[00:00:01]
Welcome in to the BuiltOnAir podcast,
[00:00:04]
season twenty three, episode eight.
[00:00:06]
This will be the last official episode for
[00:00:10]
this season,
[00:00:11]
but you may see us in between.
[00:00:14]
So good to be with you, Dan Fellars,
[00:00:16]
Kamille Parks,
[00:00:17]
Ali Alosa and special guest Rob Breidner.
[00:00:20]
Welcome in, Rob.
[00:00:21]
Hey, everyone.
[00:00:24]
Thanks for having me back.
[00:00:24]
Been on the show before.
[00:00:26]
So we’re excited to get an update on
[00:00:28]
what’s going on with Rob.
[00:00:30]
He’s got some exciting stuff that he’s
[00:00:31]
going to show us.
[00:00:33]
But a couple of things going on that
[00:00:35]
we wanted to make you aware of.
[00:00:37]
And so we’ve got there’s two hackathons
[00:00:40]
coming up.
[00:00:41]
One official one from Airtable.
[00:00:43]
That one’s on site in New York.
[00:00:47]
Dates again.
[00:00:48]
What are the dates on that?
[00:00:51]
March first.
[00:00:53]
March first?
[00:00:54]
As the beginning of when you could do
[00:00:56]
your submissions, sorry.
[00:00:59]
Oh, okay.
[00:01:01]
There’s too many.
[00:01:01]
I know.
[00:01:04]
I think the Airtable one’s the nineteenth,
[00:01:05]
I think.
[00:01:06]
Okay.
[00:01:06]
But I don’t know.
[00:01:10]
If you’re on site in New York there,
[00:01:12]
you can do that.
[00:01:13]
Big prize money for that one.
[00:01:15]
But there’s also a community-driven one.
[00:01:18]
Allie, you want to talk about that one?
[00:01:20]
Yeah, as Kamille had just started saying,
[00:01:23]
the window for the community-led hackathon
[00:01:27]
opened on March first.
[00:01:30]
So this goes until March seventeenth.
[00:01:33]
And during that time, you can submit,
[00:01:36]
build and submit a Airtable base video
[00:01:42]
that goes along with it if you go
[00:01:44]
to the website,
[00:01:44]
which I’ll post it in the link right
[00:01:48]
here.
[00:01:49]
Perfect, thank you, Dan.
[00:01:53]
You can submit,
[00:01:54]
and all the rules are there.
[00:01:56]
There’s several prizes.
[00:01:57]
There’s several categories of prizes,
[00:02:00]
and one of those categories is a jury-led
[00:02:03]
selection.
[00:02:04]
And for the second time,
[00:02:06]
Kamille and I are working together on that
[00:02:08]
jury, along with Erika Pasquale,
[00:02:10]
who won two out of the three prizes
[00:02:12]
last time with her dodgeball base,
[00:02:16]
which was super, super cool.
[00:02:18]
So we’re really excited.
[00:02:20]
The theme this year,
[00:02:21]
we actually did a theme this time.
[00:02:23]
Last year it was wide open.
[00:02:25]
This time we chose a theme because a
[00:02:26]
lot of people submitted games.
[00:02:29]
Like there was like a huge percentage of
[00:02:32]
them were games or like just fun themed
[00:02:36]
bases or builds.
[00:02:38]
So the theme is called Let’s Play.
[00:02:40]
So anything that makes you think of or
[00:02:44]
inspiration that you drive from that.
[00:02:46]
We’d love to see.
[00:02:47]
And this is all really spearheaded by Mike
[00:02:50]
Simmons, I should say,
[00:02:51]
Mike Simmons of Automatic Nation.
[00:02:53]
Check him out on LinkedIn and check out
[00:02:56]
the website that Dan just posted in the
[00:02:58]
comments.
[00:03:00]
We’re excited.
[00:03:03]
Yeah.
[00:03:04]
Rob might have something to submit to
[00:03:07]
that.
[00:03:07]
Yeah.
[00:03:11]
I know I’m on the jury,
[00:03:12]
but I want to submit one as like
[00:03:14]
a,
[00:03:16]
we’re obviously not going to vote for this
[00:03:17]
one,
[00:03:18]
but I’ve already built my own sort of
[00:03:20]
game recreation in Airtable and just,
[00:03:22]
you know, for the funsies of it all.
[00:03:24]
Yeah.
