a
result,
as
I
get
more
interested
in
things
like
large
language
models
for
AI
or
geospatial
things
with
maps
and
so
on.
I
can
use
plugins
to
play
with
those
so
I'm
sort
of
building
out
data
set
itself,
but
in
this
kind
of
isolated
way,
where,
if
something's
a
waste
of
time,
if
it,
if
it
ends
up
being
a
bad
idea,
it
doesn't
matter
to
me
so
that
that's
been
really
useful
for
me
as
well.
A
B
A
I
love
the
software
engineering
and
just
true
for
that.
The
fact
that
we
want
to
share
things
and
help
others
to
grow
and
instead
of,
like
you,
said
repeating
ourselves
over
and
over
again,
we
have
one
solution
that
we
all
use.
I
was
I,
I
didn't
really
realize.
That
was
the
reason
that
we
that
people
started
embracing
open
source
more
was
because
of
like
npm
and
stuff
like
that,
but
I
think
that's.
B
A
B
Yes,
that's
the
the
ambition
is
I
want
to
bootstrap.
The
paid
version
of
data
set,
basically
I
feel
like
of
all
of
the
open
source
business
models.
That's
the
one!
That's
that's
been
proven.
The
most
you
know,
WordPress
did
it
with
wordpress.com
gitlab
have
done
it.
So
many
of
these
different
companies
have
have
demonstrated
that
you
can
do
it.
You
can
have
an
open
source
project
and
then
people
pay
you
to
host
it
for
them
because
you're
better
at
hosting
than
anyone
else.
So
that's
my
that's.
B
A
Exciting
for
when
that
comes
out,
okay,
I
actually
wanted
to
Pivot,
like
you
told
me
that
you
wanted
to
talk
about
blogging
a
bit
and
I'm
actually
excited
to
talk
about
that,
because
I
blog
I
try
to
blog
every
week.
The
only
times
I
really
fall
off
is
if
I
I'm
at
a
conference
or
something
so
I'd
love
to
learn
about,
like
the
role
blogging
has
had
in
your
in
your
career
and
all
that.
B
Cool
so
I
started
a
Blog
at
simonwillison.net.
It's
called
Simon
willison's
weblog.
It's
got
a
nice
original
name,
I
started
it
in
2010,
oh,
my
goodness,
2002
I
think
was
it
2002
yeah.
So
it's
it's
coming
up
on
its
21st
birthday,
it's
nearly
old
enough
to
drink
in
this
in
the
United
States
and
back
then
like
2002.
There
weren't
that
many
of
us
blogging
about
like
blogging
was
pretty
new,
is
a
thing
and
blogging
about
web
development.
Just
writing
about
web
development
technology
stuff.
B
There
were
probably
only
a
few
hundred
of
us
doing
it,
which
was
kind
of
cool,
because
you
got
to
know
everyone
else.
You
know
you.
You
ended
up
building
this
network
of
of
just
other
bloggers,
because
this
was
before
Twitter.
This
was
before
Facebook
the.
What
the
the
the
social
network
was,
that
you'd
blog
and
you'd
link
to
each
other's
blogs,
and
we
all
had
RSS
feeds
and
RSS
readers
and
that
kind
of
thing-
and
it
has
had
a
I,
mean
I
feel
like
the
best
time
to
start
blogging
was
21
years
ago.
B
The
second
best
time
to
start
blogging
is
right
now
today,
because
blogging
still
works.
In
fact,
there
was
a
period
so
back
like
20
years
ago.
One
of
the
things
about
blogging
was,
if
you
blogged,
you
would
be
on
the
top
of
the
Google
search
results
for
anything
that
you
wrote
about,
because
Google
rewarded
websites
that
updated
frequently
and
other
people
linked
to
and
that's
what
blogs
were
and
that
that
faded
off
over
time
so
come
sort
of
2010
2011,
just
blogging
about
something
wouldn't
put
you
on
the
front
page
of
Google
anymore.
