Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:13
What a friends will go back.
0:15
This is that She's law. We
0:17
feature the hackers the leaders in
0:19
the innovators drive in the future
0:21
of software. I the Cove Yak
0:23
editor in chief here at She's
0:26
Well Today anti in a Thomas
0:28
Paul me and co founder and
0:30
Ceo of Recast. We discuss what
0:32
it means to be productive on
0:34
a Mac going beyond their free
0:36
tier, the extensions built by the
0:39
community for recast, and the future
0:41
of the recast store other executing.
0:43
On recast a I Chat which
0:45
aims to be a single interface
0:47
too many l a Lamb's and
0:49
how they've gone beyond being an
0:51
application launcher to fall on productivity
0:53
mode with access the ai paving
0:55
the way of their future. A
0:58
massive thank you to our friends
1:00
at Fly.iowa for been the home
1:02
of Change log.com We launch our
1:04
abs near use it because Fly
1:06
transforms containers into microbiome for us
1:08
that run on their heart. when
1:10
thirty plus regions on six continents.
1:13
It is too easy you could
1:15
do to launch. never free at
1:17
fly.i owe. Our
1:28
friends, a Fire Hydrant offer
1:31
modern engineering teams less stress
1:33
from ring to retro for
1:35
and and incident management alerting
1:37
on call and of course
1:39
streamlining every aspect of your
1:41
incident process from web hook
1:44
to alert trigger to notifications.
1:46
two incidents open to retro
1:48
task to meantime two Eggs
1:50
Analytics: everything is inside Fire
1:52
Hydrant For modern engineering teams
1:54
and what you're about. the
1:57
here are. Real reactions from Patriot
1:59
users when things signals from Fire Hydrant
2:01
for the first time paid your duty
2:03
and want to say their he's all
2:06
but they're an evil that we've had
2:08
to maintain. I know all of our
2:10
engineering teams as was myself for interested
2:13
in getting this moving the correct direction
2:15
as right now just managing a maintaining
2:17
our user seats has become problematic that
2:19
so that's that's really good actually the
2:22
services to stick consistent problem for us
2:24
and taints is at covering these sorts
2:26
of ad hoc timeframe says is very
2:29
difficult Ah mean. I putting in
2:31
like either rods and specific dies
2:33
and different do shit is is
2:35
quite onerous on you did the
2:37
most important which is didn't tie
2:39
them together. That's half the problem
2:41
with Tgt right is A at
2:43
all these alerts and that I
2:45
get an incident per alert. and
2:47
generally speaking with go sideways. You
2:49
get lots of alerts because lots
2:52
of things are broken but you
2:54
only have one insistence. Yeah I'm
2:56
super impressed with that because being
2:58
able to find a different. Themes
3:00
is an issue for us because I'm
3:02
like the one of one alert flyers
3:04
for one team and anything like that
3:06
to bounce around and a never doubted
3:08
I wish then means that we have
3:10
time to communication issues because like people
3:12
aren't updates now I in which had
3:14
to be open and honest I went
3:16
and we switched. Our
3:19
cue the next step is to go
3:22
to firehydrant.com/signals assemble the scene and worth
3:24
of problems that a single swivel on
3:26
the chair far hasn't delivers end to
3:29
end incident management. And on
3:31
call learning for a modern
3:33
software seems guess or for
3:35
free Once again, Fire hydrant.com/signals.
4:01
Well some is one of my favorite tools
4:03
to use. his rake has all of you
4:05
know this but or the about that are
4:07
like it that the other really great ask
4:09
you use Rica. I do like the bay.
4:12
it's. A is it also every
4:14
valid that? So yeah I'm quite
4:16
familiar with the to visit get
4:19
old like building and using the
4:21
tool like how does it's how
4:23
do you separates a know avid
4:26
user happy user. To you
4:28
know. Cofounder. Of
4:30
the company leading the team, etc. I
4:33
think I don't separated at that,
4:35
so I think Gerbe. Since.
4:37
Day One: Everybody works
4:39
at Rey cause that's for themselves.
4:42
Who. It's. It's one of our sequence
4:44
sober powers like we build a tour. We
4:46
won a half. So. We can be
4:48
quite a penny it is about it. Was
4:50
a business. And then
4:53
luckily there are enough people like us
4:55
out there. that and so exactly the
4:57
same tool. And. His mother when not
4:59
listening to the other a some and we have
5:01
the community. That. Bombarding also seed
5:03
bagel the time. But. We surrounding
5:06
us with the seed bed and built for
5:08
ourselves so whenever Mr So even though I'm.
5:10
Technically the Ceo of the company, I'm
5:12
still a user and. This. Is how
5:14
I. Will definitely the
5:17
Ceo assist. Okay, Why
5:20
I love recasts. I. A
5:23
for you What I used before these laws
5:25
far before and several others over the ear.
5:27
and there's always been some sort of launcher
5:29
for Mac computers Now Mcelwain was it was
5:31
A was Ten back in the day. Now
5:33
it's my go. As almost said in a
5:35
way. Similar. Operating System
5:37
mean. There's always been some
5:39
sort of once referred I think that's. A
5:42
wonder what makes a launcher. Important.
5:45
For. Any user especially
5:47
like a productive use I
5:49
feel like launchers or the.
5:52
Or for the the few is they're not really
5:54
for the money and me. that makes you sad no
5:56
way. and in the Middle East or one expand
5:58
that to many. but I feel like. Everybody
6:00
wants to use a launcher. Or.
6:03
Feel the need to. They dislike clutter
6:05
their bar down there. whatever that will
6:07
that bar is the bottom of a
6:09
Mac Africa even scarred by doc the.yeah
6:11
okay thing to have that a doc
6:13
this club.with all these icons they're all
6:15
times and he sometimes even like keep
6:17
the dog. Presence. Not
6:20
hidden and which I'd like hide my.obviously I'm
6:22
or a Cassius or sorry. cancers are like
6:24
hide everything they keep it away and the
6:26
closer when of they want to know do
6:28
you think it is that makes this particular
6:30
platform Because I don't think it's very popular
6:32
on windows? I could be wrong. On
6:34
that a Windows user. But. What you
6:36
think makes a launch or. That.
6:39
Rakesh represents Now you have expanded the
6:41
ai and more things of course more
6:43
tooling extensions, but just simply an application
6:46
launcher. What makes that Poplar. On.
6:48
The map platform the air said
6:50
that question for the voice. Been
6:53
around for ever a ride. so.
6:56
I think the oldest one of them is
6:58
called Quicksilver. Haemophilus and progressive
7:00
for many with it love quicksilver he
7:02
adds like the old see basically a
7:04
thing they really one of the first
7:06
on some make popular. And
7:08
then launched by Alfred came around.
7:10
They. Basically. Took this to a
7:12
next level. And. Then x the spotlight
7:14
them afterwards which is by the fact that many
7:16
people now. says. But let them
7:18
afterwards. And I think spotted really sort of
7:21
commercialized the Droid. So when Apple build something
7:23
and it becomes part of though s. A
7:26
sort of like mainstream but then
7:28
I know people that have a
7:30
nice since let alone transport lead
7:32
before. Because. It's like I
7:34
said that know it exists. Like
7:36
it's not really visible on the computer. I
7:39
those know on a little but new conflict.
7:41
But. It's not very albus. But. and
7:43
when you experience for the first time
7:45
it's just really make click like hey
7:47
i wanna launch an app. Iti
7:50
been the up name oppressor turn. Hey.
7:52
I wanna such a file. I do exactly
7:54
the same. Hey. I even can
7:57
do calculations. Hey can I check
7:59
the weather and. Kind of go on this
8:01
wondering what you can do with it and. Off.
8:03
To see use cases it becomes muscle memory
8:05
and a thing it's all about. The muscle
8:07
memory was dose in the face the the.
8:10
F A Merrill. The. Com and go
8:12
as fast as possible. And. The
8:14
wax optimizing for as little time spend in the
8:16
tool which is mary counter intuitive to a lot
8:18
of on the software. But. This
8:21
is icing. What for me? The perfect
8:23
Vondra Dust Like you have something I
8:25
can then and intend in your mind.
8:28
And you want to perform this as quick
8:31
as possible. And this can be as
8:33
little as launching an app, which it probably does
8:35
to like dozens or hundreds of times a day.
8:38
But. It can also be been bigger
8:40
My may be interacting with other third
8:42
party tools. Creating. Some issues
8:44
checking a pull request feathers. And
8:46
the of the force. But. This is
8:49
what befell like hey, this interface. Is
8:51
super powerful because it's so simple. And
8:54
you learn to want this interface once and
8:56
then you can apply to. And.
8:58
Plus awful things and is what you've done
9:00
to cause I I We have expanded to.
9:03
Quick. He's to center my
9:05
windows to maximize certain windows
9:07
of pretty stupid sol. Of
9:10
the snippets. Feature.
9:12
I've got several in there. I've got
9:14
sore cause to certain a folders die
9:17
go to a Director's ago too frequently
9:19
the on my Mac downloads screenshots those
9:21
would like to. I'd college Dolan navigate
9:23
to everything else on my navigating to
9:25
personally. Second, to have my
9:27
finder default to like my most important place
9:29
to go to Some from there on a
9:32
my navigate around visibly two different thoughts on
9:34
like. I just want to go to.
9:36
Downloads are only going to screenshots because I want to
9:38
share this thing quickly. I wanted to go find where
9:40
screenshots is Cetera, I. Think that's
9:42
been fun for me. Like exploring
9:45
unique ways to use a loss
9:47
for obviously applications has been key.
9:49
I haven't done much edition or
9:51
multiplication or any sort of math
9:53
have activities. I don't really
9:55
use your yeah feet return. I am a
9:58
pro user. I mainly keen for pro. Think
10:00
my settings and I did not stay
10:02
for a i am open at some
10:04
point Ai gets better. And. Or
10:06
me got great did that. But I just got
10:08
a chance. you beauty.com. Or whatever the U
10:11
R L is for it and the and go
10:13
there. I don't understand why would want to go
10:15
somewhere else but maybe you can help minister like
10:17
what does the lure? For. Folks expand
10:19
the be on application launch and what
10:21
is it that makes people go. And.
10:24
Do what I've done, which is snippets and
10:26
circuits and a I am et cetera. Yeah,
10:29
So. I started was like we're building the to
10:31
for ourself right? And others like how
10:33
many of those features on Right costs so.
10:36
Pick. The window management. Ah, come on that,
10:38
it's just mentioned. Like. I
10:40
personally remember a few odd case to do
10:42
a window on the less than under the
10:44
ride. But. In some them to have
10:46
his other positions and you don't really remember. Hot
10:48
case I can't remember. How many of
10:51
those. So. That's a lot outweigh the
10:53
second leg. Instead of remembering all this the
10:55
hard case, maybe I can trust fellow this
10:57
Co Monterrey cost nothing to have such them.
10:59
Said. A dad and then disliked turn
11:01
out hey other people like the other
11:04
small because as much more accessible. So.
11:06
We added this. Then. Is like
11:08
oh, we oftentimes open directories an hour
11:10
save for directories a web sites so
11:13
we added quick links like a quick
11:15
way to open something. And. It's
11:17
like a lot of those features that you see
11:19
nowadays. Their base a came out. Of.
11:21
Our own problems. Initially,
11:23
Before we opened up re cost
11:26
without a p I to other
11:28
people extend the functionality. Then.
11:30
It's not just about our own problems, but the
11:32
problems of. Thousands. Of developers
11:34
that are willing to like put in
11:36
the effort to automate those little. Paper.
11:39
Cut throat a day essentially. Putting.
11:41
That's like the first thing that people I guess
11:43
soon as down the stand. The system aside, oh.
11:46
I. Can extend that and can make my own.
11:48
There's. A lot of on presently say some like
11:50
you mentioned something that he use. We.
11:52
Have this some youtube video serious with
11:55
Scott! What's in my Rey cause of
11:57
into few rey cause users. And.
11:59
All of them. Rey cause very differently which I
12:01
think just a fun thing like it's one
12:03
tool. But. He can use it
12:05
and mary very different ways and some
12:08
people discover of features after the been
12:10
used for several months or years which
12:12
is kind of crazy so it's super
12:14
feature rich. And probably do a
12:17
pretty bad shop explaining all this features but
12:19
you also don't wanna obama person like hey
12:21
there's other literal thousand things you can do.
12:24
So. It's like. Almost like a bit
12:26
of a game. Like a video game that bestowed
12:28
doing a few little things. And. Then
12:30
you stumble up across something and it's
12:32
another the school need pig a you
12:34
feature at you. Adopt for you to
12:37
books and he go forward. To. Think
12:39
all of those things planned to come to better. But.
12:42
Coming to a eyes like basic the same
12:45
principle. So. When I
12:47
think about modern software in
12:49
central. When. A word and wanna
12:51
be. Very. Focused. I
12:54
wanna stay most likely in one two or
12:56
where do my job? This can be mighty
12:58
to write code. This. Can be
13:00
no snow, another added or to.
13:03
Write. My content or whatever I'm doing. And.
13:05
Then whenever I need to switch that there
13:08
was a context, which animal and this that
13:10
usually. Oftentimes ends up with like
13:12
you ending up somewhere else. Do you wanna
13:14
go? For example, it's had sympathy. For.
13:16
The new in the browser he is he
13:18
old tab which was twitter. You going
13:21
down that way the rabbit hole has and you
13:23
lose his. It's right. So with right as we
13:25
sort of in person that we wanna keep you
13:27
in the flow. who want to make sure that
13:29
he can do you work. You can
13:31
open rey cause and do as many things out of
13:33
there as possible that it don't need to go somewhere
13:35
else. This. Is what we see Many
13:37
people use reckless they i for. Either.
13:40
Way to open re costs and also quick
13:42
question get the answer carry on. With.
13:44
A do dozens of time during interviews like Samba
13:46
went terms come up that I don't know. It's
13:48
as like. Come. On space what is
13:50
excellent that get the answer on tickle know
13:52
and I'll be more. Or
13:54
people use it during coding and other things wasn't
13:57
they? I sat on the side and can answer
13:59
see. Sort of the assistant where
14:01
they can ask questions asked, a work through
14:03
a problem and half that just sitting right
14:05
next to the to they use do. So.
14:08
We don't need to open up a new tab,
14:10
it's just always with you wherever you work across
14:12
any make up, the case and. L
14:15
of that as you are describing your
14:17
use case and I use it especially
14:19
the are part of it is is
14:21
assigned how do I get their. The.
14:24
To be that the most. Hard.
14:26
Thing I suppose, but you made a pretty easy
14:28
biased. Making. It the com and way
14:30
you get to. The. Launcher Wishes or
14:32
the prompt amateur would you would call
14:34
the main window of recast. But.
14:37
Command Space gets you there and then
14:39
Tag get you to an Ai prompt essentially
14:41
and I'm assuming sort of. Honor
14:43
Any sort of rules were have set a
14:45
different. Lm. as my default
14:48
or if I've got. You. Know
14:50
Turbo. You. Know maximize
14:52
are minimized or whatever the options
14:54
are there. It's pretty
14:57
good that is that close away even
14:59
if is three point five for. Does.
15:01
Everybody gets. Access. Three Point
15:03
Five is that a anybody who goes right
15:05
now and down was recast gets that. They're.
15:08
Now we have like the free
15:10
tier that's. All. Of it's that
15:12
we had before. Which. Was always our they
15:14
want to give away a tool for free. Was.
15:17
They I features we have on their part
15:19
of the pro plan So if you subscribe
15:21
you get their cigarette cloud synchronization a to
15:23
use you get a I. Knew. That
15:26
cost them seems and a few other bits and
15:28
pieces. And. Then. When.
15:30
You wanna have one of those more advanced
15:32
models for farmers who before or cloth three
15:34
all posts. From. And throw faked than
15:36
we have like a high a price t
15:39
of for individuals. More or less to
15:41
base a cover, the costs of Ai and away. From
15:44
yeah. I guess I've been taken
15:46
my free said Cbt three point five
15:48
for granted because I just haven't even
15:50
been using it at all. Really? Exactly.
15:52
A message seems that. Why? Don't
15:54
you get here? Even like I mean, I totally
15:56
understand where the use cases because I'm a user
15:58
obviously. But how
16:01
did you even get to building recast
16:03
self more? The journey to this to
16:05
why this all would get excited about
16:07
have had even get here Yeah definite.
16:10
So. We started bag and twenty
16:12
twenty or so. peter me those
16:14
incentives. We work together at Facebook
16:16
before. And we saw
16:18
lighters. So many tools to have
16:20
to deal everyday with and this
16:22
felt. Somehow. Counter Intuitive
16:25
because who sell the lot of the
16:27
tools that we use? Their. The.
16:29
Not really optimized for the works us
16:31
we have. Often mc want to
16:34
quickly together know analytics or like of feats of
16:36
flak. What's the status of this and for that
16:38
any to. Open a browser and
16:40
we go to the page find the
16:42
right thing and as like artist since
16:45
so cumbersome. And we were always
16:47
really keen about productivity and then at some
16:49
point we saw that. That's figure out
16:51
a way how to make this better. Than. Must
16:53
be one, right? And. This
16:56
was a time when I think
16:58
C'mon pallets became hip again. I
17:00
guess. Developers: we know them for years like
17:02
you know them from supply him. From.
17:05
Be as code nowadays. But. In
17:07
other apps have found really common.
17:09
So apps like Linear and Superhuman.
17:12
And even as it notion back in the days,
17:14
that was enough to first tools that made them
17:16
a bit more. Yeah. Commercialized
17:18
some mainstream. And.
17:20
We sewed like hey, The. Best on would
17:22
probably be when we have this one just
17:24
globally like it sits on top of everything.