[00:03:26]
There’s two other categories are the AI,
[00:03:28]
there’s an AI prize that’s chosen by
[00:03:30]
Airtable themselves.
[00:03:32]
And then there’s a community chosen prize.
[00:03:36]
So just the most liked submission.
[00:03:39]
So…
[00:03:40]
You could put it under a pseudonym,
[00:03:43]
Kamille.
[00:03:48]
Maybe we could do,
[00:03:49]
we’ve talked about this,
[00:03:50]
maybe we’ll do a special podcast and go
[00:03:54]
through what you built as maybe motivation
[00:03:58]
for others.
[00:03:58]
Yeah, I think that’d be really fun.
[00:04:04]
Yeah, so we’ll do something like that.
[00:04:05]
We’ll definitely highlight the
[00:04:07]
participants and winners.
[00:04:16]
So yeah, check that out,
[00:04:17]
hackathon.airtabletemplate.com.
[00:04:21]
The submission is there.
[00:04:22]
So anytime between now and March,
[00:04:26]
So be sure to submit yours and we’ll
[00:04:31]
highlight them once they are coming in.
[00:04:35]
Well, with that, we’ve got Rob on.
[00:04:36]
Rob has dived deep into the weeds of
[00:04:40]
AI and building.
[00:04:42]
And before we get into some of the
[00:04:45]
cool stuff that he’s been working on,
[00:04:46]
we thought we’d talk a little bit about
[00:04:48]
kind of the state of AI and some
[00:04:50]
of the projects that are out there.
[00:04:53]
So yeah, Rob,
[00:04:54]
what are your thoughts on having spent
[00:04:56]
some time in this world?
[00:04:58]
Yeah, definitely.
[00:05:01]
It’s definitely an exciting time.
[00:05:03]
It brings me back to ten plus years
[00:05:07]
ago when Airtable kind of first launched
[00:05:11]
and it was kind of the first of
[00:05:15]
these accessible,
[00:05:17]
easy to kind of use and understand
[00:05:21]
databases,
[00:05:22]
I think in general that like people could
[00:05:24]
have access to.
[00:05:25]
And it kind of, you know,
[00:05:26]
there was this progression between or from
[00:05:29]
Excel sheets to like Google sheets.
[00:05:31]
And then as soon as it was in
[00:05:32]
a Google sheet, it was like online.
[00:05:34]
And that was pretty awesome, you know?
[00:05:36]
And then there was FileMaker, which was,
[00:05:38]
you know, as we know, like,
[00:05:40]
the Scots of the world,
[00:05:41]
the Scott roses of the world,
[00:05:42]
like the geniuses, uh, behind FileMaker,
[00:05:45]
but then it, uh, you know,
[00:05:47]
then Airtable kind of became this like
[00:05:49]
always on database thing.
[00:05:51]
So I think we had this wave of
[00:05:53]
AI with like the chat GPTs and the
[00:05:56]
chat, um, interfaces with AI.
[00:05:59]
And I think we’re in this next wave
[00:06:01]
of AI.
[00:06:02]
We’re now,
[00:06:03]
now that we’re really kind of like, uh,
[00:06:06]
leaning more into like AGI and more like
[00:06:08]
actual agentic artificial intelligence.
[00:06:12]
I think it’s like this next big wave.
[00:06:15]
And so I’m just hoping that, you know,
[00:06:19]
I’m confident that it’s gonna be the
[00:06:22]
future.
[00:06:22]
It’s been extremely helpful for me.
[00:06:25]
And I’ve even been telling people like
[00:06:27]
with some of the,
[00:06:28]
like with the tools that are out there
[00:06:30]
and with the workflows that are out there
[00:06:32]
now, it feels like for the first time,
[00:06:35]
like computers and systems have been able
[00:06:37]
to actually keep up with my brain.
[00:06:39]
And I’ve been able to like, you know,
[00:06:41]
actually be able to like legitimately kind
[00:06:43]
of, you know,
[00:06:45]
work on multiple things at one time.