B
I
have
a
hunch
that
that's
quietly
flipped
back
again,
I
feel
like
so
many
people
switched
to
posting
on
Facebook
and
in
private
groups
and
stuff
that
actually,
if
you
blog
like
it's
2003,
your
stuff
is
actually
going
to
start
ranking
again.
So
you
can
so
that
SEO,
the
sort
of
SEO
juice
is
back
for
blogs
as
far
as
I
can
tell.
B
You
know
there's
a
whole
bunch
of
that
sort
of
initial
candidate
analysis.
Where,
like
does
this
person
like?
Have
they
got
a
clue
about
how
to
do
certain
things
and
if
they've
got
an
article
that
they've
written?
That's
it.
That's
enough.
I
feel
like
it's
the
same
thing
with
having
a
GitHub
like
having
public
GitHub
repositories.
You
don't
need
many.
B
If
you've
got
like
one
project
that
you
did,
that
you
put
on
GitHub
and
in
the
readme
you
put
screenshots,
because
the
code's
probably
not
going
to
work,
no
one's
going
to
read
the
code,
but
if
you've
got
screenshots
of
what
it
did,
then
that
hiring
manager
they
get
a
bounce
phone
they'll
find
that
they'll
skim
through
you
read
me
and
they'll
be
like
okay.
Well,
I
don't
have
to
ask
fizzbuzz.
This
person
knows
how
to
write
code.
That
works.
B
Those
kinds
of
things
so
yeah
so
I
feel
like
there's
a
level
of
blogging,
which
is
very
much
just
write,
a
few
things
and
put
them
out
there,
and
you
can
pretty
much
forget
about
it
and
that
will
still
give
you
that
little
corner
of
the
internet-
that's
yours
and
and
a
good
impression,
and
then
what
I've
been
doing
over
the
past
few
years.
I
got
as
an
habit
of
writing
week.
B
Notes
where,
once
every
two
weeks,
because
weekly
was
slightly
too
much
I
just
write
a
blog
entry
about
what
I've
been
working
on
and
originally
that
was
to
try
and
keep
myself
accountable
on
my
open
source
project,
but
also
it
just
turns
into
it's.
What
like
like
Duolingo,
it's
a
streak
like
now,
I'm
like
well
I've
got
to
do
my
week
notes
because
I've
done
them
for
two
years.
I
don't
want
to
to.
You,
know,
lose
the
streak
of
doing
those
but
yeah.
B
B
What's
the
point
of
writing
something
if
it's
not
a
new
piece
of
information
for
the
world,
which
is
a
trap,
because
if
you
think
that
you're
never
going
to
write
anything
you'll
end
up
with
hundreds
of
draft
posts
that
you
never
published
what
you
want,
what
I
do
instead
I
actually
have
a
separate
blog
for
this.
My
til.s
simonwilliston.net
blog
is
anytime.
B
I,
learn
a
new
thing,
I
jot
down
a
few
notes
like
a
few
paragraphs
or
a
few
sentences
and
I
publish
a
thing
about
what
I
just
learned,
and
this
can
be
stuff
like
I
learned
how
to
do
a
for
Loop
in
Bash
and
the
fact
that
everyone
else
knows
how
to
do
a
full
loop
and
Bash
doesn't
matter,
because
the
expectation
of
til
is
well
I.
Just
learned
this
thing
right,
I've
got
I,
think
400
posts
on
there
now
of
all
sorts
of
little
bits
and
pieces
and
I
don't
care.
B
If
you
know
it
already,
but
if
you
don't
know
it,
it's
useful
for
you,
if
you
stumble
enough
a
Google
search,
maybe
it'll
answer
your
question
and
it's
really
it
takes
me.
Each
of
these
posts
takes
me
10
or
15
minutes
to
write.
So
it's
a
very
low
sort
of
input,
low
effort
way
of
publishing.
So
that's
the
the
one.