17:27
And that we started basically billing the ads
17:29
pretty quickly realized like you need to cover
17:31
all the basic said spotlight does because it
17:34
would only have one of those in the
17:36
phases running on the a Mac. He.
17:38
Will not have. Different. Topic is full
17:40
of launched a different launchers. You want to have
17:42
one that does everything for you. To.
17:45
Basic in the okay, we need to cover
17:47
all the basics of spotlight and then we
17:49
can start building on top of that to
17:51
do the other things. This. Was back
17:53
in the days when we were was see who are
17:55
part of. The. Window Twenty Twenty Bats
17:57
This was basically started the batteries.
18:00
That's built this as he thought it
18:02
was like free the of intense building.
18:04
Peter. And I flew to San Francisco.
18:07
State. In mountains you and was twenty four
18:09
Seven says low pay You wake up at eight
18:11
am. The. New builds until you buzzing
18:13
fall asleep and then he do the same
18:15
thing again. It was voted most
18:17
productive time in my life because he had
18:20
no obligations whatsoever our to. It's really fun.
18:22
And. We then quickly
18:24
realized, hey, They. Were
18:27
where it's heading to. Other engineers of the
18:29
was seen at work. There were all using
18:31
Sera. And. That wanted to
18:33
have a quick a way to access this exact before.
18:35
The reasons I mentioned like they have a quick buck
18:37
report. And. They would need to open a browser.
18:40
Navigate the Tehran. Pressure. British
18:42
you button above we've been derived.
18:45
Is so the first extension that rebuild
18:47
that was sort of outside of the
18:49
classic spotlights go. Was. A sensitive,
18:51
raw. And then we went to the
18:53
internet's we talk to and gave it to them.
18:56
And. Did lie go A sort of a
18:58
manual on board in so where I was hoping
19:00
on a video call with them. Was.
19:02
Walking through. I was
19:04
in progress. Leads cause this was like
19:06
building things very quickly to validate the
19:08
ideas. And. Then we were fixing it
19:11
behind the scenes and on board the next person.
19:13
And so of plugged ourselves to was
19:15
like okay this this works and and
19:17
we saw people. Were. Sticking with it
19:19
and we're using it on a daily basis. And
19:22
then move quickly. Add a few. Engineers
19:24
and companies like a been
19:27
be Lifts. Me, you name it, And
19:29
say okay, there is something here. And add
19:31
sort of gave us the confidence of. Going.
19:34
Off the this. As. Humans are
19:36
quite nice. Interface: Where is. Almost.
19:39
No other company doing that anymore. Ahmed my Be
19:41
gets a bit more popular now with they I
19:43
will People wanna come back to the desktop. But.
19:45
Pure Desktop Software. And.
19:48
This special form factor is not really
19:50
like that many companies going off the
19:52
that right everything sort of moved to
19:54
the web. For. The homeless counterintuitive would
19:56
we do and go in for the
19:58
operating system? Is it. And.
20:01
Even if you look at your manifesto which
20:03
I I do not touch on I suppose
20:05
you believe it in as necessary but. When
20:08
you say we right Native code. I think
20:10
even when you think about building a business
20:12
around this tool, it's also building a business
20:15
around native code that's written for Mac. O
20:17
S it's Not Is is built on top
20:19
of Apple's primitive not. You. Know.
20:22
Some. Sort of web software, which not. It
20:24
is not bad, it's just that you went
20:26
to the need a route to make it
20:28
as fast as possible, use the least my
20:30
resources as possible to be really as Apple
20:32
native as possible without being an Apple. Actual
20:35
application written by Apple? Yeah! and you're also
20:37
competing with Spotlight which is pretty fast and
20:39
does have a lot of seats are find
20:42
myself disabling a lot of it. doesn't really
20:44
need a lot of it. And
20:46
I find myself like, just not
20:48
really. obviously. using. Re has
20:50
not using Spotlight. really it all having
20:53
zero need for it. And. So
20:55
you can. I have doubled resources if you don't
20:57
go in disable. Manually.
20:59
Is in of the some of the
21:02
things that happen for Spotlights now there's
21:04
some. Find Her search featured You gotta
21:06
keep enable because I do find myself
21:08
command effing and in finder and searching
21:10
for certain things and directories. Nuts very
21:12
helpful as the nicer like Spotlight as
21:14
a search for files I don't find
21:16
it very useful. Dot. Vs on
21:18
on the how recast years or so and of
21:20
August. Here. So. I
21:22
think this is what I'm and like down. Since.
21:25
You know they use and treatise different
21:27
that you can say a build a
21:29
tool for people that have a new
21:31
spotlight right? and then you kind of
21:33
bring them on. But. We were in
21:35
the camp like hey. The. People we
21:37
target which is sort of. Again,
21:39
rebuild for ourselves so incentives technical
21:41
people, They. Know spotlight
21:44
right? So we take this as a crime
21:46
but and co pay the Know spotlight. Economy.
21:49
To replace that first and if you
21:51
have Dad's and by. Send. You
21:53
can move forward. There. Is the question?
21:55
How do you replace it like one is like
21:57
you make? They exist in features better. It
22:00
for the typical i ten x better than
22:02
people switch kind of stuff. Or.
22:04
Yet more features and I think we did a
22:06
good mix of both. We. Took.
22:09
The features. That's. What
22:11
I have and were making them. Better.
22:13
For I'll use cases we have like
22:15
a more reliable search for apps. We
22:17
made it like faster than Spotlight and.
22:20
That. Is like sort of the table stakes.
22:23
And. Then we added new features with
22:25
simple ones like window management. Was clear
22:27
what history. With. Other features that
22:29
base again. Many. Of our
22:31
users have you to Lawrence of the For.
22:34
We. Find ourselves like hey, If
22:36
he can use more of the same
22:38
interface afraid cause it becomes basic be.
22:41
Even. Was digging cause who enjoy it more he
22:43
can use it more often. So. We
22:45
added more of those little features over time.
22:48
And. Then it can pile up to something really, really
22:50
big because you have. So. Many things
22:52
to do every day. And we see
22:54
users like wheezing and hundreds of times a day
22:56
like the use it's we have is crazy. Which.
22:59
We also see an hour and of the Dixville. Yeah.
23:03
That's true. Do. As. Opposed to
23:05
mention that analytics. Let's also tent on the
23:07
other manifesto which is privacy. What I thought
23:10
was pretty cool. And just
23:12
because you mentioned illness, why mention
23:14
this? Because I didn't notice until
23:16
I was preparing for this conversations
23:18
or revisit. Your manifesto? What? You.
23:21
Declared. You got four points in the
23:23
say: we respect your privacy, We do the
23:25
business we right, Native code, we grow community
23:27
and so. When I think it's important
23:29
to tongue when analytics is like how are you tracking
23:31
your users because you give this. You.
23:33
Know the large majority of the usefulness
23:36
that is Ray Cast for replacing spoiling
23:38
for example as you mentioned before as
23:40
the default and then some Of course
23:43
with extensions for the default recast the
23:45
south of is non pro non paid.
23:48
Is. Quite liberal in terms of it's use
23:50
like you is free. anybody this listing this
23:52
can go and check it out, download that
23:54
and then again the next question is okay.
23:56
Great Nave get this. Free.
23:59
Tool and Though. The hands of a
24:01
lot of people attempting to adopt this tool
24:03
and potential become a paid user on a
24:05
team or an individual or whatever. But.
24:08
Didn't you think? well, how are you tracking those
24:10
people? So I guess I'll just let you speak
24:12
for yourself versus we read your manifesto. Are.
24:15
You tracking folks? We're.
24:17
So. The. Thing
24:20
that happened to be on board of the first
24:22
people and we had to burgle. I pay any
24:24
to lock in for it is up to use
24:26
it. And. Then head and literally
24:28
I think the first person said like. What?
24:30
Had to any to la fin on like. Well.
24:33
The ballot question like. Technically you
24:35
don't are. There is nothing which requires you
24:37
to look in. Which. We say like.
24:40
Yeah. Forget this feedback that early. the maybe
24:42
should revisit that sorry like stripped away on
24:44
the lock in both it and said hey
24:46
if you can. Use the tool.
24:49
Without. A login that is technically
24:51
required. Let's. Do it like it's the
24:53
least friction the a half. When. We want
24:55
to have people down on himself to than it. So.
24:59
Be removed that so you can still to this
25:01
day it down the drain cause to use the
25:03
app. And then you find yourself in that
25:05
didn't need to look in. We. Only require
25:07
lock him and is technically required. So when
25:09
you don't have people up the end there
25:11
is a question like what can you actually
25:13
track. So. The second question, we ask ourselves,
25:15
what are we actually interested Then. We.
25:18
Very much builds by intuition.
25:21
They. We wanna see and in connect
25:23
with people what you want build. And
25:25
so did that go beyond the most important metric want
25:28
to have who just wanna know how many users use
25:30
or a gloss on a daily basis like as a
25:32
tool to use every day is who wants us know
25:34
how me that act If he was we have. A
25:37
me said like for us not important who
25:39
this person is. If is a
25:42
daily active users. That's all we care about.
25:44
Soured analytics Completely anonymous. We don't know who
25:46
this person is that we track. We.
25:49
Have very little product metrics.
25:51
We. Still had stayed mostly caring about
25:53
daily active users. And the
25:56
now that we also revenue. He.
25:58
Always a have a bit of different maximum
26:00
since they're like was revenue that's like for
26:02
the business which is also an hour manifest
26:04
We wanted to mention early on that business.
26:06
Something. We want to build a business about. One
26:09
who beat France pound around as. Was.
26:11
Revenue you obviously have like other ways
26:14
to measure but then as what really
26:16
matters is like okay. How
26:18
many paid uses you have to this
26:20
day with you Like thus the revenue
26:22
crow. For. All of that we don't need
26:25
to know the individual person. We just want to
26:27
make sure. That. We know the data
26:29
and for the product metrics us most important
26:31
for us to know. How we can
26:33
improve the tool. Like. Do people use
26:35
this feature of yes, know? So
26:38
we remove to feature. Is. It heavily
26:40
used how have a is used? What are the
26:42
top features? We as. So. We have
26:44
those those things and then can work with our
26:46
intuition and was all the feedback we receiving. To.
26:49
X Reform. Our. Roadmap, so
26:51
be not really like a super data
26:53
driven company, at least for now. And
26:56
and rather than one a half like. Side.
26:59
Were touched when that be going in the right direct from
27:01
I'd so I see. Analytics or something
27:03
like. I doing the right things? Are
27:05
you going in the right direction? To.
27:07
The things work exley. And
27:10
and for that we think like. Privacy.
27:13
Is a big topic because we want to make sure.
27:15
That. People can use us and our confidence
27:17
is us using them. How. Do
27:20
you then tattoos are So if you if
27:22
you don't collect a lot of data how
27:24
then do you keep in touch point to.
27:27
Real. People. And What? This?
27:30
As a mess with freeway slick one
27:32
we have an inept. Feedback.
27:34
Mechanism so you type in, send feedback
27:36
and re cost. Since. An email to
27:38
us. We. Answer every email we
27:40
get. Everything we get you will hear from
27:43
us. I'm even though that sometimes is.
27:45
The answer is like yeah Thanks for the feedback
27:47
we are building at. Even that
27:49
is a single valid answer. Cousins
27:51
transparency. That. Is the First
27:53
Way which is heavily use all
27:55
the feedback we collect throughout the
27:57
day and send the Daily Digest.
28:00
In the morning to the entire team so everybody
28:02
can stay on top of things. And
28:04
the interesting piece there. Is.
28:06
That everybody surrounds himself constantly
28:08
was feedback. When the web
28:10
and features that can extract information the
28:13
need for their features. And then
28:15
can bring that into that they the decisions.
28:17
That. Like one loop that be have the other
28:20
one. We have this like we have. Almost.
28:22
Twenty thousand people and the slack
28:24
community Sophie go to write gossip
28:26
com/community You can come and join
28:28
us in the community. We. Basic.
28:30
constantly telling their people are ask
28:33
questions we are entering. It's the
28:35
direct this way to reach out to us so
28:37
you can literally the am the until a team.
28:40
Please Don't Do This. The
28:43
last lady other on his Us Twitter
28:45
so we tweet everything we built. We
28:47
oftentimes tried to share how we build
28:49
something or why we it and trying
28:51
to be transparent. Because. At
28:53
the hard we are build or thrive like me
28:55
do the stuff because we love building the product.
28:58
And. We wanna say why we doing
29:00
certain things. And then also
29:02
people are equally excited about as and
29:04
reach out to us on twitter was questions.
29:07
Opinions. Idea Earth. And
29:09
we love seeing all of that. At the
29:12
moment, there's a lot of sound designers
29:14
poking around with concepts for rey cause
29:16
virus. Which. Is something we started working
29:18
on. So. It's extremely cool to
29:20
see. South. Ram people on
29:22
the internet exploring ideas how right us
29:24
could work. And I was. Which.
29:26
Is something we explore ourselves. but it's like really
29:28
nice to see that. That. People think
29:31
alike while think creatively outside of
29:33
the box. I guess the
29:35
the real nuggets there is is how do
29:37
you you know. How do
29:39
you get the right kind of information
29:42
to lead right? You can have my
29:44
community, You can have analytics, You can
29:46
have more or less data from application.
29:48
This just garner-only and ultimately would you
29:50
wanna know is are you doing the
29:52
right kind of thing. right? How
29:55
do you? How do you feel you're doing
29:57
on that? If you're if you're grading yourself,
29:59
if you're the teacher yourself. So the speaking
30:01
you're giving us have a great honey. So
30:03
you're doing in terms of meeting. What?
30:06
Your. Your. Person is ours.
30:08
I supposed to use you all lead by
30:10
what you want to build but the same
30:12
time you are a business in your growing
30:14
it honey for you're doing on the on
30:16
the skill of building the right thing or
30:18
not. They're. So. Good
30:21
question. I. Think First of all
30:23
you need to define what is good and
30:25
where you wanna go. Right on. The thing
30:27
that's the first the to. Continue.
30:29
To think about. Center. Really was
30:31
like sort of what we build a
30:33
company and the business and everything. We.
30:36
All the say. Wanna. Get more users.
30:38
We wanna make more revenue and broad
30:40
strokes. But. An interesting, less. Revenue.
30:44
Is more the outcome Rights of
30:46
Revenue is actually the very fire.
30:49
Is. It doing well. And even if it
30:51
don't have revenue which we didn't have for long time.
30:54
You crowing User base is.
30:56
The. Outcome: If you do it well. Suited.
30:59
As far as like early on it
31:01
was like clear like save our user
31:03
base growth. We. Doing something
31:05
well because people. Enjoy!
31:07
Using the tool. People. Stick
31:09
with the tool. Ah, people
31:11
recommended to other people all of
31:13
those things. Basic. A leading
31:15
to as you are a user base
31:17
crows. So. I think that's the best indicator
31:20
and is a very. Easy indicated
31:22
of Gatorade so you don't need much
31:24
to get this information. And
31:26
then you can sing about okay when business
31:28
the outcome, What? Actually, variables
31:30
to move this right? There. Is
31:32
one thing you can be more features. Sure. Ah
31:35
the other thing is like you can
31:37
told more are better about it. Which.
31:39
Is I think sometimes. Especially
31:41
was early stage company is undervalued.
31:44
On Because oftentimes it is not
31:46
about having more features, it may
31:48
be presenting it better. Was.
31:51
Building. Like more for prance around to
31:54
product. Which. We I think. Early
31:56
on. sort of picked up. We.
31:58
Always were very. The about. The.
32:01
Sign in general and. And. How we
32:03
talk about the prime than really thought about it as
32:05
well as a brand in a sort of a movement.
32:08
Like. Hey we have here. To like.
32:10
Give. You time back to make you more
32:12
productive to. Keep. You in the slow.
32:15
And so bad results. Oftentimes true
32:17
inspirations outside of tech like looking
32:20
into prance that actually advertise and
32:22
how they do it and how
32:24
to build. Had a
32:26
gated basically get people excited about this.
32:28
Setting. That's one of the things that over
32:30
that the hears like me now four years in
32:32
which kind of crazy when i think back. But
32:36
over the time. As any got
32:38
better and better. We. Always in north at
32:40
the end said oh it does Always something you can
32:42
do better. But. The thing we
32:44
understand more. What? People are
32:46
interests then would be interested in what we
32:48
could add. would we not so good at
32:50
and how we want to distinguish ourselves. And.
32:53
As into the constant evolution like
32:55
to. Figure. Something out and
32:57
then you. Get. Better and
32:59
better and better. I'm a huge
33:01
fan of light compound effect like
33:03
every tree to put out. Every
33:06
release you do. Have Buck
33:08
Sixty do every. Podcast.
33:11
The parents who do. All. Of
33:13
those little things when you looking at that. Isolated.
33:16
Yeah. They don't. Move much. But.
33:18
When you release every two weeks. When.
33:20
He tweets multiple times a week.
33:23
When. You do more of those things. Those.
33:25
Things were to make a difference by this
33:27
and us very very hard to be consistent
33:29
rather than half this momentum. Which.
33:31
Is like often times where we feel. Where.
33:34
We really strike when behalf this woman too
33:36
many seats us like how. The. Entire
33:39
team community, Gets. Really
33:41
passionate excited about it. That's a best times
33:43
but it's so hot getting into the state
33:45
and it's so easy getting out of the
33:47
state and I think we all have experienced
33:49
this in one or the other way. Absolutely
33:51
how big is the team now
33:53
Days like how hard is it
33:56
to keep x number of people
33:58
motivated? Yes, it is is also
34:00
something it seems quite rapidly so
34:02
we. Novices: Dot A Twenty Twenty
34:04
Two The two of us. As
34:06
in, we ended twenty twenty with Five. Then.