[00:06:47]
And whereas it would have been,
[00:06:50]
like a lot of people talk about context
[00:06:51]
overload and context in general,
[00:06:53]
when it comes to AI is,
[00:06:54]
very important to talk about um and to
[00:06:57]
understand but um i actually feel like
[00:07:00]
it’s it’s you know context switching is
[00:07:02]
extremely taxing uh but it’s nice to be
[00:07:05]
able to kind of for me at least
[00:07:06]
offload my ideas and like offload my
[00:07:09]
internal context and just put it into
[00:07:10]
somewhere else and just know that like
[00:07:15]
it’s there and i guess that’s like the
[00:07:17]
biggest that’s actually the biggest um
[00:07:20]
That’s the biggest thing that I’m trying
[00:07:21]
to actually tackle right now is like,
[00:07:23]
how do we have the trust and the
[00:07:24]
confidence to know that when we put
[00:07:26]
something into this AI slash like agent
[00:07:29]
world, how can we actually trust it?
[00:07:31]
And I think we’re all systems people here.
[00:07:34]
We’re all kind of like computer people.
[00:07:36]
And so when you…
[00:07:38]
talk to the computer,
[00:07:39]
you kind of expect the computer to do
[00:07:40]
what you ask it to do.
[00:07:42]
And that’s the beauty of computers.
[00:07:44]
It’s the beauty of like calculators and
[00:07:45]
things, you know, things like that.
[00:07:46]
So, you know,
[00:07:48]
I think that’s the biggest thing that we
[00:07:50]
all should be worried about right now is
[00:07:52]
like, you know,
[00:07:54]
it’s great that these AIs can take a
[00:07:55]
lot of authority for us on our behalf,
[00:07:58]
but like from a,
[00:08:03]
from a trust standpoint and from like a
[00:08:05]
risk standpoint, I don’t know,
[00:08:07]
there’s an inherent risk.
[00:08:10]
And when it comes to like risk mitigation
[00:08:12]
and all that stuff,
[00:08:12]
like I think that’s where a lot of
[00:08:14]
the conversation should be actually going
[00:08:15]
into now because the technology is,
[00:08:18]
is like growing exponentially and it’s
[00:08:20]
growing with itself, right?
[00:08:21]
Like AI is like self-improving itself.
[00:08:23]
And so I think we,
[00:08:24]
as the humans need to like be talking
[00:08:25]
a lot more about risk mitigation and talk
[00:08:27]
about like,
[00:08:29]
how do we control this uncontrollable
[00:08:30]
thing that yeah.
[00:08:33]
So that’s kind of my two cents,
[00:08:35]
I guess.
[00:08:39]
That makes sense to me.
[00:08:40]
I mean,
[00:08:41]
one of the biggest things when you work
[00:08:43]
as part of a team is you should
[00:08:45]
be able to check each other’s work just
[00:08:48]
to make sure everything comes out right.
[00:08:50]
And if you’re working with a team of
[00:08:52]
people, it’s, oh, you’re only human.
[00:08:54]
A mistake is going to come in there
[00:08:57]
at some point.
[00:08:59]
Today, AI is still built by humans.
[00:09:02]
So by its very nature,
[00:09:05]
it’s going to mess up every now and
[00:09:06]
again as well.
[00:09:07]
So as long as you can use it
[00:09:10]
as a tool and just do your own
[00:09:12]
due diligence at what it outputs for you,
[00:09:15]
you’re probably in the clear.
[00:09:17]
You just keep your one hand on the
[00:09:21]
steering wheel.
[00:09:22]
That’s where I’m at right now.
[00:09:23]
Yeah.
[00:09:26]
I’ve been kind of telling people that
[00:09:29]
we’re in this transition where we used to
[00:09:31]
have atlases and maps.
[00:09:34]
And then we got map quests.
[00:09:37]
And then we got tom-toms and garments and
[00:09:40]
things like that.
[00:09:40]
And then nowadays,
[00:09:44]
the maps are kind of the context.
[00:09:45]
And the maps are actually kind of
[00:09:46]
irrelevant.
[00:09:47]
What’s actually important is the
[00:09:48]
directions that we are taking.
[00:09:51]
And so maps are kind of becoming this
[00:09:53]
thing of the past.
[00:09:55]
but maps are what are fueling the
[00:09:57]
directions of like getting us from point A
[00:09:58]
to point B.
[00:09:59]
And we’ve inherently just trusted that
[00:10:01]
like when we get directions to something,
[00:10:03]
it’s gonna take us to where we wanna
[00:10:05]
get to.