The
one
format
I
love
is
Tils.
The
other
format
that
I
think
everyone
should
embrace
is
anytime.
You
you
finish
a
project.
B
I,
keep
on
getting
distracted
and
I
do
lots
of
little
things
and
I've
said
to
myself:
okay,
if
I'm
going
to
spend
like
half
a
day,
building
or
a
day
or
two
days,
building
something
the
cost
is
I
have
to
then
write
about
it
and
put
a
screenshot
of
what
I
built,
and
this
is
a
massive
engine
for
Content
growth,
but
also
I
feel
like.
If
you
don't
do
that,
you
are
basically
throwing
away
90
of
the
value
that
you
created
right.
B
If
you
spent
time
building
something-
and
you
learned
something
you
made
something-
you
should
put
a
screenshot
shot,
you
should
at
least
give
give
yourself
a
chance
of
communicating
what
you
built
and
then
in
five
years
time,
when
somebody's
talking
about
something
you
say:
oh
yeah,
I,
built
that
five
years
ago,
here's
a
link-
and
so
you
end
up
referring
back
to
your
archives,
a
whole
bunch
of
times
so
yeah.
That's
some
my
advice
for
writing.
Okay,
I've
got
one
last
piece
of
writing
advice.
B
The
trick
to
blogging
is,
you
have
to
publish
it
while
you're
still
actively
unhappy
with
what
you've
written,
because
if
you,
if
you
hold
out
until
it's
perfect,
you
will
never
publish
you
will
end
up
with
this
folder
of
drafts
that
you've
never
done
so
I
I
almost
take
pride
in
hitting
publish
on
things
where
I'm
like
it's
not
good
enough.
Yet
it's
not.
It
doesn't
capture
everything
I
wanted
to
say:
I
I'll,
you
know
it's
I,
I'm,
not
happy.
B
I'm,
gonna
publish
it
anyway,
and
then
what
I'll
do
is
I'll
cut
it
out
there
and
then
an
hour
later,
I'll
reread
it
and
I'll
fix
a
few
typos
and
I'll.
You
know
you
can
you
can
make
changes
to
stuff
once
you've
put
it
live,
but
that
has
been
absolutely
crucial
for
increasing
my
publishing
rate,
just
just
accepting
that
if
you
wait
until
it's
perfect,
you'll,
never
publish
and
that
doesn't
like
if
you're
working
on
a
corporate
blog
where
you've
got
you
get
somebody
to
if
you've
got
the
benefit
of
an
editor
fantastic.
B
A
A
Also
your
your
regular
blog
is
simonwillison.net
and
then
also
for
people
that
want
to
check
out
datasets
github.com,
Simon
W,
slash,
dataset,
I,
love
everything.
You
said
about
blogging
and
I
agree
with
you
I
think
like,
like
you
said,
like
people
get
a
little
bit
nervous
to
actually
post
a
blog
because
they're
like
oh,
it
needs
to
be
brand
new
or
I.
I
know,
like
you
know,
software
engineering,
it
kind
of.
Has
this
song
I,
don't
I,
don't
know
what
it
is,
but
I
like
this
like.
A
Oh,
it
has
to
be
like
I've
created
this
new
thing
that
no
one's
ever
thought
of
like
that
Vibe
yeah.
So
it
could,
it
could
be
discouraging,
at
least
for
me.
I
I
had
always
been
a
blogger,
but
not
like
a
tech
blogger,
because
I
was
scared
of
like
putting
that
stuff
out
there,
but
once
I
started
working
at
GitHub,
it
would
I
realized
how
helpful
it
was
for
me
to
be
able
to
blog,
because
one
I
express
my
thoughts
like
way
better
in
writing.
So
then
people
were
like.
A
Oh
okay,
like
you
said
like
this,
is
how
brick
results.
Brain
is
actually
working
when
I
talk
it
I
I
am
like
more
stumbly.