34:09
We went up to ten. And
34:11
the nowadays we since this week with
34:13
Twenty seven. So. The team crew
34:15
by quite a lot. Ah
34:17
my! Oftentimes is quite the team as
34:20
one product him like we have basic
34:22
a majority of people engineers and designers.
34:25
And. We. Were very
34:27
highly product driven company so
34:29
we. Basically. All our efforts
34:31
going to products. Since. Reason that
34:34
we have a team called Hype. Which.
34:36
Is like a mix of a marketing and
34:38
crosse team. Said they. The. First Team
34:40
we've found ourselves and said a fracas
34:42
which doesn't actively fill the product. But.
34:45
At Talk about a product. Which. Was
34:47
something which was always a bit of a
34:49
sad responsibility but over time have figured out
34:51
this is so crucial. To the success
34:53
that we wanted to have. That. We
34:56
talk about a product of petrol is
34:58
leading that. You. Might seem and on you
35:00
tube. You. To became a big pile
35:02
of the strategy. And how
35:04
to keep all his people motivated. I
35:07
think it's us like. Go. Forward
35:09
and keep this momentum as soon
35:11
as you move. It
35:14
doesn't matter. How big the company
35:16
is if there is movement. As. Soon
35:18
as the movement slows down and was be
35:20
a company says overseas easier because it's just
35:23
a big a mass to move. But.
35:25
As soon as it slows down, I
35:27
feel like it sometimes use loses the
35:30
culture of like. Oh how quickly
35:32
we can get this out or Tembisa but
35:34
next week. Was like today.
35:36
like. Fixing. A buck right
35:38
off the release than like mitt the hot sex.
35:41
I got of those a little things builders really
35:43
culture of like. Hey, we want to just.
35:45
Keep. Doing keep build and keep moving
35:48
forward and as like you'd need to keep
35:50
the urgency no matter how big the to
35:52
miss a thing as soon as the urgency
35:54
goes away. Everything like becomes
35:57
a bit harder to do. How
35:59
does it. What's your practical
36:01
way is keeping the urgency. Is
36:04
it a must ship by Friday? Is.
36:06
It's like you said, every two weeks
36:08
we're releasing. Like, how do you maintain
36:10
a level of urgency? Yes sir, We
36:13
sip every two weeks. So therefore. We.
36:15
Also want to highlight of seats for every two weeks
36:17
so we have if you look an hour train slug.
36:20
You. See Basically there was a was one
36:22
highlighted feature which is something very visceral.
36:25
And to be. Need. To
36:27
have something every two weeks so there was a
36:29
bit of plenty balls. Were we say okay? We.
36:32
Have often times many projects in parallel.
36:35
With protests and never really know when
36:37
they're finished. cried and so we started
36:40
Brookside to then over time. They.
36:42
Get to the finish line and then with usually.
36:44
Say. Okay, which release can be target?
36:47
So at some point we have sort of
36:49
a target in this release. And. Then
36:51
we work to with stat. And. Then the
36:53
team than best he say a pet are things
36:55
we not shipping and the relays and reduce the
36:57
scope. Of. We make sure to make the caught.
37:00
And so is that. I think he always have
37:02
this thinking like okay, what do we released this
37:04
week? What release in two weeks? What were these
37:07
end of the month. And attitude
37:09
is long enough. everybody on the gets into
37:11
the same thing. That's one of the things
37:13
that we do which works well for us,
37:15
but it needs a bit of coordination like
37:17
every now and then you see like always
37:19
ship. Suspect. And was three weeks
37:21
because like out the thing was bit lower. And.
37:24
That's fine, but I think that's that's one of
37:26
the the things. The. Other thing we
37:28
have this them we're using all right
37:30
cause so we experiment All the features
37:32
ourselves have like nightly builds. For.
37:35
Every day you have a new version of
37:37
for cost so even though some features come
37:39
out later we testing them into only for
37:41
Wilde. Which. Is also like if you
37:44
see something or with coming in and the
37:46
tool everyday it's kinda nice to see like
37:48
odors progress. Wagga say as
37:50
a as a user for a cast I
37:52
don't see that momentum. Okay, and
37:54
I'm a minute negatively. And I mean
37:56
that's like now that you've got to say team. Pitches.
37:59
Don't. The job I love the work is don't
38:01
I think the worth it's happening on you tube
38:03
spot on like it. Whatever you do on their
38:05
keep doing it doubled down especially if it's move
38:07
the needle in you're measuring it and what I
38:09
mean by. I. Don't see it is
38:12
not that I don't see a great tool
38:14
to use is that I'm not seen. Behind
38:16
the scenes or even some sort of middle
38:19
seen like the blog isn't very frequent. I
38:21
don't know where to check as a user
38:23
to pay attention to the to the activity
38:25
of the the focus. As you said on
38:28
prior to focus on comments on the urgency
38:30
you're saying that. But. I
38:32
like to see a for you to bring that
38:34
out into the world and so which. As
38:36
as as a diehard fan over a cast
38:39
and somebody who wants to pay attention to
38:41
the Neil moving, how can I do that?
38:43
In what ways can I do that? Yeah,
38:45
I think that eases. islam. To.
38:47
Twitter like. That's why we most active.
38:50
That's where we're. Always to.
38:52
It's something that's how they put out the that
38:54
same smoked a missile stuff. Was. Quite
38:56
interesting with the events and they're like I
38:58
can see that as sometimes. Too
39:00
much focus into the building where
39:02
some things we. Might be not
39:04
sharing and we basically to dedicated to that.
39:07
Bird. Like the easiest one is like going
39:09
on on twitter. I think that's where. Most.
39:11
of the information gets into. Was.
39:14
Due to would be doing nowadays. We.
39:16
Find ourselves the not always the to release
39:18
something new to talk about. Something like to
39:20
play. Guess who mentioned earlier this? so many
39:22
features in there so there might be things
39:25
that we can just use straight away and
39:27
talk about. and we have already in the
39:29
products so that we don't really need to
39:31
sell something new to talk about it. While
39:47
counselors develop, a week is over. But
39:49
there is so much to cover from
39:51
that week and I'm here to give
39:53
you a round up. So eerie. Go
39:56
there for distributed surplus database the one
39:58
when ga it now suppose. Twenty
40:00
two bytes of data and the
40:02
added new exploiting solutions an inflatable
40:04
hyperdrive, woods accelerator, posts, grass, and
40:07
My Siegel databases also went ga
40:09
you know that must be workers
40:11
people in the offer How much
40:13
that costs? Five dollars? Yeah, Five
40:15
dollars. And how much does a
40:17
costly, completely spielberg database operations using
40:19
Hyperdrive. Zero. Yes, zero
40:22
dollars with choose you can now
40:24
sand and acknowledged messages from any
40:26
East be client the also have
40:28
the ability to. Delays.
40:33
Yes the At A Delays workers
40:35
Alex engine was revised analytics as
40:37
scale footed years which to be
40:40
launched. A brand new a playground.
40:42
let's explore all the hosted miles
40:44
on workers ai which by the
40:46
way also went yeah that's right
40:48
Production gray global ai inference that
40:51
you don't need to deploy all
40:53
available seamlessly and workers or directly
40:55
from a rest easy I call
40:57
The also announced a partnership with
41:00
Hugging Face Signal Quickly deploy an
41:02
app. Using these models and fine
41:04
tunes are here. Yeah, that's right.
41:06
The offer Lore support upload your
41:08
fine tunes from Wrangler and apply
41:10
them to their most popular alums.
41:12
There are so many ways the
41:14
the bill with I'm Using Club
41:16
where it's awesome and Claflin doesn't
41:18
support my son wrong they do
41:20
Now Python workers is here from
41:22
the same command. Using Wrangler you
41:24
could now launch a worker. They
41:26
come fast a P, I, Lane
41:28
chain, non pie and more. Are
41:32
to get event notifications. You can
41:34
get notified now when an object
41:36
is created, changed, or deleted. inhale
41:38
that event in your worker. And
41:41
who says you can't spell Sdk
41:43
without Sdk? Craig did and that
41:45
means they have Nu Se Casey
41:47
to use Postscript Go in Python.
41:51
And that is just a few
41:53
announcement from classless. Avila for week.
41:55
Take it all out for yourself
41:57
at Claflin I Com/develop a week.
42:00
Once again, claflin.com flash developer. So
42:16
what I did notice is that you have
42:18
you very nice teams long and I think
42:20
that's a linear popularized beautiful scene was like
42:22
this or think you may have follow their
42:25
lead as a good friends there at linear
42:27
and are so I guess that's one way
42:29
to it is is what's on Twitter. Now
42:32
I'm not against twitter actually acts by any
42:34
means but I'm not on there anymore like
42:36
I was before. To follow
42:38
a brands. Updates.
42:41
List to say. And. So I
42:43
feel like what you've done you with
42:45
the cheese on is pretty cool. However
42:47
I gotta go to it's vs it.
42:49
Come the me or for it defy
42:51
me. To. Tell me all the cool things are doing
42:54
so that's that's where I want to fund the Gap. Would you think
42:56
about that? Yeah, node is a good
42:58
good point. I think that's harder and harder. Read:
43:01
Sing People nowadays because it's so
43:03
pressure and that rights bill humans
43:05
like twitter or acts. Are
43:07
smarter the on threats whatever for everybody.
43:10
Had sex something different. We've been on
43:12
and on acts a twitter since the
43:14
beginning so we have. Quite.
43:16
As of sense of followers up there. And
43:18
this is like sort of still our would
43:21
star a t where we communicate with people
43:23
but we try to fragment that a bit
43:25
so we pranced out to you tube as
43:27
he chatted about this is like we talk
43:29
about releases but also more about other content
43:32
says like about features that we have well
43:34
liked how to use rey cause in to
43:36
see also people how they it was right
43:38
cause it's. Not. Just about
43:40
the product announcements. We.
43:43
Try to remember our block at the moment, even
43:45
so, that's a sin way. Also, need to go
43:47
with hours as this kind of stuff it's a
43:49
bit easier for people who stand top off. On
43:52
and will be doing now is also
43:54
be bambee how email newsletter this one
43:56
this i quite interesting. As
43:59
in response from the gwen. There we don't wanna
44:01
spam people right? Nobody likes of full
44:03
in books. So. We're doing this just one
44:05
of the month. And. It's more like a
44:07
recap like hey, what happened at Ray clustering
44:10
mom. Who. Mention what were released,
44:12
Who. Medicine you people join mention anything
44:14
else that popped up. And
44:17
Id and website this fun category what's
44:19
called team picks so it's just like
44:21
everybody and a team like sometimes experience
44:23
of the new the made by something
44:25
I do different than. So. We
44:28
just sat from our personal lives. essentially
44:30
things like. Sometimes. Error: Which
44:32
board game somebody played recently of stuff
44:34
that made it a bit more human
44:36
than not just about eight years old
44:38
and new stuff. And you can
44:41
become but whoring? This. Might be
44:43
a challenge or thing for everybody is how do
44:45
you have an active. Connection.
44:48
Right to your audience. Who wants to
44:50
lurk slice? Listen frequently. And.
44:52
I think that and act of blog is more
44:54
way. I. Think an active blog
44:56
that isn't robotic like a brand blog, but
44:58
more of like a branded. Person.
45:01
All natured written blog. You know it's
45:03
written by a person is not. Recast.
45:06
The brand and it's just like something that's
45:08
a bit more connected. And. Then obviously
45:10
you know I think for me as this is
45:13
gonna go one way or deeper but I just
45:15
went to Twitter Now it's actually called the Twitter
45:17
app on Mac O S. All the enough is
45:19
not the X F as the twitter app. So.
45:22
That's why this Constance ebb and flow
45:24
of what does it actually called continues
45:26
because nobody gets it. But why go
45:28
to my note of this: Were driving
45:31
away from Twitter Sly Sex. Is
45:33
that when I go to my notifications and
45:35
this is not a Thomas problem or a
45:38
Rake has problem this is a I think
45:40
everybody problem. My notifications are no her mind
45:42
like the first one in here is Elon
45:44
Musk posted. Something. In
45:46
my notifications and then somebody name Armand
45:48
who I follow. I'm sure maybe I
45:50
don't follow me double checked luggage these
45:52
things that are not even and even
45:54
fall. this person through my notifications with
45:56
is like Star Thing and Dax and
45:58
Risks and and. Hell, he Hightower some
46:01
had to pay attention to, but not
46:03
like this. The Prime Again, Christian Lampre
46:05
like other people are. Like all my
46:07
notifications, I find that Twitter Now is
46:09
so. Noisy. Like.
46:11
I want a I want to keep connect
46:13
with people in the high signal look low
46:15
noise. Ratio. The way and
46:18
there really is no right way to do
46:20
that. You can publish newsletter. Maybe that. I
46:22
see nothing though, because like all. In.
46:24
A Maybe you can discomfort your. Your.
46:27
Team Lot a newsletter were blog post that's
46:29
more ways to sort of tap into people.
46:32
As. A What I. Was. Oom out
46:34
though and and suggests or
46:36
desire. Especially. As somebody
46:38
who run something and somebody who consumes
46:40
things is as I believe in the
46:42
hub and spoke model. We're. Sort
46:44
of hanging out with your. Best. Content
46:47
basically to said on a spoke,
46:49
not a hub. Sunita. Hub
46:51
as recast, ran recast orchestrated he plays
46:53
Reagan. Go and check out and as
46:56
browse or subscribe to any push that
46:58
stuff out the everywhere else as necessary.
47:00
I think you need to go back
47:02
at Hub Model. We. Have spoken
47:05
to hub and right now you're
47:07
treating Twitter. Like. The mean place
47:09
but not everybody's there's you're missing out.
47:11
You're missing out. The me entertained and
47:14
least that serves. Yeah, I think there's
47:16
also a single center. A problem with
47:18
some. Other social media bids because
47:20
it's marry sought lift. So when you put.
47:22
Out, let's say a tweet today. It's.
47:25
Forgotten about tomorrow, right? So.
47:27
This. Is why for combo experiment was
47:29
here too because as either lox much
47:32
slower. Lifetime. And that
47:34
were similar. There are proposals and stuff.
47:36
they can be read. Surveys that can
47:38
be found again. There's. A
47:40
bit different think there's like, etc
47:42
content saints. Quite. A lot.
47:45
Also. Like the style of Clinton
47:47
on video is much more prominent nowadays
47:49
fits when years as they have like
47:51
a visual prend quite nicely because f
47:53
something to so and talking about the
47:56
thing this also interesting when a real.
47:59
Thing. About a different. Well. What
48:01
medium do they used to showcase himself
48:03
on? And be quite
48:05
different I don't imagine will solve your keeping
48:08
connected. Prob me on the podcast I do
48:10
want to mention because. I
48:12
just I do wanna pay attention and
48:14
I think what? What? Is
48:16
on your cheese on it is
48:18
mostly. When. I want to hear.
48:21
But. It doesn't feel like a guide
48:23
through the bi weekly caden so
48:25
to speak. Of this momentum
48:27
harkening back to you know the inception
48:30
of this part of the conversation as
48:32
you know how you keep a team
48:34
of eggs most motivated, the urgency in
48:36
your time i really cycles of every
48:39
two weeks and as a daily active
48:41
user over a cast in a pain
48:43
user. I. Just mentioned, I don't see
48:45
it in I'd like to see. I like to find a
48:47
way to see it. In a
48:49
way that doesn't seem my app experience in
48:51
others. One application called Warped that I love.
48:55
And I'll have you know you if you use
48:57
the terminal application called work on Mac O S.
48:59
I love that application. I'm a big fan of
49:01
them and their team. I'm not
49:03
a fan of how loud it is. When
49:05
they have updates, it's very loud. It's it.
49:07
I did and my turn to be simple
49:10
minimalist. don't bother me, you can tell me.
49:12
But that as an example of. What?
49:14
I would consider way too loud of telling
49:16
me about your updates like I don't need
49:19
to know every tiny bug fix. our know
49:21
about the important things not the you think
49:23
it's important things. first this doesn't doesn't seem
49:26
right as soon as it's ready to away
49:28
the work in the don't want to be
49:30
disrupted by any updates when it's right for
49:32
that reason x The. Build. Our
49:34
own app above data that works in
49:37
the background and says like pincer hey,
49:39
here's something new you might be interested,
49:41
but it all happens automatically. An
49:43
ad. It's all pretty much in the background.
49:45
They don't need a terrible that. Will
49:48
give the single egg if it's a tool that he is. Constantly.
49:51
Every day. You just wanna
49:53
be bothered updating. I suspect a thing
49:55
of the past and of in a
49:57
way subtle nudges. I like subtle noses.
50:00
As an Asian user, Of. Change
50:02
Like I don't mind scenes. I don't
50:04
mind even knowing there's not date. I
50:06
believe A mind. Pausing.
50:08
My work flow if it makes sense than
50:10
the do it. But. Tell me
50:12
in a subtle way a color change
50:15
a a small thing in the interface
50:17
that that is subtly different than it
50:19
had been. Mom's. Beforehand as modify
50:21
to mention the email feature the
50:23
send feedback feature. And
50:25
out while you're here. Talk and I
50:28
just quietly. Sent. An email off
50:30
to your team and says he. Ah,
50:32
And Pakistan real Thomas on the cheese
50:34
like eve of the group. Great work.
50:36
And then Daniel and let's raise his
50:38
last name said Carol? maybe. Ah
50:40
and Remains or a cast response A
50:42
How To Thomas for Me And by
50:44
the way, when you're done podcasting, please
50:47
run the check for updates Command to
50:49
get recasts latest version: One Point Seven
50:51
One.one. And. The and so I
50:53
did that I don't even know there was a check for updates.