[00:10:07]
And I think that like,
[00:10:09]
not there yet with ai but we’re at
[00:10:11]
the point where we know that ai can
[00:10:13]
define a couple different ways to get to
[00:10:15]
an end point um we just need to
[00:10:17]
like influence it a little bit so i
[00:10:19]
kind of like think of this analogy of
[00:10:21]
like taxi drivers you know like there’s
[00:10:23]
taxi drivers and then there’s like a waymo
[00:10:25]
or a subway or something it’s like the
[00:10:28]
subway’s only gonna go
[00:10:31]
this like line of the subway you know
[00:10:33]
like it’s not going to take a a
[00:10:34]
left or a rights to like miss something
[00:10:38]
you know uh unless it’s given some
[00:10:39]
direction to to kind of like reroute
[00:10:41]
itself whereas like a taxi driver who
[00:10:43]
knows the back streets or you know knows
[00:10:46]
its way to get around town and knows
[00:10:48]
how to get around traffic and knows when
[00:10:49]
school’s about to be out and all that’s
[00:10:51]
like um yeah there’s that human aspect to
[00:10:55]
it of that like human intuition uh
[00:10:59]
But there’s a ceiling to that,
[00:11:01]
and there’s a ceiling to, like,
[00:11:03]
a taxi driver doesn’t necessarily know if
[00:11:04]
there’s a traffic accident, you know,
[00:11:07]
one mile ahead because that’s happening in
[00:11:10]
real time,
[00:11:11]
and they might not have access to, like,
[00:11:12]
that real-time information.
[00:11:14]
And that’s where, like,
[00:11:15]
the computers and the systems come into
[00:11:17]
play.
[00:11:18]
Yeah.
[00:11:20]
My thought was a question for you is,
[00:11:23]
so Rob is the one that built the
[00:11:25]
Olympic Games app.
[00:11:27]
We highlighted it a couple of weeks ago.
[00:11:29]
Very impressive, very cool stuff.
[00:11:32]
BuiltOnAirtable,
[00:11:34]
but vibe coded or AI generated on top
[00:11:38]
of that.
[00:11:40]
Where do you see the Airtables of the
[00:11:43]
world fitting in this world of AI?
[00:11:46]
It’s very interesting dynamic.
[00:11:48]
What are your thoughts there?
[00:11:50]
Yeah, I mean,
[00:11:51]
my thoughts are when it comes to AI,
[00:11:55]
it’s all about the context.
[00:11:56]
It’s all about the information that we’re
[00:11:58]
giving this AI in order to get an
[00:12:00]
output.
[00:12:01]
from the AI and the better the information
[00:12:04]
that we can give it,
[00:12:05]
usually the better the information that
[00:12:07]
comes out of it.
[00:12:08]
And that’s just kind of like a system
[00:12:09]
in general, right?
[00:12:10]
Like the better the plan for something is
[00:12:12]
the better.
[00:12:13]
Usually the outcome is of whatever,
[00:12:16]
like whatever goal you’re trying to
[00:12:17]
accomplish.
[00:12:19]
And so I think Airtable is uniquely
[00:12:21]
positioned as an established single source
[00:12:25]
of truth.
[00:12:27]
for information, for businesses,
[00:12:29]
for personal use, for whatnot.
[00:12:32]
Like it’s established itself as a single
[00:12:34]
source of truth.
[00:12:35]
And in this AI world,
[00:12:36]
like where we have to have the confidence
[00:12:40]
of what is true.
[00:12:42]
Like I’ve been experimenting and playing
[00:12:44]
around with like open claw rights,
[00:12:46]
for example.
[00:12:48]
which is this like incredible AI open
[00:12:50]
source, like personal assistance,
[00:12:53]
quote unquote, right?
[00:12:54]
And one of the things that I tried
[00:12:57]
to have it do for me was research
[00:13:00]
some topics and some people that I wanted
[00:13:02]
to kind of like on a daily basis,
[00:13:04]
give me an update and like,
[00:13:06]
let me know if anything is new has
[00:13:08]
happened in the news throughout the night
[00:13:10]
or something, right?
[00:13:11]
And so you kind of just like chat
[00:13:13]
to it and you’re like,
[00:13:13]
give it to them.
[00:13:14]
And you’re just like, here, do this.
[00:13:15]
This is what I want, you know?
[00:13:20]
And then you wake up the next morning
[00:13:21]
and the thing you asked for isn’t there.
[00:13:25]
Your research isn’t there.
[00:13:28]
And it’s like, why is it not here?