Less
eloquent,
like
my
thoughts,
are
all
over
the
place.
So
getting
your
thoughts
all
out
on
one
place
is
good
one
for
employers
and
other
people
to
be
able
to
network
in
and
connect
with
you
and
then,
like
you,
said
for
for
hiring
as
well
for
people
to
be
like.
Oh,
this
person
actually
does
understand
what
they're
doing
they
understand
the
process.
A
B
Actually,
there's
one
more
thing:
I'll
say
about
writing
I
feel
like
if
you
want
to
be
a
senior
software
engineer
or
a
staff
engineer.
Whatever
writing
is
one
of
the
core
skills.
You
know
clear
communication
is
what
separate
what
is
is
really
what
distinguishes
senior
Engineers
from
from
their
senior
engineers
and
blogging
is
such
a
great
way
of
of
developing
those
writing
skills
I've.
B
Actually,
when
I
work
for
companies,
one
of
the
first
things
I
do
is
I
start
up
a
little
internal
blog,
using
whatever
I
can
find
like
Confluence
has
a
blogging
thing
that
you
can
turn
on
that
kind
of
stuff,
because
then
you
can
do
the
same
thing,
but
for
internal
bits
and
pieces
today,
I
learned
how
to
set
up
a
development
environment,
or
my
team
just
launched
this
project.
Here's
a
screenshot
that
kind
of
stuff
which
is
it's
weird
like
very
few
people,
do
this
and
it
massively
increases
your
influence
within
the
organ.
B
You
know
if
there's,
if,
if,
if
your
team's
projects
get
screenshots
and
a
few
paragraphs
explaining
what
you
built
and
crediting
the
people
who
worked
on
it
and
so
forth,
that's
a
great
way
of
elevating
your
team's
work
and
and
sort
of
helping
helping
people
understand
what
you've
been
doing
and
it's
great
we're
practicing
your
writing,
skills
and
stuff
as
well.
Yeah.
A
B
It's
the
best
kind
of
documentation,
because
if
it's
in
a
blog,
nobody
expects
you
to
maintain
it
like.
If
I
read
a
blog
entry
from
two
years
ago,
telling
me
how
to
install
postgres-
and
it
doesn't
work
today-
I'm
not
like
wow-
that
that
was
misleading
I'm,
like
oh
yeah
sure
years
ago,
whereas
the
moment
stuff's
in
the
official
documentation
you're
making
a
commitment
to
keep
that
up
to
date
in
the
future.
And
so.
A
B
A
Love
that
some
people
say
that
I
had
no
I
love
the
acknowledgment
about
blogging
and
passion
about
sharing
your
progress,
regardless
of
the
outcome,
and
someone
was
like.
So
even
if
you
don't
like
your
work,
you
still
publish
it.
I
actually
I
relate
with
you,
because
okay,
so
GitHub
has
their
official
blog.
But
then
we
also
have
a
dev.2,
Blog
and
I
have
a
tendency
to
just
like
I.
B
Right
corporate
right,
yeah
I
mean
the
key
thing
I
think
is,
and
this
is
true
of
all
creative
Pursuits.
Nobody
else
knows
how
good
the
thing
you
wanted
to
build
was
yeah
like
you've
got
this
idea
of
this
blog
entry.
That
will
be
completely
amazing
and
explain
this
concept
that
people
have
never
seen
before
and
then
you
type
it
up.
B
You're
like
yeah,
it's
kind
of
it
doesn't
fit
that
dream
I
had
in
my
head,
but
you
publish
it
and
nobody
who
reads
it
will
have
the
slightest
idea
that
it's
lacking,
you
know
they're,
comparing
it
to
not
publishing
anything
at
all.
So
you
know
it's!
It's
not
like.
There's
when
I'm
reading
other
people's
blogs,
sometimes
I'll
skim
it
and
I'll
be
like.