50:56
Command and I did that So like I
50:58
had a daily active user. There's a new
51:00
of the every cast and is barely even
51:02
telling me and it's a positive thing is
51:04
I if I've never had a pause and
51:06
say is my recast updated it's just always
51:08
been. The recast I know
51:10
loves I am. And praising you
51:13
for that. Subtle. Way of up in
51:15
your application. That's the way to do it. The we're doing
51:17
is the right way. That's good to
51:19
know those us moink have a it is
51:21
what Daniel Daniel men slid to you. We
51:23
had bought in like an earlier version. One
51:26
point, six and nine. Which. Basically
51:28
pro go app up there are
51:30
so now. We. Have a lot
51:32
of users that are outdated and with sort
51:34
of need to read them somehow of like
51:36
I believe. Runs I fab. That's
51:39
a get on the latest out as so
51:41
yeah, this happened like minutes away. Likely you'd
51:43
figure that out this week than recorded. We.
51:46
And I We spoke about it on Twitter today as
51:48
well. As. One of those things
51:50
that unfortunate happened was assembled copy paste
51:52
around the code which was miss. And
51:55
so now it's as they need to make. sort of
51:57
people get to the latest version. So. Yeah
51:59
that. Sec. Wow that's so
52:01
alive. So prepared Now was assessed.
52:04
So do you have a large population
52:07
of user base that essentially can't be
52:09
talked to cause there's no updates? They
52:11
can get the updates and was happening
52:14
so that's a tricky bit. So basically
52:16
amends like be updating the up in
52:18
the background and it is happens automatically.
52:21
So. Now what it is de Las Vegas
52:23
broken that it busy doesn't check for the
52:25
update. On. One point, Six
52:28
Nine. Which. Means like with, they'd allow
52:30
the button one point seven. Seventy.
52:32
Because. We'd know when this ship a new
52:34
version. Fifty percent of the uses up said
52:36
nothing the first twenty four hours or so
52:38
so he extremely have a high adoption of
52:40
like updates. And so. We
52:43
noted that someone like with getting
52:45
feature requests. That. Be has six than in
52:47
the a worse than. As i something seems
52:50
off and then be tactical I go. Like.
52:52
Many people are still on the old worsen
52:54
and so we figured out one one. So.
52:57
Basically, people can update so they need
52:59
to manually run it if you ride
53:01
manually. Then it up. that's the app which
53:03
is good. But. We need to tell
53:05
people and coming back to when we talked about
53:07
before. Privacy. Ride so.
53:10
We. Don't know who are is people
53:12
we don't have a lock in the recon.
53:14
Send them an email and so we try
53:17
to find creative ways. To. Reads as
53:19
many people as we can. So. Quite
53:21
finally mentioning it yourself. Yeah, make
53:23
sure that everybody gets back into
53:25
the auto updating scale. feals get
53:27
that have been spoken in place.
53:29
That way, least minimize your Euro
53:31
disconnection accessing your. Ah, yes, the
53:33
communication is really key, right? So.
53:36
For things to go well but all those things
53:38
don't go well for sure your they want to
53:40
have that in place so do you don't even
53:42
have the email address? have any of users like
53:44
is usually of and have them logging as you
53:46
have you have no information to reach you know.
53:49
The. General free population that uses
53:51
recast. Yes, We have
53:54
like a big email
53:56
newsletter. Which. Arguably
53:58
are a lot of users. That we
54:00
can reach those. Bugs.
54:02
Me: not every user is lock them so not
54:04
every user we know the email of Survivor. I
54:06
don't actually know how on what's the ratio of
54:09
it that would need to check. but yeah so
54:11
this is. I got an interest in talents we
54:13
need to overcome the the next couple of days.
54:16
It's interesting to think about the
54:18
problem said, because hopefully this is
54:20
probably never happens again. But.
54:22
You know in terms of the updating getting broken
54:25
but if it does and you have up the
54:27
frequency you a sense you wanna have. Some.
54:29
Sort of logic in the application says
54:32
I haven't been updated in a beyond
54:34
the normal ratio of of updates or
54:36
frequency of updates. See. Kind of
54:38
want to have in as some sort of second
54:41
as I can might be stale. Yes so the
54:43
net the a few and residents and. Arrows
54:45
happen right sooner or later. Any era
54:47
your medicine will happen right? So.
54:50
That's fine, I thing it's much more important
54:52
having. Mechanisms. In place
54:54
to recover. From. Those things right, And
54:56
this test of of the case of that particular.
54:59
Quite. Nasty because. It's
55:01
as upset as sip the Sept Nine
55:03
recon Three bothered, right? Like with web,
55:05
Okay, you refresh the page non. Western
55:08
of applicants that unfortunate isn't the
55:10
to. So. It was
55:13
quite cool to see our greatest team got like
55:15
out a ways that we can sort of up
55:17
they'd and I've house and me communicate how do
55:19
we read so many people as possible but also
55:21
how do we know the no way people that
55:23
has been updating so finding like good way to
55:25
in dad is that it's a bit of a
55:28
challenge for a threat now but I'm sure we'll
55:30
figure it out. By. Adding and also
55:32
liked us going forward. That.
55:34
Gets an incident finding ways to like recover
55:36
from that in the future. So it's to
55:38
ever happens again that you have a way
55:40
to sell I pay x the at that
55:42
that up they the app an episode of.
55:45
A. Mechanism in place for this to make
55:47
sure. That. This will not
55:49
happen again. This incident. So. It was the
55:51
auto update or code at the Gods. Copy
55:54
paste is essentially that's what happened. The auto
55:56
Update: Here's what's not. So. The
55:58
one saying that you have employees. The like
56:00
know. If. There's an update. You.
56:03
Simply have a bat signal in your application.
56:05
Seeing a a slightly uneven update but that
56:07
best in like it's the lights, broken eggs
56:10
like let his legacy the line of code
56:12
as it always is right? There
56:15
though, things happen unfortunately. But.
56:17
We. Do our best to get everybody back
56:19
on track and that people run at Mandalay.
56:22
That. Is ever that works for that's that's a
56:24
good thing about it. Will Maybe I was running
56:26
one point, six nine it on these be updating
56:28
to one point seven one.yeah maybe that's why you
56:31
are missing the moment or may not have been
56:33
ironed out now. another know. A
56:35
deal of the moments in the and I think in a will
56:37
were camping out on for a little bit. Here I think is.
56:40
Is how do you? How do you lead? Now.
56:43
Is is a team in terms of mental. And
56:46
urgency. But then how do you
56:48
gain the right kind of feedback from. The.
56:50
Police of users that she serves. To.
56:53
Keep building the right car pratt. I
56:56
am in the struggle. Any founder has any
56:58
Ceo as as like are we but in
57:00
the right thing at the right time. In.
57:03
The right quality. Of. Measures that
57:05
we adhere to, etc, Do.
57:07
We have the right amount of people to be to have
57:09
less to have more. Do. Either have a
57:11
in isolated hi team vs. everyone sort of
57:13
has this you know ad hoc marketing job
57:16
and I think you've done it right. And
57:18
in the Java Peters Dornan as a team
57:20
sees you working to me. Ah
57:22
you know so I think we're looking
57:24
at as the different scene was you
57:26
take in as a leader to say
57:29
this is the right direction it seems
57:31
like to me. You've. All
57:33
been really focused on artificial
57:35
intelligence. Different A and Prompts
57:37
even have a thing an
57:39
entire website. As dedicated to
57:41
like prompts of the what is the will
57:43
of the Euro for this thing of thing.
57:45
it's like prompts.radar so. It's
57:47
out s L pop
57:49
Prompts.ray.as oh. And I
57:52
this kind of cool. I am. I
57:54
think there's lots of different prompt engineering.
57:56
Nothing we've seen the way of prompt
57:58
engineers go. where we thought. Initially,
58:00
When a i was becoming hot that. prompt,
58:03
Engineering would become a new jobs hell
58:05
I think that's gone by the wayside,
58:07
basically replaced by things like prompts, Actually,
58:10
I'm I'm sorry, let me try to. It's prompts,
58:12
Plural. Dot Re.sl said to
58:15
clarify that. But. Yeah, so
58:17
Europe and zooming out back in two we
58:19
are focusing on is like it seems like
58:21
you're focused on artificial intelligence? Is that what's
58:23
been giving you the most signal to noise
58:25
ratio in terms of like what the build
58:28
and what you've been personally getting the most
58:30
value out of with recast. Yeah
58:32
so. That's. Always the
58:34
trickiest question like what to build? Try
58:36
it on. Because. There's.
58:39
Almost. An infinite amount of stuff
58:41
you can help. And so I've
58:43
been at Ensnare right? and of
58:45
there is something interesting which I
58:47
still struggle with. but if it
58:49
ensnare. Yeah, it's really. Tight.
58:52
Feedback Loop Be a build something. You.
58:55
Compile. Which. Is the first
58:57
test. And then it passed feature
58:59
at works and that disappeared. Which.
59:02
Is a really good because the feedback
59:04
movies the sense seconds right though the
59:06
really really sort feedback loop. It's.
59:09
A Like a founder or and not necessarily
59:11
it's as the founder. if you work on
59:13
something bigger to seedbed, no business. Second, that
59:15
might be months or even ears depending on
59:17
what you do in. Through. It's
59:20
a it's really really hard to
59:22
know what you build. And
59:24
if that is so right thing to build because
59:26
you'd never really know. Even if
59:28
some segments may be so you do,
59:30
it's it's correct. It. Might be not.
59:33
So. What we always dad's and to steal
59:35
anything and i've seen working for for me
59:37
and my and else and away like we
59:40
were always go and where we got the
59:42
strongest pull from. Where we
59:44
see that people are center interesting
59:46
than something. they keep asking us
59:48
about. Something that's what we're doing.
59:51
And then you see offered them a small on
59:53
board. What pool of sad so. Bad
59:55
that Basic is proud us to I so
59:57
and twenty twenty two right and going into
59:59
the to try to be t came around.
1:00:02
With. Suddenly changed the whole industry.
1:00:04
I think nobody. And I. At
1:00:07
least Matthau hasn't really seen that comment
1:00:09
or even remember. And twenty twenty we
1:00:12
had like machine learning. Start.
1:00:14
Ups in our Y C bachelor thought myself
1:00:16
that oh machine learning added to sing but
1:00:18
been around for ten years but never really
1:00:20
left up to his promos. Let's see where
1:00:22
they are. So now they're replicate so they're
1:00:25
twins I think quite well was all the
1:00:27
the wave of Ai right now with all
1:00:29
the bottles, the host and and so on
1:00:31
before. So. I think
1:00:33
it's like it took. everybody was surprised and
1:00:35
away. And similarly outside we
1:00:38
had plans we wanna do this and
1:00:40
that And and is that out there
1:00:42
are some dissing. And falls
1:00:44
it made really quickly sleep because
1:00:47
everybody was sort of looking for
1:00:49
textbooks. To insert a prompt.
1:00:52
And as a singlet rey cause. It's
1:00:54
a textbook. Plant is available everywhere. So.
1:00:57
That said, like a really really good fit. So
1:00:59
we put one in one together. Quickly.
1:01:01
Ship like a beater. Saw. Like
1:01:04
huge interested the weightless from one day
1:01:06
to another. Which. Gave us confidence
1:01:08
that out there is something that's go
1:01:10
for that. Is basically turned and also
1:01:12
to right cause Pro. Which. We launched
1:01:14
May last year. It's almost a year ago.
1:01:17
And will be launched it. Said.
1:01:20
Was ordered a moment when we launched monitor say
1:01:22
son. And. Launching one of the season
1:01:24
is this really weird, scary thing like we
1:01:26
had for three years look really monetized. We.
1:01:29
Put it out that the pro plan asking
1:01:31
people for pain for. And were
1:01:33
quite surprised by the response because from
1:01:35
day one there was like a huge
1:01:37
interest in it and with like oh
1:01:40
okay. That's. Like what
1:01:42
he hoped for. But. You never read
1:01:44
his think this will happen right? which is
1:01:46
another thing. Like basically we got confidence. That.
1:01:48
Is is like a right direction to go for
1:01:51
us and guts again like so where the pool
1:01:53
is coming from. One went and. And.
1:01:55
So. Over time we bases
1:01:57
okay. There's high demand for that and.
1:02:00
And then we were just working with a community
1:02:02
closer to get a what the actually won a
1:02:04
half from Ray coast and a I. Together.
1:02:07
And Athena to things like one.
1:02:10
They. Want to have similar what we did
1:02:12
miss extensions way have access to so
1:02:14
many tools. Across all the
1:02:16
other things you have almond combined and
1:02:19
in one interface is this at of
1:02:21
want to have the same experience with
1:02:23
A. Because. There's so many
1:02:25
different the I model staff or benefits but
1:02:27
the radio on they use them. The.
1:02:29
Same way with one app for that would
1:02:32
be trusted. We released. And. Profit Complexity
1:02:34
Open A I models that all of in
1:02:36
the same you. I can use them on
1:02:38
the same way. And it'd take the model
1:02:41
he will have. And the so
1:02:43
I think over time what we wanna
1:02:45
of what B C is busy happening.
1:02:48
Like. A I got like
1:02:50
a huge for activity booths. too
1:02:52
many people write on paper. the
1:02:54
his copilot. And all other
1:02:56
things and became really quickly or
1:02:58
this for myself to let I
1:03:00
use regularly. And gives
1:03:02
me a lot of well you're my daily
1:03:05
tasks. So. We really will see.
1:03:07
A. I is really most powerful on
1:03:09
the operating system level, where we operate
1:03:12
on right. Though. We want
1:03:14
to make like I'm Dave. Generally.
1:03:16
I accessible across tools that it knows
1:03:18
about you that you can connect to
1:03:20
your tools. To the greedy sits
1:03:22
there. And. Become sort of This a
1:03:24
system that everybody dreams of, right? Like.
1:03:27
The saying that he consulted Vogue and
1:03:29
you can ask questions you get the
1:03:31
answers can perform actions for you. Really?
1:03:33
Like this interface there is always ready
1:03:35
for you. And is always
1:03:37
aware double crossing advisers, tools and
1:03:40
any well. And
1:03:42
I think like we have like quite a nice position.
1:03:45
As the thing about it that we have. One.
1:03:47
In the south and extensions in the store right
1:03:49
now they connect him to or the different tools
1:03:51
which a thing as a t. To. Really
1:03:53
make this. A reality. We
1:03:56
after perfect interface. It's a for merrill
1:03:58
it's it's until before the. Their applications
1:04:00
and I think nobody really is interested
1:04:02
having and they i checked and all
1:04:04
the different software out there. It's
1:04:07
just wanna have one that knows that.
1:04:09
But. Not have a fragmented as we have
1:04:11
nowadays was apps. Because. Whenever you
1:04:14
switch those apps, does things perform
1:04:16
differently? You. Need to pay five
1:04:18
times for the from saying. So.
1:04:20
Me really see ourselves like hey,
1:04:22
that's consolidate that. And become
1:04:24
bases in the face for it. I.
1:04:26
Think the key thing here is like. Not.
1:04:29
Everything needs to be. I don't
1:04:31
many good things that don't require
1:04:33
a I. Would. You want to
1:04:35
keep bride so not at me now do
1:04:37
with and a hardcore perverted become like for
1:04:39
the I trip and. Thing.
1:04:42
Like. Things like would you mentioned earlier was the
1:04:44
launcher and all of those things that like valid
1:04:46
things right. We need them and I see them
1:04:48
as left the building blocks and and you can
1:04:51
build abstracts of on top of that. Some.
1:04:53
Of them might be a i trim on some of
1:04:55
them, might be not. But. They all fit
1:04:57
on the the same nice and prolong for launcher
1:04:59
in the face of a come palette. The C'mon
1:05:02
bar. and that's where my with
1:05:04
see like to think those two things really
1:05:06
nicely come together is where again we see
1:05:08
like people putting us forward than and we
1:05:10
work with some to terrible make it a
1:05:12
happening. He. Mentioned store. I do
1:05:15
want somebody. I have it. I want a pause
1:05:17
and tongue about the store. And.
1:05:19
I want to talk about. How.
1:05:22
Much or it really want. My question is necessarily believe.
1:05:25
How do you have a store? How do
1:05:27
you support. The. Folks in there,
1:05:29
so I'm I guess my questions of my
1:05:31
concerns come around. Like. Had an audio nurses
1:05:34
this as a founder and leader. Comes.
1:05:36
From my usage with Obsidian, there's lots of
1:05:38
really cool extensions that are built on top
1:05:40
of a city and and in like
1:05:42
is you begin to rely upon these things
1:05:45
the code becomes on maintained, goes stale, is
1:05:47
not first party citizens in terms of a
1:05:49
and extension or a store or whatever you
1:05:51
might wanna call in. A however recast
1:05:54
terms these things to help me on terminology
1:05:56
and he caught the stored you com all
1:05:58
extensions of the caught application. You.
1:06:01
Know how do you support our population
1:06:03
of folks who wanna, you know? Dig
1:06:06
deeper and provide value to their
1:06:08
a cast audience. But. The
1:06:10
same time make for those applications as
1:06:12
your to certain standards. Even as you
1:06:14
mentioned before, the brand of recast efficiency,
1:06:17
productivity, native code speed, all these things
1:06:19
sort of. Are. Synonymous with
1:06:21
the recast brand. How. Do you?
1:06:23
are you leave that? well? How do you have a
1:06:25
store? That. Maintains a level
1:06:28
of. I. Guess high quality.
1:06:30
And. And maintain ability. Say.