Oh,
this
isn't
useful
to
me
and
I
will
close
that
Tab
and
I
will
not
think
anything
worse
of
the
person
who
wrote
that
you
know
it's
just
like
okay.
B
This
wasn't
the
thing
I
needed
to
read
right
now,
but
yeah
I
think
that
and
it's
psychologically
it's
so
important,
because
it's
so
easy
not
to
publish
like
it's.
It's
so
easy
to
to
think.
You
know
what
it's
not
quite
there
yet
and
I've
got
a
folder
with
100
drafts
in
myself.
You
know
I've
not
managed
to
overcome
that
yet,
but
yeah,
the
that's.
The
key
I
think
it's
knowing
that
nobody
else
can
see
the
Perfection
that
was
in
your
head.
A
B
So
this
is
why
I
love
today
I
learned,
because
honestly,
the
thing
is
like
the
the
the
best
kind
of
til
is
one
where
you
try,
you
search
for
something
and
you
couldn't
find
the
answer
straight
away,
and
then
you
spent
like
half
an
hour
poking
around
and
figured
out
how
to
do
that
thing,
because
you
know
that
that's
new
and
that
information.
Nobody
has
posted
the
thing
that
you
were
trying
to
search
for
so
that
you
found
a
little
hole
in
the
internet.
You
found
a
tiny
little
Gap.
B
B
And
see
I'm
on
Mastodon
and
Twitter
and
I
played
around
a
bit
with
blue
sky
and
so
forth,
but
also
mainly,
it's
mainly
I
I
blog
these
days,
I'm
blogging
something
most
days,
and
some
of
those
are
just
little
link
blogs.
You
know
it's
a
link
in
the
description
and
then
once
or
twice
a
week,
I'll
put
up
a
full
a
full
sort
of
post
I've.
Also
I've
got
I've,
got
a
sub
stack
newsletter
now,
which
is
exactly
the
content
of
my
blog,
just
copied
and
pasted
into
the
newsletter.
A
A
B
Sorry,
okay,
so
this
is
my
blog.
This
is
actually
a
Django
application
running
on
Heroku
using
postgres
and
it's
got
a
bunch
of
content
on
it.
This
is
my
newsletter
and
if
we
click
through
to
the
latest
newsletter
entry,
you'll
see
that
it's
got
my
latest
blog
entry
and
then
seven
links
and
three
quotations.
Those
are
things
that
come
in
from
that
sidebar
on
the
right.
So
as
I
scroll
down
here,
there's
my
blog
entry
and
then
link
colon.
This
quote
colon
the
quote
from
somebody
all
of
that
kind
of
stuff.
B
B
Here
we
go,
this
is
my
Django
database
turned
into
sqlite
and
backed
up
and
up
here,
which
means
I
can
do
things
like
run
SQL
query.
So
here's
a
complex,
SQL
query
that
joins
together,
my
blog,
my
bookmarks
and
my
entries
and
my
quotations,
and
it
puts
them
all
in
a
big
list
here
so
like
I,
did
a
bookmark,
a
bookmark,
an
entry,
a
quotation,
etc,
etc,
etc,
and
then
I've
got
this
thing.
B
This
is
an
observable
notebook
and
what
this
does
is
it
composes
that
giant
SQL
query
here
and
then
it
runs
that
SQL
query
against
dataset
by
using
the
data
set
Json
API.
So
a
neat
thing
about
data
set
is
everything
that
you
can
see
in
the
interface
you
can
get
back
out
as
Json
as
well,
including
like
you,
can
pass
it
a
full
SQL
query
and
get
back.
B
So
it's
like
a
an
API
for
running
read-only
SQL
queries,
so
it
pulls
that
content
back
and
then
I
muck
around
with
it
a
little
bit
and
then
I
generate
HTML
and
render
that
on
the
page.
So
this
right
here
is
the
HTML
for
my
newsletter
generated
by
that
SQL
query
and
then
I've
got
a
big
button
at
the
top
that
says:
copy
Rich,
Text
newsletter
to
clipboard,
so
I
click
that
there
we
go
and
then,
if
I
go
to
sub
stack
and
do
new
post
and
hit
paste,
that's
my
newsletter!
B
A
B
All
I
have
to
do
is
add
the
title
and
add
the
subtitle
and
click
Send
and
that's
it
so
yeah.
So
it's
kind
of
fun
this,
because
it's
a
really
fun
example
of
a
whole
bunch
of
moving
parts
right
I've
got
GitHub
action,
I've
got
the
Django
app
for
the
blog
I've
got
GitHub
actions
that
download
that
postgres
database
convert
it
to
sqlite
and
then
post
and
then
deploy
that
with
data
set
here.
B
I've
got
data
set
itself
running
that
SQL
query
I've
got
the
observable
notebook
I've
got
all
of
these
bits
and
pieces
and
it
works.
You
know
it's
and
then,
if
you
want
to
see
more
about
how
that
works,
if
you
search
for
observable
on
my
blog
and
sort
by
date,
I
think
I've
got
automating
my
week
notes
and
new
semi-automating
a
substance
newsletter
with
a
deservable
notebook.
So
this
is
a
write-up
of
the
thing
that
I
just
showed
you
but
yeah.
B
So
that's
really
fun
and
that's
that's
kind
of
the
mentality
I
have
to
software
engineering
around
this
stuff,
because
it's
all
about
it's
kind
of
the
Unix
philosophy
right.
It's
little
tools
that
do
one
thing
and
then
you
pipe
them
all
together,
except
that
you
don't
have
to
use
Unix
pipes
anymore.
You
can
now
use
Json,
API
calls
and
JavaScript
running
in
observable
notebooks
and
I
love
that
the
final
step
of
this
thing
here
is
copy
and
paste
I.
Think
there's
something
almost
a
bit
subversive
about
saying.
A
B
One
has
complained
like
nobody
has
said:
oh
your
newsletter's
wasted
time.
I
already
read
your
blog.
If
somebody,
if
somebody
cares
they'll
just
stop
subscribing
to
the
newsletter,
and
so
instead
I've
got
like
a
1300
subscribers
now,
I
think
who
just
basically
it's
people
who
don't
use
RSS
readers
can
use
the
newsletter
instead,
but
yeah
I
I
found
that
I
I
think
that
that
pattern
of
automating
newsletters
through
copy
and
paste
is
something
that
more
people
should
do.
I.
A
B
At
it,
this
is
another
one
of
those
things
about
blogging,
so
I
started
blogging
about
AI
in
September
last
year.
Really,
when
stable
diffusion
came
out,
I
put
up
a
Blog
interview
where
I
said
that
stable
Fusion
is
a
really
big
deal
because
it
was
the
first
time.
One
of
these
modern
AI
models
was
something
you
could
run
on
your
own
computer,
which
I
found
really
exciting
and
then
I've
been
writing
about
and
then
chat.
Gpt
happened
and
you
know
chat.
A
A
B
It,
though,
even
though
Everything
Has
Changed,
because
of
this
and
so
I
started
blogging
about
AI
related
stuff,
and
it
turns
out
that
you
know
if
you
start
blogging
about
things.
People
start
reaching
out
to
you,
so
I
get
invited
to
dinner
at
weird
Mansions,
full
of
AI
researchers
and
Silicon,
Valley
and
stuff,
which
is
super
fun.
You
know
and
very
odd.
B
That's
just
because
I
blog
right,
I
write
about
AI
people
who
are
doing
work
in
AI.
Read
that
I
get
invited
to
things.
You
know
it's
it's
a
great
example
of
how
much
influence
you
can
have
just
through
writing
about
stuff
like
I'm,
not
an
AI
researcher
I
mean
I
I,
published
the
experiments
that
I'm
doing
with
things,
but
basically
I
I've
kind
of
found
myself
in
the
situation
being
a
bit
of
an
AI
pundit.