1:06:32
You must have been talking about ai it so
1:06:35
much lately solicited up with it. will come back
1:06:37
to adore. I do want to have now I
1:06:39
got sisters or know is there is about this
1:06:41
point won't will come with bay I know a
1:06:43
gets so. When. We started
1:06:46
was thinking about. Than
1:06:48
a P I to extend extensions. At.
1:06:50
The same moment we started thinking about
1:06:53
the store. Because. His see like they
1:06:55
go and hand in hand. We. Really want
1:06:57
to build like an end to end
1:06:59
experience seemed about Iowa's like. People.
1:07:01
Can build apps that and popular said. Are.
1:07:04
The people can discover it and could downloaded.
1:07:06
To does is really what be. Did
1:07:09
our sub swell with build an
1:07:11
Ip? I dad lets people build
1:07:13
native extensions with riding React node.
1:07:16
And then we build a store. That.
1:07:18
Those developers can publish their extensions and they
1:07:21
can be discovered and installed. Was the scene
1:07:23
of plate. By. Our array cause
1:07:25
users. Do we wanna make
1:07:27
the the rated nights like the really
1:07:30
easy eco system. So. I was the first
1:07:32
thing. But. Any was very clear to us
1:07:34
like. Does. Extensions become core part
1:07:36
of what rey cause it's. So.
1:07:39
We wanted to make for a day. Extremely.
1:07:42
Well designed to the have a good you
1:07:44
accept that fast that reliable. And.
1:07:47
Over and over again we came back
1:07:49
to like yeah this basically means and
1:07:51
to do a review process of extensions
1:07:53
and nobody really likes review process of
1:07:55
right there is enough housing, passing on
1:07:57
apple and other of those bad for.
1:08:00
The day and the gate keeping this kind of thing.
1:08:02
So. We thought like okay. What?
1:08:04
Is the best way to review
1:08:07
extension for developers. And
1:08:09
then we exit. Didn't need to go
1:08:11
that far because every developer has review
1:08:13
process of. It's a pull request.
1:08:15
It to upon request. Another developer
1:08:17
comes looks at a code. This is
1:08:19
something up your much you could go.
1:08:22
To. Is is would do. All. Expenses:
1:08:24
Open Source We have a method for Apple
1:08:27
which is good as it comes less ray
1:08:29
costs less expensive. If you
1:08:31
want to add a new extension, you cry
1:08:33
the opponent quest against that. One.
1:08:35
of our team members comes takes a look.
1:08:38
Hopefully. Ever seen a skirt than
1:08:40
tested quickly and then you in the
1:08:42
store? Are. We also use
1:08:44
this opportunity to provide feedback just like
1:08:46
as he would do a pull requests
1:08:49
I paid like we beauty don't. Tend.
1:08:51
To look too much at a coat but rather
1:08:53
on the functionality on the higher level like. Giving.
1:08:56
Feedback on what six you could do better
1:08:58
to like said our you x. And.
1:09:00
An Insipid so this works really well. The
1:09:02
same I guess you contribute to an extension
1:09:05
as a business. Really nice because you just
1:09:07
use the right look it up books. Love
1:09:09
you! Prance. Odds you do
1:09:11
it changes the fuck it, and then you
1:09:13
bring it back into the repository. So.
1:09:16
It also immediately gifts to as
1:09:18
collaboration. With. Dressing this really, really
1:09:20
powerful. So. I think it's like
1:09:22
this thing which which helps us to
1:09:24
really maintain the quality. Though.
1:09:27
There was one thing. If you have
1:09:29
a thousand extended and that's a lot of more you
1:09:31
like we getting. More. Liverpool requests every
1:09:33
day, right? So be going through this. We
1:09:35
have a dedicated team looking into this. By.
1:09:38
Now we have like one person, Tom. Boy.
1:09:40
Had some eggs the goes over some
1:09:42
of that extensions than and even improve
1:09:44
them. I get especially the topic sentence
1:09:46
that everybody has installed. We try to
1:09:49
stay on top of that if terror
1:09:51
to box which bomb users can report.
1:09:53
We. Held fixing them, We. Work
1:09:55
with companies and create us to better to
1:09:57
improve The Simpsons The printing debate as. First.
1:10:00
We. Have other started launching their things and
1:10:03
in our store the really been like
1:10:05
an eco system around that but first
1:10:07
and foremost centered. Around to
1:10:09
simple things. It. Wanna keep quality and
1:10:11
want to make it easy for developers to
1:10:13
contribute? And. Funny enough,
1:10:16
like the pull requests this little
1:10:18
it's threats and said all developers
1:10:20
know for a while works magically
1:10:22
for that. Comes. With a
1:10:24
few try it out the repository as a
1:10:26
huge right now but we try to like
1:10:28
find solutions to to make it easy to.
1:10:31
Keep. Maintaining does extend that
1:10:33
contributed them easily. But.
1:10:35
Nevertheless, this is like something which makes it
1:10:37
really, really powerful. And the funny thing to
1:10:39
wrap this up: his. New
1:10:42
users. Said. On even know
1:10:44
what was a search party extends know what's
1:10:46
an office full extent and anymore. Because.
1:10:48
They're look and feel. Exactly the
1:10:50
same. So. They might
1:10:52
not even know that this is like an open
1:10:55
source Extends we highlight added the store that is
1:10:57
built by people and stuff like that. And.
1:10:59
When they use a day extension. It's
1:11:01
so immersive. Been to direct experience,
1:11:04
That. You have this the ceiling and I
1:11:06
think this is like the really important
1:11:08
bit here. I get the sad ownership
1:11:11
between us as company but all the
1:11:13
contributors. And. Be helping each other to get
1:11:15
to this quality bar. I like
1:11:17
the idea of it being part of
1:11:19
the repository the to have when your
1:11:21
own get out by the that actually
1:11:23
helps you not house. And
1:11:26
you want to have to distribution in terms of
1:11:28
your user base obviously, but you don't have the
1:11:30
code be. Distributed. In
1:11:33
so far as good as already distributed.
1:11:36
Version troll System so barring the awareness
1:11:38
of how get works with. Go.
1:11:40
Above that and say this is all located
1:11:42
at. Recasts. User your
1:11:45
org on get hub in
1:11:47
the extensions repository. In. The
1:11:49
extensions directory within that repository.
1:11:52
And. So all of it is in a way. As.
1:11:55
A High level of control. From a
1:11:57
retest perspective you have a level of
1:11:59
control team. Remove arrow Probably to veto
1:12:01
or to be like this. Probably doesn't fit.
1:12:03
Let's. Not merge this Pull requests. Rikers
1:12:05
is not gonna be you know in
1:12:07
master or in mean. Whatever. You've
1:12:10
chosen if you've gone back and renamed. Ah,
1:12:13
your your primary branch I'm isn't going to
1:12:15
be in that mean branch until you've accepted
1:12:17
the pull request, right? So there's a level
1:12:19
of said in a betting I think that
1:12:22
I wasn't quite aware of. Now.
1:12:24
That makes a lot more sense. You know how
1:12:26
you're maintaining that control. There. Is an
1:12:28
interesting side effect that helps us also
1:12:30
to correct the store. So. Fussy.
1:12:32
There was a extension. That.
1:12:34
Is the same or similar to another,
1:12:36
extends and. We. Can hints the
1:12:39
aws of like hey maybe we
1:12:41
contribute to extra features to the
1:12:43
existing extensions. For. Me use
1:12:45
of respect they have. this is the like
1:12:47
five different accents and in the store you
1:12:49
might not know the differences though. We.
1:12:52
Tried it a dose things as well which is
1:12:54
like less on the extends. Never but warned a
1:12:56
higher levels makes for it like a. Really
1:12:59
nice experience. And. Then the
1:13:01
other side effect that was quite helpful as
1:13:03
like. The Past A Sinner repository.
1:13:06
It's perfect for teaching as well
1:13:08
for agitating. Cause. Every new develop
1:13:11
a com than one know how to build
1:13:13
next and know how to use this feature.
1:13:15
There's no endless of features are like example
1:13:17
the defense Us linked to the hate. Us:
1:13:20
To: Is this react Hope Here is example: Check
1:13:22
it out. The some people take one
1:13:24
extend, some. And that sort of
1:13:26
taking it as a template and then
1:13:28
praying in their own changes for like
1:13:30
a new extension methods like something which.
1:13:33
Leg. Was like we thought about it initially but
1:13:35
was a bit of a surprise. How about sex?
1:13:37
The help. To like spearhead that because
1:13:39
we like to use in and got to like
1:13:41
a thousand extensions. Read it the with the quickly.
1:13:44
Which. Was like. The. Open Source
1:13:46
know it's like nice because people. Really?
1:13:49
Associate themselves with struggles, community,
1:13:52
And like praying and like this does
1:13:54
openness. And like the developer line said
1:13:56
around that. Lot. Of activity here
1:13:58
for this for sure. Instead,
1:14:00
you know, maybe one more layer
1:14:02
deeper from a technical level, but
1:14:04
not quite actually technical, but more
1:14:06
like. Technically. Speaking
1:14:08
license wise. The.
1:14:11
Entire repository is Mit license and grants
1:14:13
on being in though not just for
1:14:15
years and free but for his and
1:14:18
Liebreich the in a true permissive nature
1:14:20
of open source. And
1:14:22
I guess that means that every extension
1:14:24
that's part of this says it makes.
1:14:26
Then there is also permissive license open
1:14:28
source software that can be free to
1:14:30
use. However, literally in quotes, open Source
1:14:33
can be used The real definition of
1:14:35
of the source. Yes, Who
1:14:37
We built our Ip I that
1:14:39
allows people to build those extensions.
1:14:42
We. Build a former to bad basically
1:14:44
allows to render those things and rey
1:14:46
cause this is the permanently host photos
1:14:49
extensions on and he wanted to make
1:14:51
sure we have had like this. But.
1:14:53
We also wanted to make sure
1:14:55
that people builders extensions that there
1:14:57
is this concept of feather the
1:14:59
sad owner said bright. As.
1:15:01
You mentioned as part of our repository so
1:15:04
we own what comes in there. Are.
1:15:06
If other people that decide like a i don't
1:15:08
wanna. Sheridan. The Store.
1:15:11
Or maybe one of us distributed outside
1:15:13
so they keep down repository. And. Then
1:15:15
people can like downloaded from there on
1:15:17
added to. Drag. Us from there.
1:15:19
So. I think this is was assessed the
1:15:22
flexibility mom. I think it's a
1:15:24
steak. It's quite interesting. like when when
1:15:26
you work with developers and on steam.
1:15:28
Basic about myself with well, because the
1:15:30
system one of Bush's people are. I
1:15:32
did one A Basically, be honest with
1:15:34
people, I mean ten or a suit.
1:15:36
but I think developers. They. Like
1:15:39
extremely good at like picking outdoor
1:15:41
things and are like logically not
1:15:43
connect together. You don't make yourself
1:15:45
friends with it. so. I.
1:15:48
Think that we which has good citizen when
1:15:50
going back they didn't how we would like
1:15:52
to habit so. I. Think there's no harm
1:15:54
having and like this so we just trying
1:15:56
to be transparent around us topics as well.
1:16:00
Adding it myself about that so I
1:16:02
guess to tell off the the store
1:16:04
story Mrs Bella time slot extension by
1:16:06
the way to browse only episode sydney
1:16:08
half an ip I we do have
1:16:10
as I think it's a behind the
1:16:12
scenes a pianist never been first class
1:16:14
citizen. After that
1:16:16
would said was during a know he'd made it a
1:16:19
priority but. I don't know how
1:16:21
well documented or a pi if
1:16:23
it exists. How it's being
1:16:25
used are consumed set me now they
1:16:27
would be fun. Like having a little.
1:16:30
Come on and rey cause the open Saints
1:16:32
like that com you see all the episodes
1:16:34
nice covers, the new conflict in it opens
1:16:36
it up and. To. Prefer podcast
1:16:38
plans would be really cool. Never.
1:16:40
Be kinda cool I guess if you have like settings
1:16:42
you can say. It's a based
1:16:45
on install applications, spot a fire, apple,
1:16:47
podcasts or whatever and in Asia opener
1:16:49
for it and extension. Preferences.
1:16:52
And. The winner to you, You know? Fresh
1:16:54
teams are gonna know. As the
1:16:56
horses species on of episodes
1:16:58
are in a latest Sees
1:17:00
log podcasts I'm no. To.
1:17:03
Be for new world and would be cool. Isn't
1:17:05
we will be dope says as I would certainly
1:17:08
use that because in I used for a cast
1:17:10
daily and if I guess have one more thing
1:17:12
in there. I I just installed
1:17:14
pills because we're talking. I was like okay
1:17:16
those browsing storm I would see here. where
1:17:18
do I use I use one password. I.
1:17:21
Use tail skill. I use docker.
1:17:23
I. Use that as my editor, you have
1:17:26
enough Visa Dick is that is too
1:17:28
nice which is cool and what I'd
1:17:30
do with as As Nedra in In
1:17:32
Re Cast would make any sense. as
1:17:34
the only was some low hanging fruit that are to use
1:17:36
only won't. Tell. Skill Sweet. What would
1:17:38
I have their let me list my exit node
1:17:40
on the me to find my exit node or
1:17:43
whatever might be or less my status and whatnot
1:17:45
so as chemical. And that's actually
1:17:47
I'm noticing on the store. it's it's contributed.
1:17:50
You know when you go to
1:17:52
recast.coms last tail skills last till
1:17:55
skill. And in a while there's
1:17:57
a double up there. But as you can explain that
1:17:59
is I see. By. The tail
1:18:01
skill user I guess then you have
1:18:03
contributors which I believe is fi pulled
1:18:05
from the get herb a H E
1:18:07
I in the contributors on get hub.
1:18:09
But. You know it's icy eighteen skill.
1:18:12
Org. Or user. I mean, I know,
1:18:15
How they get their name of their tail
1:18:17
skill yeah, that much like excellence and said
1:18:19
similarly, we kind of follow the get a
1:18:21
pattern there. So. What do you
1:18:23
have is the the can have. One. Organization
1:18:26
or user and than when
1:18:28
your contribution extension. We. Basic
1:18:30
append that for similar to a repository
1:18:32
way have he isn't a rebel name
1:18:35
is Maria Username Extent name. And
1:18:37
then be also have. Companies.
1:18:39
That we collaborate with because often times.
1:18:42
When. You install an extent and then it
1:18:44
comes from the of see sold company
1:18:46
or yeah, makes it more trustworthy right? So
1:18:49
that's why be. Collaborated: With
1:18:51
or com is next tail scale of
1:18:53
one password that linear to like the
1:18:55
head of like the official x than
1:18:57
some because then you can sort of
1:18:59
trusted on and it gives access this
1:19:01
extra boost of confidence. To. Makes
1:19:03
up here this the sounds sounds
1:19:05
reasonable. As. I mentioned all except
1:19:07
the open source throats. Henry can always look
1:19:09
up. If you developer what happens in
1:19:11
there we just sort of also security measure to
1:19:14
be know. What? Goes on there and
1:19:16
like a transparent for people to look into. it. But.
1:19:18
Doesn't like something which we find
1:19:20
extremely valuable like having. This
1:19:22
sort of official exception. And.
1:19:25
So she loses in the tank. It suggests
1:19:27
on details you have officers in a you
1:19:29
have author obviously and you have owner and
1:19:31
the the owner in this case for Tesco
1:19:34
is obviously they'll skill. Says
1:19:36
chemical which is kind of interesting like wire
1:19:38
why we have to two different ones lap.
1:19:40
It's a nice birthday is a nice detail.
1:19:43
Like. Oftentimes we have. A
1:19:45
Person Starting Villains Extension Thirty
1:19:47
Author: Yeah. And then
1:19:50
disclose and popularity.
1:19:52
And and we wanna make it unofficial
1:19:55
accents and. And. Still
1:19:57
wanna keep the or term author of.
1:20:00
They put so much effort into it was
1:20:02
to make wanted it. I'm the we have
1:20:04
them still as the original author and data
1:20:06
that get the extra credit. And
1:20:08
that half basic a D d owner
1:20:10
separately as the organization. with
1:20:13
somebody for example. Similarly, the notion
1:20:15
extension. With. Cynically was built
1:20:17
by of friends pm. And
1:20:19
he really build this and really really well. And
1:20:22
then we made it. Like together with Notion
1:20:24
said that hey let's make this an official
1:20:27
extension to so user fight This Really the
1:20:29
official Notion extension. But. Still wanna
1:20:31
keep the credits to the Ireton author so
1:20:33
we have them in there as well. As.
1:20:35
They assess this nice touch us which is like.
1:20:38
As. You build this yourself in one
1:20:40
of still have discredit right? so tested
1:20:42
one a trend for that away artificially
1:20:45
agree with that. I do agree also
1:20:47
that you need to have an owner
1:20:49
mechanism. You. Know, even if it's
1:20:51
the original author, you might have author
1:20:53
Sam. When Bill. An onerous
1:20:55
in limbo until til skill. The Yeah recast
1:20:58
will collaborate with you and take ownership of
1:21:00
that and. Let. Their become part
1:21:02
of our first class citizen extension ecosystem
1:21:04
out there because what you give that
1:21:07
brand is is really. He. Added
1:21:09
user experience and I guess in
1:21:12
some ways more usefulness and potentially
1:21:14
distribution because. I'm a
1:21:16
good your store and and discover something. I
1:21:18
mean who doesn't about one password. Or.
1:21:20
Even tell skill. but. Till. Skill by the
1:21:22
way. as as a sponsor and so is. One.