You
know,
I
put
out
put
out
I
I
try
and
keep
people
up
to
date
with
it.
B
B
A
B
Yeah
I've,
never
in
my
career,
encountered
film,
where
the
ethics
are
so
non-obvious,
yeah
that
that
there
are
no
easy
answers
to
any
of
this
stuff,
which
is
I
feel
like
honestly,
if
you
have
a
degree
in
philosophy
and
Linguistics
and
so
forth,
you're
in
a
better
position
to
work
about
reason
in
the
space
than
those
of
us
who
have
just
done.
Computer
science,
yeah.
A
No,
that's
a
good
point.
I
would
love
to
hear
philosophers,
perspectives
on
this,
oh
yeah,
so
we're
at
the
59
Mark
I'm
gonna
ask
the
the
wrap-up
questions
and
instead
of
asking
you
what
to
do
in
the
open
source
project
that
you
would
like
to
create
one
day,
because
you
told
me
that
that
it
was
yes
yeah.
Somebody
asked
what
did
you
learn
today,
so
I'm
gonna
ask
that,
after
after
the
other
one,
so
my
first
one
is
what
is
the
first
programming
language
that
you
ever
learned.
B
A
B
I
mean
this
is
so
I
sold
a
startup
a
few
years
ago,
so
I
have
I'm
currently
self-funded
with
a
Runway.
That
is
not
unlimited
at
some
point
in
my
life,
I
need
to
start
earning
money
again,
which
is
one
of
the
reasons
that
I
well
actually
with
data
set
I
want
to
and
I
want
it
to
have
enough
Revenue
to
hire
other
people
to
work
on
it
with
me.
So
it's
not
just
about
covering
my
own
expenses.
I
want
to
like
grow
a
team
around
it,
but
yeah.
So
basically
it's
it's.
B
A
B
Can't
tell
you
what
I
learned
today,
because
I
got
up
to
do
this.
I
should
have
got
a
I
should
have
got
up
earlier,
but
yesterday
I
learned
a
whole
bunch
of
stuff
and
I'm
actually
going
to
write
a
til
about
this
I
learned
about
null,
but
using
null
byte
delimiters
in
command
line
tools,
so
that
you
can
have
one
tool:
output
like
three
chunks
of
text,
with
a
null
byte
delimiter
between
them,
which
you
can
then
run.
B
Three
have
three
other
commands
that
run
against
those
really
obscure,
took
a
lot
of
fiddling
I
was
learning
how
to
use
hex
dump
and
to
Output
slash,
zeros
and
all
of
this
stuff.
But
it
was
useful
and
I'm
going
to
write
that
out.
That
was
I
posted
a
blog
entry
yesterday
about
some
new
tools.
I've
been
building
for
command
line
language
model
use,
so
you
can
like
curl
the
New
York
Times,
pipe
it
through
a
thing
that
strips
HTML
tags
pipe
that
through
a
language
model
and
tell
it
to
summarize
it.
B
A
B
A
A
Right
well,
this
was
really
really
exciting.
I.
Thank
you
so
much
for
like
your
time
and
your
knowledge
sharing
like
even
today,
but
then
overall,
with
your
blog
post
and
and
all
the
open
source
projects
you've
been
putting
out.
I
also
think
the
the
audience
because
they
were
really
like
engaged
and
had
a.
A
It
was
better
than
the
ones
I
wrote.
I
was
like
I'm
ditching
them
for
the
the
questions
and
the
comments,
but
we've
reached
time
y'all.
So,
oh
actually
do
you
have
any
last
things
you
wanted
to
highlight.
I
I'll
highlight
again
your
your
repository
for
data,
set
your
blogs
as
well
yeah.
That's
it!
Okay,
cool
awesome!
Thank
you!
So
much
everyone
for
tuning
in
and
stop
by
next
Friday
for
the
next
open
source,
Friday
bye,
cool
thanks.