1:21:24
Pass for the and become upon a sponsor
1:21:26
Really since I'm excited about the Ebony a
1:21:28
life alone user of one passwords I'm excited
1:21:31
our the I'll answer it right here in
1:21:33
the near future but you know we use
1:21:35
all their features in or for structure and
1:21:37
as a developer he's a lot of their
1:21:39
feel I features and on my as these
1:21:41
keys are stored in there I biometric on
1:21:43
the things anyways ad is over for one
1:21:45
pass from a big fan though cool to
1:21:48
close the loop like seven something that you
1:21:50
can use a great for long personally and
1:21:52
adding sponsored by it's like interesting thing as
1:21:54
yes. Love they love helped clear was
1:21:56
developers so. I. Think there's a lot of
1:21:58
your the number one pass or but no up. Who use
1:22:00
their developer toying. Or. Have a
1:22:02
clear awareness in oh that's part of the
1:22:05
key to take part of distribution for brands
1:22:07
and communication is go where the people are
1:22:09
in a was totally makes sense why you're
1:22:11
on Twitter next because that's where a lot
1:22:14
of people are. I think in the last
1:22:16
couple years we've seen that that ecosystem fracture
1:22:18
bed and you have more people who completely
1:22:20
a man that never coming back and have
1:22:23
moved Mastodon or slack or does nothing. There
1:22:25
is a sort of abandoned social media maybe
1:22:27
now the lyric on Instagram instead of. Being
1:22:30
active on a platform which is cool
1:22:32
too. But. As a brand is
1:22:34
trying to maintain communication, loves or folks
1:22:36
and whatnot, it's become in Philly harder
1:22:39
and today's age really is what I
1:22:41
mean. Ride like you can. busy tourism
1:22:43
at his stuff. But. As people
1:22:45
those cowards doesn't matter right? and it's really into
1:22:47
doesn't doesn't matter how big or small companies is
1:22:49
see a half this proper brought in one of
1:22:51
the other way his best and think by way
1:22:54
of and be one a prance out. And
1:22:56
for the new target audiences from wanna. Diversify.
1:22:59
Your teacher said the bet you kinda need
1:23:01
to done. For. The for he learned
1:23:03
his mechanics okay well the new users I
1:23:06
reached. Out to I
1:23:08
can make my brand attracted to them
1:23:10
to the know about the print differently
1:23:12
before super interesting. I feel like this
1:23:14
is like yeah especially. And
1:23:16
develop a land as like a breed interested. States
1:23:20
of America. Was
1:23:37
a friend's This episode is
1:23:39
brought to you by our
1:23:41
friends as Century Sigma as
1:23:43
Century.i O S E N
1:23:45
T R Y.i owe a
1:23:47
Co breaks six it fast
1:23:49
with Century. Don't as observe
1:23:51
take action. The only admiring
1:23:53
platform out there bill for
1:23:55
developers. they just as root
1:23:57
cause for every issue. Ninety.
1:24:00
Thousand plus growing teams use
1:24:02
Century to find their problem
1:24:04
Fast good. Have Disney Vm
1:24:06
were air Table Autodesk monday.com
1:24:08
Miro Tonal. I could just
1:24:10
go on. Literally Ninety Thousand
1:24:12
plus teams trust Century to
1:24:14
fix their problems fast. But
1:24:16
how they do that or
1:24:18
they have air monitoring says
1:24:20
replayed performance monitoring code coverage
1:24:22
All the things it takes
1:24:24
to make an application run
1:24:26
well. Production Century has got
1:24:28
you covered again. Sega. Mouth
1:24:30
as sensory.i oh that's
1:24:32
sentry.io and Macys are
1:24:34
coupon code chains while
1:24:37
to get one hundred
1:24:39
dollars for error monitoring
1:24:41
with Century once again.
1:24:43
S. E N T are
1:24:46
wide Io century.io. And
1:24:48
by our friends at Factor as someone
1:24:51
has always been my i'm full of
1:24:53
when I put my brain and when
1:24:55
I put my body I was so
1:24:57
excited to get It seems to work
1:24:59
with Factor that a fan of Factor
1:25:02
for a while and recently had a
1:25:04
chance to get them on as a
1:25:06
sponsor. Now this is a little Us
1:25:08
center from our normal sponsor style because
1:25:10
a it's about food but I believe
1:25:13
you are only as good as what
1:25:15
you put into your body and that
1:25:17
begins with food eating. Better can be
1:25:19
easier with factors: Delicious ready to
1:25:22
eat meals. Every fresh never frozen
1:25:24
meal is chef crafted dietitian approved
1:25:26
and ready to go in just
1:25:28
two minutes. You'll have over thirty
1:25:30
five different options to choose from
1:25:32
every week, including calorie Smart Protein.
1:25:35
Plus that's mine in Quito. Also,
1:25:37
there are more than sixty add
1:25:39
ons to help you stay filled
1:25:41
up and feeling good all day
1:25:43
long. Now we wait and for
1:25:45
guess it's a day and get
1:25:47
asked year. eating goals to minute
1:25:50
meals pancakes movies know prep
1:25:52
no mass meals flexible of your
1:25:54
schedule factor is the perfect solution
1:25:57
if you're looking for fast premium
1:25:59
off with no cooking required.
1:26:01
And of course, you can sign
1:26:03
up and save. We've done the
1:26:06
math. Factor is less expensive than
1:26:08
takeout and every meal is dietician
1:26:10
approved to be nutritious and delicious.
1:26:13
So head to factormeals.com/change
1:26:15
log 50 and use
1:26:17
our code changelog50 to
1:26:19
get 50%
1:26:22
off. That's right, 50% off.
1:26:24
That code again is
1:26:27
changelog50 and go to
1:26:29
factormeals.com slash
1:26:32
changelog50. That's
1:26:35
factormeals.com slash
1:26:38
changelog50. All
1:26:43
right, now Thomas, it's
1:26:46
time to talk about AI. Are you ready? Are
1:26:48
you ready? I
1:26:57
was excited for you to talk about AI. I mean, I
1:27:00
wonder, it's now become the norm.
1:27:02
And in a lot of cases, there's people who
1:27:05
are fatigued by discussions of
1:27:07
AI and it's become a meme
1:27:09
in lots of ways too where
1:27:11
XYZ brand has a new launch and artificial
1:27:14
intelligence is now baked in in
1:27:16
some way, shape or form. And
1:27:18
then obviously you launched Pro. A
1:27:20
while back, I came for the
1:27:22
setting sinking personally and did
1:27:25
not come for the AI. But now I can
1:27:27
appreciate the AI because I
1:27:29
got 3.5 for free or I guess
1:27:31
for paid for free with my pro plan.
1:27:34
Right? For free with my pro plan. But
1:27:37
then you also have the option to have advanced
1:27:39
AI in Raycast.
1:27:42
This has got to be a boom for
1:27:44
you. One way to have a single
1:27:47
interface, a single pane of glass as a
1:27:49
lot of people in tech say to
1:27:52
AI tooling, the various LLMs. What does
1:27:55
this mean for Raycast? Have you seen
1:27:57
a massive growth? Is there new research
1:27:59
internally? How does this work for you all? Yeah,
1:28:01
especially I get it. I mean, I
1:28:03
think everybody's fatigued about AI, right? So
1:28:05
this has been like around now for
1:28:07
one and a half years. Every
1:28:10
week, there is something new. It
1:28:12
keeps getting better and like
1:28:15
new stuff enables and whatsoever. Every
1:28:17
company indicates it, every company talks
1:28:20
about it. So I
1:28:22
definitely get this. And I think it's also, I
1:28:25
personally wish like this gets a bit like less
1:28:27
about AI, and more about
1:28:29
what's going on. Actually, it's possible or
1:28:31
what actually features people build, where
1:28:34
AI becomes just like the
1:28:36
enabler, right? Like the technology that enables
1:28:38
it. But you talk more about
1:28:41
actually the innovations you do
1:28:43
and less about AI and
1:28:45
model. That being said, we
1:28:48
recently added more of those models, because
1:28:50
we're in this really weird timing
1:28:53
right now, where it's
1:28:55
one of the only things I can remember from
1:28:57
my life where it comes out. And
1:29:00
immediately, everybody knows about
1:29:02
all the details, is an
1:29:04
expert in it, knows when to
1:29:06
use which model knows the preferred model
1:29:09
whatsoever. And it's like it's,
1:29:11
it's kind of hard to go against this, this
1:29:14
flow and so like you completely neglecting it,
1:29:16
which is basically we did our pivot there in
1:29:19
a way of like we first wanted to just
1:29:21
support open AI models, and said like,
1:29:23
deals are the clear winners, making
1:29:25
it easy, we're just going to have those models. But
1:29:28
then loud and clear our community say like,
1:29:30
Hey, hold on. And traffic is
1:29:32
actually really good for x, y, z. And
1:29:35
then there is perplexity that is really great
1:29:37
for when you want to have quick answers
1:29:39
around channel knowledge. And it's like, oh, hold
1:29:41
on. Yeah, this kind of makes sense. And
1:29:44
it's not so black and white, the details
1:29:46
in the devil. And this
1:29:48
is why we sort of added more of
1:29:51
those LMs to Raycos. And we really
1:29:53
see that now it's like Raycos AI
1:29:55
is sort of this one interface. For
1:29:58
many different LMs and you can pick can choose
1:30:00
the ones that suit you
1:30:02
personally the most or suit
1:30:04
the task at hand that you want to do
1:30:06
the most. You might be need like
1:30:08
an extremely fast LLM which is
1:30:11
like less capable but it's good enough for
1:30:13
what you want to do like fixing spelling
1:30:15
of a sentence or something. But
1:30:17
maybe if you're diving down into a coding problem, you
1:30:19
want to make sure that you have the best LLM
1:30:23
and it can reason about the problems you have
1:30:25
and it's fine if it takes a bit longer
1:30:27
because you get more value out of it. So
1:30:29
that's like sort of the conclusion we came to. Personally,
1:30:32
I don't think like that's sort of the end state.
1:30:35
We're already thinking about can we
1:30:37
do something automatic so that we pick the
1:30:40
right model for your problem and you don't
1:30:42
need to care about it. So it's
1:30:44
more like sort of the implementation detail behind the
1:30:46
scenes like we pick the one or the other
1:30:48
and make
1:30:50
a good guess of like what's the right model for
1:30:52
this. So like because
1:30:54
if I think about AI, I don't
1:30:57
think about picking a model right. I just
1:30:59
think about there is like an intelligence that
1:31:01
helps me doing something that I would need
1:31:03
to do otherwise myself. So that's really what
1:31:05
it is about. And then the
1:31:08
other angle to it which now you know
1:31:10
want to talk more about the features. What
1:31:13
Raycrest is interesting you if
1:31:15
you use Raycrest you do it with the keyboard,
1:31:17
you insert the text and that
1:31:19
we sort of interpret this text. At
1:31:21
the moment, this is very static like you search
1:31:24
a command to center your window, we center
1:31:26
your window. There is another command to
1:31:28
center it left or to move it to the left.
1:31:31
And so it's very statically defined what you can
1:31:33
do. But what AI allows
1:31:36
you to do is that this becomes more of a
1:31:38
fluent thing that this becomes more of something can enter
1:31:40
something I want to do x y that and
1:31:42
then we essentially figure out for you what to do.
1:31:44
And the interesting
1:31:47
bit here is when you think about
1:31:49
the store and the extensions we chatted about all
1:31:51
of them are connected to different tools. So
1:31:53
they connected to your issue tracker to your
1:31:56
wiki to whatever you
1:31:58
have. And so So you can
1:32:00
then start thinking about oh, this is quite interesting
1:32:02
so now I can connect things together that I
1:32:04
would do otherwise. And this gets in
1:32:06
much more around like what is things that
1:32:09
you do on a daily basis very often
1:32:11
that would be good to abstract away and make
1:32:13
it quicker to do. And so
1:32:15
this is something which we're thinking about how can we get
1:32:18
out of this very statically defined feature
1:32:20
set. This is the I don't
1:32:22
know, a thousand things you can do to like
1:32:25
more hey, you can shape
1:32:27
them up yourself by basically chatting
1:32:29
or talking or inputting the text, what do you want
1:32:31
to do about this stuff. And
1:32:34
with that all sort of problems come
1:32:36
which is like fun problem. So it's
1:32:38
not straightforward otherwise we'll probably have build
1:32:40
it by now and it's extremely
1:32:42
hard to build something that is reliable and
1:32:44
fast which basically Rayco stands for
1:32:46
right like we had all the way at the
1:32:49
beginning about how to replace things like spotlight, you
1:32:51
need to be really fast, you need to be
1:32:53
reliable, you need to be 10x better and then
1:32:56
now you gain always the opposite direction to
1:32:58
like LLMs, they hallucinate,
1:33:00
they're slow, how do
1:33:02
you combine those two things to make like a good
1:33:04
product. And some of
1:33:07
those things will hopefully be
1:33:09
solved over time, things just
1:33:11
naturally get better and faster with
1:33:14
those language models. But how can
1:33:16
we actually solve the unique problems to our
1:33:18
use case and how can we make those
1:33:21
really, really good for what we wanted to do. I
1:33:25
think the challenge
1:33:27
is interface, right? How
1:33:29
do you maintain some sort of native approach like
1:33:31
I know when I use chat GPT proper on
1:33:34
the website, there's this analyzing
1:33:36
aspect like it has this
1:33:38
spinning wheel, something's happening
1:33:41
while you're waiting some sort of feedback loop.
1:33:45
And when I asked the same
1:33:47
question on chat GPT for native
1:33:50
on the website or
1:33:52
inside of Raycast, The
1:33:54
output is different. It's a little bit, I
1:33:56
Don't know what makes it different or how
1:33:58
you... Even what
1:34:00
you do behind the scenes to take what they feed you v
1:34:02
the a pie to be different than what they have on the
1:34:04
web. I suppose you
1:34:07
could see the need of Chelsea pretty way. And
1:34:10
so that's always the my concern is like how to you.
1:34:12
How. Do you build? A. Tool successfully
1:34:15
around a moving target that is
1:34:17
a I. That. Has
1:34:19
unique ways to have a
1:34:21
user experience. First class. You
1:34:24
know when you're at their proper and
1:34:26
and wouldn't the way in a wouldn't
1:34:28
the the purists so to speak to
1:34:30
really care about the director their movie
1:34:32
kind of the only guy mean i
1:34:34
you know that I'm just thing like the
1:34:37
purists is like what is it must
1:34:39
be shot on film kind of thing.
1:34:41
back to like I must go to
1:34:43
Chelsea pretty proper to get my chance.
1:34:45
You be t four hours daily or
1:34:47
whatever they have their had a why.
1:34:50
How do you make sure that Re
1:34:52
cast is superior? In. All the
1:34:54
way you want to be superior as
1:34:56
an application, but not less than the
1:34:58
actual need of way. We
1:35:00
know Claude may operate or the Gp
1:35:03
t me operating for those. Those examples
1:35:05
at us near thing is quite interesting
1:35:07
one because. The. I've set
1:35:09
off for the immense unaided sensibility
1:35:11
as many hours or so doing
1:35:13
something similar. So. You ask yourself
1:35:15
the question like what should be exit
1:35:17
do like coming back the like. I'm
1:35:20
excited about ponders like okay, how do we know
1:35:22
it did the right thing. And.
1:35:25
As a my take on this. Is probably
1:35:27
change because they I'd send so quickly
1:35:29
but at the moment it feels like.
1:35:31
Everybody tries to build is. Super.
1:35:34
Scenario day I that solves everything
1:35:36
right. All of our problem in
1:35:38
such a simple sad and why
1:35:40
this might work and we we
1:35:42
can get there eventually. It's
1:35:44
quite. Hard to build something
1:35:46
that's America. Because. He
1:35:49
usually we do the other way around to
1:35:51
build tools for very specific things. We.
1:35:54
Have an Id either write code. We.
1:35:56
Have a notion to write not code but
1:35:58
also type of text right? They
1:36:00
complete a different interface, completed
1:36:02
different tool. But. Both of them
1:36:04
produce text that the and. So.
1:36:07
Now going back to the basic this.
1:36:09
like what is actually the unique stuff.
1:36:12
That. He do with rey cause. What?
1:36:14
Is actually the things that we can do Good.
1:36:17
To. See, it's enough and with more come
1:36:19
from this angle. And when I think aware
1:36:21
that it's more okay. We kind of
1:36:23
about how can we make people. Staying.
1:36:26
In their slow better how can you make sure
1:36:28
they get information that need quicker to do that
1:36:30
shop? How do we make sure that. Really?
1:36:33
Moved us distraction because all the tools they use
1:36:35
on a daily basis. Believe it or not the
1:36:37
have an upset about it I want to put
1:36:39
on top of your i don't wanna make. You.
1:36:41
Visit this tool. For. Better of
1:36:43
words and but you have he own agenda so
1:36:45
we want make sure of the can stick to
1:36:47
your own agenda. And. Some of
1:36:50
those in the them and can be
1:36:52
self was a i think an analysis
1:36:54
intelligent the way by connecting doesn't feather
1:36:56
to better but some also maybe not
1:36:58
so. As. And goods it's quite
1:37:00
easy at a moment to see like the
1:37:02
friend feel play always hope everything was a
1:37:04
I'd It's always a bit mindful about. Cause.
1:37:07
Sometimes the best solution might be not with they
1:37:09
I might be just like. The
1:37:11
Good: Old classic code that we
1:37:13
ride that produces much more value than
1:37:15
me. Think. Other night by the
1:37:17
when you come with this mind that it's about more. What?
1:37:20
Sort of the secret sauce that we have. That.
1:37:23
Sits our target audience that are really
1:37:25
the problems that have on a daily
1:37:27
basis. And that sort of first
1:37:29
class. Like product filling. I
1:37:31
could go to your users if figure out
1:37:33
what problems that have you sold them. With.
1:37:36
Or without a I didn't as I basically them
1:37:38
find a solution. but at the moment I feel
1:37:40
like it's a lot around are we talking about
1:37:42
a solution? It's ally but we don't really know
1:37:45
the problem and outside to think of a dangerous
1:37:47
like isn't if is thought this way around. While.
1:37:49
The one that you can recommend or at
1:37:51
least mention is the amount of miles Like
1:37:53
he released the blog posts literally today that
1:37:55
they were recording April Tenth, the Spirit forty
1:37:58
percent by the way. For. whom. listening,
1:38:00
you've got all the models essentially, at least the
1:38:02
ones that seem to matter most. You've got the
1:38:05
chat GPT 3.5, which really, if
1:38:09
anybody's using that, I just feel bad for them
1:38:11
basically. But it
1:38:13
is about access. Then you have
1:38:15
GPT 4, you've got Cloud 3 Opus, you've
1:38:17
got Turbo versions of GPT 4, which I
1:38:19
didn't even know existed. I didn't even know
1:38:22
you can Turbo GPT 4, but
1:38:24
you can have extended versions of it. You've got Cloud
1:38:26
3 Haiku, Cloud 3 Sonnet,
1:38:28
Sonar Small, Sonar Medium, Llama,
1:38:31
Mixroll, Code Llama. You've got
1:38:34
all the different LLMs out
1:38:36
there, at least the ones that are most popular today. So
1:38:38
you've got one interface to sort of access
1:38:41
the most popular ones out
1:38:43
there. How will you maintain,
1:38:45
I guess this,
1:38:47
Raycast began as a launcher and
1:38:50
now is a full-fledged access to
1:38:52
LLMs application. Where
1:38:55
do you go from here? Good question. So
1:38:58
Raycast itself is a set
1:39:01
of productivity towards an hour mind. It's
1:39:04
basically, we see the launcher
1:39:06
is the centerpiece of it. The
1:39:08
launcher is essentially making sure
1:39:10
that you can access all the information and
1:39:12
that's where we're coming from. We already have
1:39:15
something like Raycast floating notes, just
1:39:17
little notes I can have on the side to
1:39:19
basically take notes during a meeting or
1:39:21
during a day. I use it to
1:39:24
plan my week and check off my tasks. So
1:39:27
we have sort of this concept
1:39:29
of having multiple applications in one.
1:39:31
The AI chat is the latest addition. So AI
1:39:34
chat is like it lifts on the side as
1:39:36
well as it almost feels nowadays like a standalone
1:39:38
application. And we feel like
1:39:41
Raycast itself evolves more and more into
1:39:44
like a own ecosystem, like a productivity
1:39:46
suite. It's all packaged in
1:39:48
one application because nobody really likes to
1:39:50
download multiple ones. So we
1:39:52
feel like that's like a nice way of
1:39:54
like structure in your work where you have
1:39:56
this productivity suite that works really nice together.
1:39:59
The launcher is the same. The people bed. And.
1:40:01
Then you basically can. Use. Nodes
1:40:03
and I checked separately and despising enrich
1:40:06
that there might be other things coming
1:40:08
that be adding to to this mix
1:40:10
as well. Cause. Like the
1:40:12
launcher while at the crate interface for
1:40:14
a quick access. Weatherbill.
1:40:16
The ads had was. The. In
1:40:19
the size of a laundry isn't crowded. you
1:40:21
wanna have keep a conversation conversations that. Go.
1:40:23
Of I'm one of the message to maybe one
1:40:26
or refer wendy like go back to your code
1:40:28
editor away that stuff. So. That basic
1:40:30
so there is a need for like a specific in
1:40:32
the face for that. And. So that's
1:40:34
something the tree of them see that with. Start
1:40:36
with the hey that's indicative rey cause. Then.
1:40:38
With see like oh excellent. Semi.
1:40:40
Be his kids were better is outside we
1:40:43
not really sign away. Feeling. About side
1:40:45
and is a thing is like the good thing. That.
1:40:48
We saw natives and close to the metal
1:40:50
close to the operating system we can do
1:40:52
to add relatively easily and bike. Can
1:40:54
may goods of appear like all the
1:40:56
Us replications you have the to make
1:40:59
sure. The. To have this consistent,
1:41:01
consistent feel. How to
1:41:03
handle when you get. These.
1:41:05
More for windows for easy the the
1:41:08
falling windows for the notes for example
1:41:10
or even the chat window for ai
1:41:12
is since the you have recast as
1:41:14
a singular application but then he sort
1:41:16
of have like Saab applications. Within.
1:41:19
That will you eventually have more some
1:41:21
applications as you grow on. What are
1:41:23
the challenges of that? For me? You
1:41:26
know? A you are perspective.
1:41:28
How do you manage the toggling owns?
1:41:31
The. Conjuring of an application via the
1:41:33
launch bar to the Ai window
1:41:35
that then stays floating. and I
1:41:37
mean it seems challenging from I
1:41:39
like a Us perspective that you
1:41:41
have to. Dance. Around
1:41:43
in need of application land and how
1:41:45
that user experiences and does naturally done
1:41:48
on a mac is is this sounds
1:41:50
so the first thing that really it.
1:41:53
Did. was like thirty know it's which was
1:41:55
very primitive window and this what kind of
1:41:57
fine and that the whole thing was rounded
1:41:59
slowed It stays on top of all
1:42:01
the other application is a bit
1:42:03
inspired by the launcher right which also is on
1:42:05
top of all the other Application the
1:42:08
second of them we did was AI chat Which
1:42:10
we started similar like audience is just like floating now
1:42:12
It's in it it floats and stays on top and
1:42:15
then the UX channels started to appear
1:42:18
Because people and say like oh But
1:42:20
I want to have it not floating
1:42:22
and I want to have a dock
1:42:24
icon and I want to have a
1:42:26
sidebar And these are all valid
1:42:28
feedback like we said again. It makes
1:42:30
sense. So we then realized all the
1:42:33
Interface you build with the launcher and
1:42:36
this the UI components built
1:42:38
though didn't really suited the
1:42:40
air chat so we did
1:42:42
on we went on the journey of like okay,
1:42:44
cool, let's take on the feedback and rethink their
1:42:46
chat and So a
1:42:48
few weeks ago we launched the revamp of
1:42:51
that which is much more Like
1:42:53
a real chat app if you look at
1:42:55
it If you open AI chat on your
1:42:58
Mac with raycon's right now, it has
1:43:00
a toggle for a sidebar Looks
1:43:02
and feels almost like the messages app on your
1:43:05
Mac. You can have different
1:43:07
conversations You can create new ones. You can
1:43:09
pick the model you want to have for
1:43:11
a different threats you have and
1:43:13
so it became really this rich application
1:43:16
and so for that we
1:43:18
had to Revamp a lot of the
1:43:20
things before we were very keyboard driven
1:43:22
and we still are But like
1:43:24
when you work with the chat you're going to
1:43:27
select text and you're much more clicky so we
1:43:29
had to actually Build like
1:43:31
a bunch of things that we didn't
1:43:33
really had in raycos because it wasn't necessary
1:43:35
and had to build it out because you
1:43:38
just perceive this as
1:43:40
kind of window much more differently because it
1:43:42
is essentially like an app it's just Bundled
1:43:45
within within raycos to make
1:43:47
it easier to distribute all of this. Yeah,
1:43:50
I suppose a challenge comes back into play whenever
1:43:53
If you're a command tab, you
1:43:56
know, I don't know guru frequent user like I
1:43:58
am like most people
1:44:00
are I think. They are very
1:44:02
keyboard driven. Now you've got the scenario where,
1:44:04
and this is where the UX I think,
1:44:08
you know, permeates to continue to be a
1:44:10
challenge, even now, even with
1:44:12
these updates, is that whenever I wanna
1:44:14
command tab a way to another application,
1:44:17
that AI chat window is still there and I
1:44:19
can't command tab back to it. It doesn't feel
1:44:21
like, and if I can, I haven't found a
1:44:23
way to do it yet. So
1:44:26
I still feel like you can. You can,
1:44:28
so if you're in the chat and do
1:44:30
command K, you can disable
1:44:32
always on top. And what this does,
1:44:34
it basically makes the window
1:44:37
behave like a normal window. So then you're
1:44:39
going to command tab, you get
1:44:41
a dog icon, you get all of those goodies. So
1:44:44
we sort of have these two separate nodes depending
1:44:46
on how the person likes to use it. Because
1:44:48
we got the feedback, it's like, oh yeah, I
1:44:50
wanna have it command tabbing.
1:44:53
And then we also got feedback from others, oh, I
1:44:55
love how this is floating. And the third
1:44:57
of half is contradicting opinion.
1:45:00
And we often times, we don't
1:45:02
really wanna add many settings because we feel
1:45:04
like the majority of people
1:45:06
use a default setting anyways. So we wanna
1:45:08
keep them to a minimum. And every setting
1:45:10
you add also has a huge cost and
1:45:12
maintenance because with every setting you
1:45:15
have, you have to test two cases.
1:45:18
So, and then two mutations of that.
1:45:20
So we keep them usually down to a minimum. But
1:45:23
in this case, it was like, yeah, if
1:45:25
you have those strong contradicting opinions, it's
1:45:27
probably the right way to add
1:45:30
a setting for that. It's a bit hidden. One
1:45:33
of the challenges we have generally when you
1:45:35
have a quite compact UI, which is quite
1:45:37
condensed, and not wanna overvalue the
1:45:39
user, you kinda end up
1:45:41
like often times with some of those hidden
1:45:44
settings, which
1:45:46
we probably need to do a bit better. Yeah, that would've,
1:45:48
I mean, yeah, I find I'm an advanced user, at least
1:45:50
I think I am. And I
1:45:52
would imagine I'm a pretty advanced user. And
1:45:56
maybe I would've stumbled on that, but I think maybe I wouldn't have. I'm
1:45:59
familiar with Command-C. Hey, inside of different
1:46:01
recast applications are extensions are interface
1:46:03
is essentially I'm unfamiliar with that
1:46:06
being. A. Way to get the other
1:46:08
things. But. There's a lot and
1:46:10
there's like yeah, I've got a bunch of
1:46:12
pace responses of the responses copy and co
1:46:14
op een chat and are really gonna dig
1:46:16
to find that. That particular nugget.
1:46:19
Least. He. Even as part of
1:46:21
that, his colleagues hidden down there really
1:46:24
aren't even aware to look for that.
1:46:26
Thankful you do have a search rations
1:46:28
option, which I mean stellar You? I
1:46:30
mean by no means doesn't count by
1:46:32
your application, but just these are the
1:46:35
you x challenges every team faces. Regardless
1:46:37
of how good you are. Regardless of
1:46:39
how amazing your software is, you always
1:46:41
have some sort of you x feedback.
1:46:44
That. Is just like this year and have some for
1:46:46
it's. And your interface and
1:46:48
you you x is stellar. But.
1:46:51
This is one thing I think you'll have a
1:46:53
challenge of is is like. You. Know
1:46:55
this says task switcher You know
1:46:57
icon to tear back to for
1:46:59
example. A. Comes right in
1:47:02
the moments which is good of of you
1:47:04
know as your search shown with me how
1:47:06
to use this feature of unable to feature
1:47:08
to make it not float. Saw.
1:47:10
A test it out here in real time. And
1:47:13
I can imagine you have your your work for
1:47:15
yeah for sure. Only exotic
1:47:17
continually expand recast to have.
1:47:20
Sub. Applications are sober.
1:47:22
You. Know distinct feature sets. With.
1:47:25
Distinct you? why? With. The
1:47:27
distinct you acts of that you I. Within.
1:47:30
The The Global Recasts
1:47:32
application. That's. Such.
1:47:34
A challenge? Yeah, add some.
1:47:37
That's. The fun part. Is the family
1:47:39
portrait. As I know I to
1:47:41
I tend to like it. It's like
1:47:44
okay societies start building. It is study
1:47:46
is an ad and then. We'll
1:47:48
define ourselves a real than. With
1:47:51
design exploration to his dog or
1:47:53
something. A predefined, they found that
1:47:55
we start with something where we want to do
1:47:57
saints. And then betray him. Bonus Big
1:47:59
thing. And any work
1:48:01
on them design was a few days.
1:48:04
And then we go in circles. And like
1:48:06
Exley simplifying, I press the clay. And.
1:48:09
Come up with like a much better experience. and
1:48:11
is like a dismal with a slight always says like.
1:48:14
This weird to say blob is like. Ah,
1:48:17
Your side Now Finally, the simple
1:48:20
solution. But. The eggs the had
1:48:22
all of the big way to find it which
1:48:24
is like is as i like this feeling because
1:48:26
i feel like very. Very. Good that
1:48:28
you're finally made it. But. A need to
1:48:30
go to waste food is expensive to
1:48:32
sign. Period. To figure that
1:48:34
moment out. And he wouldn't
1:48:36
end up there with when in the haven't
1:48:38
Explored all the author angles. And. It
1:48:41
off mans means that you. Throw.
1:48:43
Away Ninety percent of the snow freud.
1:48:45
Nelson and say like that the
1:48:47
best Think Sybil is actually the
1:48:50
six Don't bill because like you
1:48:52
go to so many iterations. Dad
1:48:54
actually. You saw up and
1:48:56
like what you actually want to do with every
1:48:59
the to raise new node of problem better. You
1:49:01
take the solution out and more
1:49:04
detail. And then of them Sudan
1:49:06
and walk back and so they actually the simplest
1:49:08
way to do x with that but don't need
1:49:10
to do all the opposing but we explored it
1:49:12
and be Canada we don't need to do them
1:49:14
in his yeah I kind of like the mode
1:49:16
when you dare to the chinatown of them the
1:49:18
painful because the do this at this doesn't feel
1:49:20
right. This. Doesn't seem right. And.
1:49:22
To have did like four or five times
1:49:24
as that. Oh and I kind of his
1:49:26
ride but it's much simpler that be admissible
1:49:28
with severe which is good. For.
1:49:31
The same with code. right at the same
1:49:33
with Coat of Bugs You bang your head
1:49:35
against a wall. In. Writing want
1:49:37
to solve this problem and you're
1:49:39
right. it's crazy thing and then.
1:49:42
You. Sleep over it and the next
1:49:44
that actually I can do this and
1:49:46
smooth coat per month.yeah as a simpler
1:49:48
way, Yes, absolutely won't. Speaking of a
1:49:50
potentially simpler where was it. Did.
1:49:53
You explore the idea of a i chat as
1:49:55
is as if called it a couple times. The.
1:49:58
Do explored as a standalone up location,
1:50:00
but it makes sense to have a
1:50:03
standalone application that is still part of
1:50:05
the recast brand. You. Know
1:50:07
same rate as subscription, just a
1:50:10
standalone. Separates. Maybe
1:50:12
even a code base I know, like or
1:50:14
their shared libraries? The concerned with what. Have.
1:50:16
You get his did explore. That doesn't make
1:50:18
sense. We didn't really explore it because it
1:50:20
came out of the race and like the
1:50:23
initial version was in roy cause swindle way
1:50:25
that we had it's happened there. As
1:50:27
it turned out like out yeah, this
1:50:30
is not always craved because the window
1:50:32
closes. That cell phone clicking outside. And.
1:50:34
And Zoc as will be two separate window. Suburban.
1:50:37
Though when I'm. Again, Like a
1:50:40
few it's a razor blade or it felt
1:50:42
good would bears I guessing two things like
1:50:44
with one we saw the what out extra
1:50:46
to make didn't separate application and put it
1:50:48
into Mac app store and see if that
1:50:50
yield some nice result like some exposure and
1:50:52
the Mac app store. Was rate
1:50:54
of itself isn't that they're. Due. To
1:50:56
some limitations. And then the
1:50:59
other idea there would be. Maybe.
1:51:01
You could have an edge had also on the web
1:51:03
just i for people to quickly check it out the
1:51:05
than if they're like it they could say i want
1:51:07
to have this all the time where the poll on
1:51:09
a desktop had good been interesting angle to it as.
1:51:12
As. Is a tricky bit as like if
1:51:15
you make multiple applications, you start as
1:51:17
a user base over and over and
1:51:19
over again. Like. Okay, everybody
1:51:21
had downloaded launcher. Now you
1:51:23
need to get them to download the chat. And.
1:51:25
Then potatoes new application need to do
1:51:28
the same way. I think it's much
1:51:30
easier if you cut in degraded and
1:51:32
I do consider to use isn't much
1:51:34
more hate you have to software already
1:51:37
now Yeah extra functionality. And
1:51:39
I think you see something quite often and. A
1:51:41
set of big big apps like social
1:51:43
apps like Instagram, well of Facebook. They.
1:51:47
Sometimes unbundle and that doesn't really work and
1:51:49
then at a bumbling into Canada and I
1:51:51
am Now it works because it's a much
1:51:53
easier switching cost for for users. So.
1:51:56
I suppose this hypothesizing. a
1:51:58
little bit on This is
1:52:01
kind of like marketing but it's
1:52:03
also getting attention, gaining awareness, adoption
1:52:06
obviously. How
1:52:09
well do you think you do to broadcast
1:52:12
to the world that
1:52:14
there's an AI chat that is a single pane
1:52:16
of glass to the
1:52:18
world that cares about a single pane of glass to
1:52:20
all AIs? Or two,
1:52:22
I guess the end quotes
1:52:25
most popular today, artificial intelligence is
1:52:27
out there that can be used because I feel
1:52:30
like the feature, this
1:52:32
feature is locked behind
1:52:35
Raycast the application, which
1:52:38
is to some degree arguably
1:52:41
bigger than the Raycast brand
1:52:44
so to speak. To
1:52:46
say you have a single pane of glass
1:52:48
artificial intelligence chat to
1:52:51
me is like wow, that's not so
1:52:53
much the bigger thing but it's big
1:52:55
enough to be standalone. Yeah, you're
1:52:58
touching down a good topic like generally
1:53:00
with Raycast because it has so many
1:53:02
features, we oftentimes find it very hard
1:53:04
to describe it. So one
1:53:06
of the things we did like we currently
1:53:09
working on a new website that
1:53:11
comes out soon where we basically work
1:53:13
on exactly this topic of what
1:53:15
is Raycast, what are the main
1:53:18
features, how should people think
1:53:20
about it, what is the story we want to
1:53:22
tell. And so that's one of the
1:53:24
things which we can certainly do better
1:53:26
and I think we always can do better,
1:53:29
especially as soon as you throw in new
1:53:31
things in the mix, it changes a bit, right? So
1:53:34
let's say a year ago was not so AI heavy,
1:53:37
now there's maybe more AI with separate chat
1:53:39
that has access to all those glorified
1:53:42
models. So I think this is like something which
1:53:44
we can better. I think
1:53:46
it's generally a problem when you have such
1:53:49
a broad feature set or even like separate product,
1:53:51
how do you group them together, right? Like what
1:53:54
is the overarching theme
1:53:57
you give somebody, why should they get interested
1:53:59
in? this thing, like how do
1:54:01
you target them towards this or that, which
1:54:03
is something which we haven't cracked as I
1:54:06
mentioned and something which I think we
1:54:08
can work on. But it's also
1:54:10
like a continuous work. You saw that we
1:54:12
do a bunch of stuff on YouTube, which
1:54:14
I think you can like wave in this
1:54:17
message. We recently, when we revamped the chat,
1:54:19
it became more of a thing. So
1:54:22
we kind of resurfacing this and then and bring
1:54:24
it in. But yeah, it's something
1:54:26
which is quite tricky to get under one
1:54:28
umbrella with like the launcher,
1:54:30
the AI chat, the floating notes, soon
1:54:33
an iOS application. There was a
1:54:35
lot there, right? It's much more than just
1:54:38
like a spotlight replacement by now. Yeah,
1:54:40
I think that's the in your
1:54:43
particular position, this is your struggle
1:54:45
to concern with, right? Your
1:54:47
team can probably have the
1:54:49
concern and have a concern for it,
1:54:51
but it's not necessarily their job to
1:54:54
solve it, right? This is
1:54:56
really you and Peter's job to solve this
1:54:58
brand. This is a brand problem. This is
1:55:00
a, you got a brand umbrella
1:55:02
at this point. The brand is Raycast and they
1:55:04
have a sub brand called AI chat. And
1:55:07
I think, are you a fan of Silicon Valley
1:55:09
TV show Bunny Chance? No, yeah, I'm the percent.
1:55:11
Okay. So there was a particular,
1:55:14
I'll spoil this. So spoiler alert here. Richard
1:55:21
Hendricks was trying to create this algorithm. He had, he was
1:55:23
really big on the platform. And at
1:55:25
one point, I think it was season three, there
1:55:27
was a hooli chat was popular.
1:55:30
And then you had Piper chat,
1:55:32
which they almost, you know, Richard quit the
1:55:34
company for a bit. And then they ran
1:55:37
with that and Dinesh became CEO and all
1:55:39
of them happened there, right? So, you know, for
1:55:41
a bit there, the whole entire show
1:55:43
could have been derailed on pursuing
1:55:46
this platform. And so in
1:55:48
your case, Raycast is your platform. Right.
1:55:51
And Then Piper chat camp comes out
1:55:53
and it's arguably better than Hooli chat
1:55:55
ever was.. And Hooli chat was the
1:55:57
behemoth in the grand scheme of things.
1:56:00
In that world in that universe, right? And
1:56:02
they pursued it a little bit. And.
1:56:04
We never did it. But. Was a standalone
1:56:06
application, wasn't part of a platform, the wasn't party
1:56:08
algorithm a date leverage the irony to be amazing
1:56:10
for compression in a to compress all of a
1:56:13
to to make it super fast like and stuff
1:56:15
and I just wonder if you zoom out a
1:56:17
bit. Is. The idea of a
1:56:19
single pane of glass. A. I
1:56:21
a chance. Bigger than
1:56:23
recast. And first than yeah,
1:56:26
I mean it's hard to tell of
1:56:29
mine tried as things the Silicon Valley
1:56:31
and Alex is quite interesting now feel
1:56:33
like there is. I'd buy that he
1:56:35
a half looks had which isn't application
1:56:38
of the crate compression algorithm right? Them.
1:56:41
With others and of a bit different because. I
1:56:43
would say it was. Multi.
1:56:45
Product and away. I
1:56:47
the close as I could come up with
1:56:49
this probably say Microsoft Microsoft has an alpha
1:56:52
see. And an Elvis. You do
1:56:54
have an application for our. Email you
1:56:56
have an application for. A
1:56:58
calendar and songs for. So.
1:57:00
He of would say that to that say hey you have.
1:57:03
The launcher with souls or that acts
1:57:05
as a problem space and a half
1:57:07
the i chat with slept to check,
1:57:09
told them areas a i'm ones out
1:57:11
there and thoughts that dedicated problem space.
1:57:14
So. Think they get a list nicely. Next.
1:57:16
To each other but under questions like what's
1:57:18
the bigger on pop right at how the
1:57:20
a bundled together at the moment all of
1:57:23
this is called rey cause we're not really
1:57:25
and as. And as separate
1:57:27
product World's Sauce Yet. Now.
1:57:29
Says like called rey cause and the at the happened
1:57:32
to be a launcher and a happens to be it's
1:57:34
has or notes. Bundled. with their i'm
1:57:36
over time as those pieces get more this
1:57:38
to stainless like it's passage for them with
1:57:41
the at said would trees and gotta make
1:57:43
over. This. Becomes more for think how you
1:57:45
talk about it older it's like now it is. Sad.
1:57:48
Which is different to the launcher and we
1:57:50
probably need to find a way. How
1:57:52
to do that in a mood? Yet for
1:57:54
he's it for a. While. You're You're
1:57:57
pretty close to Pepper Chat. You
1:57:59
know, It's actually I chat hyper
1:58:01
chat. Seibert said we just need
1:58:03
a better or compression algorithm this
1:58:06
right. Well
1:58:08
the similarity to to silicon valley's
1:58:10
Tv shows was just draw a
1:58:13
comparison. I. Had I just wonder
1:58:15
if like eventually get to a point where
1:58:17
he i chat. Not so much
1:58:19
is bigger but could be. More.
1:58:21
Focused as a product and more focus
1:58:23
as an on boarding and more focus
1:58:25
as a pricing. Perspective.
1:58:28
Is it was a stand alone. Application
1:58:30
you don't in like he had save
1:58:32
with Microsoft as example. you don't install
1:58:35
Microsoft and get Word and Excel. Install.
1:58:37
Like probably something else the gives you
1:58:39
all that stuff. To. Get that. Most
1:58:42
next less less about what's over the horizon
1:58:44
and the you've got his, he'll be in
1:58:46
a glass now to. A. Lot more
1:58:48
popular alums out there, you resigning it as
1:58:51
we speak with deliberated over whether or not
1:58:53
it should be a steel. I'm saying how
1:58:55
that might actually work out, but yeah, what's
1:58:57
just over the horizon? Not many folks might
1:58:59
know what. what should people know about where
1:59:02
you're going. There. So that may
1:59:04
start on the macro level. on that
1:59:06
go under the micro level. So mark
1:59:08
for short term. We. Released
1:59:10
our model and the next thing which
1:59:12
might be out already when we published
1:59:14
it's blog or the spot. Is
1:59:16
that be? Also want to make it easy to. Build.
1:59:19
Your own preset for a check so you
1:59:21
can define a model. Give it like a
1:59:23
few instructions and then you can reuse it
1:59:26
or share it with others. Numbers
1:59:28
as is quite cool to print
1:59:30
and more personality those chats looks
1:59:32
like sometimes. Said. Suffer thing.
1:59:35
As annexing which one do with had it
1:59:37
about of the bed like we have goes
1:59:39
extensions and new on A and decried them
1:59:42
together with a i've to read a combined
1:59:44
those two powers I basically take the intelligence
1:59:46
of a I'd. Say. Galleon and
1:59:48
in the face and take the
1:59:50
connections we have resolved. extension something
1:59:52
that's like really excited for the
1:59:54
accented developers. And equally for
1:59:56
all users of those extension with and prose
1:59:58
me really excited about. And.
2:00:00
Then we have been spoken about this
2:00:02
public the already it's like I was.
2:00:05
That's. Like an interesting take we want to
2:00:07
pray the rey cause to i was. Been.
2:00:10
One of our request, especially with Sinclair
2:00:12
Said, but also with other things like
2:00:14
snippets so quickly to have them in
2:00:16
your pocket available when the on the
2:00:18
dope. So we working there. On.
2:00:20
A mobile application with comes out this summer.
2:00:23
Around the summer I would probably the start was a
2:00:25
bit of a beta test and than see how far
2:00:27
we can get. It's not gonna
2:00:29
be unfortunately one to one right cause
2:00:31
I was assessed to restrict the phone
2:00:33
for violence of economy to be a
2:00:35
bit crate if they're. Probably.
2:00:37
Gonna be something around a companion see after
2:00:39
most important things always with you and can
2:00:41
reuse that. For. That's exciting.
2:00:44
The. All the big piece words we
2:00:46
haven't talked publicly much about. His.
2:00:49
Windows, May. Go As as much
2:00:51
as I love it is not the only operating system
2:00:53
at us for put out there them more that. The.
2:00:56
For us, it's also questioned like. How
2:00:58
can we go to us are operating
2:01:00
system and bring the same experience. To.
2:01:03
Windows axis of magnitude bigger than
2:01:05
that are why even also Linux
2:01:07
later on. Though. We're thinking about,
2:01:10
how can we do that. Or
2:01:12
we always knew we want to do
2:01:14
this. We just always all the know.
2:01:16
it's technically quite a hard task. Because.
2:01:19
He talks about. Being in a deputation gets
2:01:21
us all the advantages being close to the
2:01:23
operating system, given us the performance, but also
2:01:26
given us a look and feel. Off.
2:01:28
A native application. So. We
2:01:31
kind of researching. And.
2:01:33
Finding ways how to do that actually on
2:01:35
on windows as well. And make
2:01:37
people happy over there because we getting
2:01:39
that l ste Marie of them like
2:01:41
com that's one of. The. Feature
2:01:44
requests a pop said every now and then
2:01:46
we don't have an ends rather than. Us.
2:01:48
It would happen eventually. But. This year
2:01:50
we want to get to a stage where there is a
2:01:53
much more concrete plan. hopefully with some
2:01:55
really be taught that people can explore
2:01:57
it and help us making it happening
2:02:00
Big ambitious goals to be
2:02:02
multi-platform for sure, especially if you
2:02:05
consider Linux down the road and
2:02:07
I suppose You probably have
2:02:09
to hire right? You probably couldn't do a Windows
2:02:12
application with the same team you probably have to expand by
2:02:14
at least one or two Maybe three,
2:02:16
you know like a small team
2:02:19
or people who specialize in Windows
2:02:22
application development and I'd
2:02:24
imagine that the real challenge is probably not
2:02:27
just that but like keeping the user interface
2:02:30
similar Right. Yeah, if
2:02:32
you've come to know raycast for what it is, you know
2:02:34
that it's got a great beautiful
2:02:36
user interface a really good
2:02:39
UX in Terms of
2:02:41
like what you can do with a single limited
2:02:43
application like it is in terms of like its UI
2:02:46
footprint is not very big It's meant to be minimal.
2:02:48
So you do a great job of Maintaining
2:02:50
that minimalism but also allowing
2:02:52
you to even expand the
2:02:55
launch bar with a down arrow For example to see
2:02:57
more applications or to see the release
2:02:59
notes or whatever Like I think those are all subtle
2:03:01
ways you allow the application to
2:03:03
remain minimal but also expand when necessary
2:03:06
That's gonna be a challenge to do when you
2:03:08
go to Windows or another platform is like
2:03:10
maintain that stability from the UI
2:03:12
Yeah, 100% I think that there's
2:03:14
a lot of challenges. There is technical
2:03:17
challenges there are design challenges It
2:03:20
can be a nice opportunity to also rethink
2:03:22
some of the UX and the design to
2:03:24
like set ourselves up for the next five
2:03:26
years It also can help
2:03:28
us rebuild some of
2:03:30
our technical foundation Like
2:03:32
also sustain longer because we mentioned like we
2:03:34
have now not only a launcher We have
2:03:36
like also a chat on the side
2:03:38
and other things we have all of that
2:03:41
So I think like that's sort of also Finally
2:03:45
quite interesting recruit a team now want
2:03:47
to grow it a bit further to like expand to
2:03:49
this and and then also making sure
2:03:52
that we have Going
2:03:54
back to like the momentum to ship on one
2:03:56
of the platforms, which is also not the most
2:03:58
easiest thing to do So
2:04:00
that's all the things we're basically figuring
2:04:03
out as we speak more or less. We
2:04:05
have already like a sort of
2:04:07
a plan that we want to explore and then
2:04:09
hopefully can share more about it very, very soon
2:04:12
when we gain sort of the confidence that that
2:04:14
is a good way for us building all of
2:04:16
this. Well, good deal. Appreciate
2:04:19
you digging deep into all
2:04:21
the details and even allowing me to push
2:04:23
back here and there on
2:04:25
different parts. Yes, definitely once you think
2:04:27
about that separate application, AI Chat, Raycast,
2:04:30
AI Chat, I don't know what to call it but
2:04:32
seems like a big deal to me, maybe potentially
2:04:34
a bigger deal or not a bigger
2:04:36
deal necessarily, more like a separate deal. It's
2:04:38
less like a bigger deal, more like a
2:04:40
separate deal. At some point the AI
2:04:43
Chat is going to have enough of a feature set
2:04:45
that it really is distinctively and it's
2:04:47
distinctively different than Raycast application
2:04:50
proper. Yeah. Raycast user
2:04:52
and I see that already from my
2:04:54
own perspective is like it's a tool
2:04:57
but I
2:04:59
suppose if Raycast has given me access to
2:05:01
productivity tools that give me predictive
2:05:03
nature, I suppose that can be argued that
2:05:05
it is a subset but I'll leave it
2:05:07
there. That's enough for now. Thomas,
2:05:10
thank you so much for digging in. Thank you so much
2:05:12
for sharing some of your
2:05:14
journey product wise and direction wise. Here
2:05:17
on the Change Log, I love that digging in with you and appreciate
2:05:19
you coming on the show. Thank
2:05:22
you so much. Yeah. Thank you
2:05:24
very much, Adam, for having me. David
2:05:26
Listener, always happy to pop behind the
2:05:28
mic and contribute to the Change Log
2:05:30
as well. Yeah, man. Well,
2:05:34
thank you. It
2:05:36
is always such a good
2:05:38
feeling to have someone like Thomas
2:05:41
Paul Mann, an accomplished CEO, leader,
2:05:44
founder, be a listener and
2:05:46
to come on the show. That's
2:05:49
like the coolest thing ever, honestly. Now,
2:05:51
I don't know about you, but I'm
2:05:53
a Raycast user, as you can tell.
2:05:56
I use Raycast for everything. I use
2:05:58
it all the time. single
2:06:00
day it's core to
2:06:02
my workflows, my toolset,
2:06:04
so much hinges upon
2:06:07
Raycast. And maybe the
2:06:09
next big thing for them really is
2:06:11
Raycast AI chat. Maybe they need to
2:06:13
go the way of Patrice and just
2:06:15
simply break it off and
2:06:17
make it a standalone application. I don't
2:06:20
know. But what I can say as
2:06:22
a user after this conversation
2:06:24
of course, because I was like skeptical,
2:06:26
I'm a user of Raycast AI chat
2:06:28
on the daily now. And
2:06:30
I have to say I really
2:06:33
wish it was a standalone application.
2:06:35
There's certain UX behaviors, certain functionality
2:06:37
you expect from an application. And
2:06:40
right now it achieves most of that. And
2:06:43
I think they're gonna break it off and make it
2:06:45
something. The only time will tell. So we'll see. The
2:06:48
cool thing is is most of what
2:06:50
Raycast is today is totally free for
2:06:52
everyone. So if you haven't yet, check
2:06:54
it out. A big thank
2:06:56
you to our friends who sponsored this show
2:06:58
today. Fire hydrant, less stress
2:07:00
from ring to retro. Our
2:07:03
friends of course at Cloudflare and the
2:07:05
awesome announcements from developer week 2024. I
2:07:07
was down at their offices hanging out,
2:07:09
having fun. It was such a blast.
2:07:11
We're big fans of Cloudflare. And of
2:07:13
course to our friends at Century, use
2:07:16
the code change law to get 100
2:07:18
bucks off the team plan. That's good
2:07:20
stuff. And our friends over at Factor.
2:07:22
Meals made easy. I love them. You
2:07:24
should check them out. And of course,
2:07:26
a big massive thank you to our
2:07:28
friends and the home of change all
2:07:31
to come fly.io. Put your application
2:07:33
near users is too easy. And
2:07:36
last but not least, thank you to break message
2:07:38
cylinder to beat freaking residents. Yes. As
2:07:40
a matter of fact, we have an episode
2:07:42
coming up this Friday on friends
2:07:45
with break message cylinder going deep on
2:07:47
all the things. You're gonna
2:07:49
love it. All right, that's it. The show's done. We'll
2:07:51
see you on Friday. you
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More