[00:13:34]
It’s there,
[00:13:34]
but it’s only there for two of the
[00:13:36]
ten people.
[00:13:38]
And it would be nice to just have…
[00:13:44]
a table of records of people that I
[00:13:47]
wanna keep track of.
[00:13:49]
And whenever I wanna add a new person,
[00:13:51]
I just have to like add a new
[00:13:52]
record to a table.
[00:13:53]
Or whenever I tell it to do something,
[00:13:55]
there’s some sort of like visual feedback
[00:13:57]
and confirmation of like, okay,
[00:13:59]
this is the list of people that I’m
[00:14:01]
gonna like research every single night.
[00:14:03]
And so taking that back to like the
[00:14:05]
Airtable side of things,
[00:14:06]
I think Airtable as it relates to AI,
[00:14:09]
like big AI in general,
[00:14:11]
is very,
[00:14:12]
very uniquely positioned to already be
[00:14:14]
that single source of truth and already be
[00:14:16]
in a,
[00:14:17]
like in businesses and organizations as an
[00:14:22]
organized, like organized context.
[00:14:25]
And so I think if we can figure
[00:14:26]
out how to like harness that or that,
[00:14:29]
that context and harness the information
[00:14:31]
that is already being interpreted as the
[00:14:35]
truth, um,
[00:14:38]
I think it’s got legs to actually be
[00:14:42]
a leader in the AI space.
[00:14:46]
Yeah, that’s my thought.
[00:14:48]
Yeah.
[00:14:49]
Yeah, it’s really interesting.
[00:14:53]
Yeah,
[00:14:53]
just this like hybrid approach of like the
[00:14:55]
no code aspect of the Airtable with the,
[00:14:58]
especially like the interfaces and then
[00:15:02]
putting the custom.
[00:15:04]
Cause like for your example with the
[00:15:06]
Olympic games,
[00:15:07]
in theory you could have built that all
[00:15:09]
standalone, you know, on its own.
[00:15:12]
Like what would be the pros and cons
[00:15:14]
of taking that approach versus doing what
[00:15:16]
you did to build it on Airtable?
[00:15:18]
Yeah.
[00:15:19]
I mean, it was honestly,
[00:15:21]
it was about speed.
[00:15:22]
Like it was about pace.
[00:15:22]
It was about speed.
[00:15:23]
It was like,
[00:15:24]
how quickly can I actually create
[00:15:25]
something that’s valuable?
[00:15:27]
And that’s something that’s not specific.
[00:15:30]
Like that concept is not specific to air
[00:15:32]
table or sorry to AI.
[00:15:33]
Like how do I build something as valuable
[00:15:35]
as possible?
[00:15:36]
And like in as least amount of time
[00:15:38]
is, is not unique.
[00:15:41]
And so a lot of the, like,
[00:15:44]
when I was thinking about how I was
[00:15:46]
going to build this,
[00:15:48]
and a lot of times when I think
[00:15:49]
about these like complex problems,
[00:15:50]
I always try to boil it down and
[00:15:52]
break it down into like, okay,
[00:15:53]
like what’s, what’s simpler here.
[00:15:55]
Like what’s already been done.
[00:15:56]
What’s the existing mental model that
[00:15:58]
people are already used to when it comes
[00:16:00]
to looking at data visualized for the
[00:16:04]
Olympics.
[00:16:05]
Um, and you know, so,
[00:16:08]
so thinking about like, okay,
[00:16:10]
metal counts are usually pretty important
[00:16:12]
and thinking about like, uh, news,
[00:16:15]
thinking about, um,
[00:16:17]
you know, like all this stuff, it’s like,
[00:16:20]
it’s essentially data.
[00:16:21]
Like all this is just data.
[00:16:22]
And so I was like, okay,
[00:16:23]
I can put this all into Airtable.
[00:16:25]
I can organize this all.
[00:16:26]
Cause I have the muscle memory myself of
[00:16:27]
like establishing an Airtable base or a
[00:16:31]
database in general for this.
[00:16:34]
But I needed to move fast.
[00:16:35]
Like,
[00:16:36]
cause the like opening ceremonies were in
[00:16:38]
like three days or two days from the
[00:16:40]
time I started building it.
[00:16:41]
Yeah.
[00:16:43]
And so, but I had work to do,
[00:16:46]
like I had other work to do and
[00:16:47]
everything.
[00:16:48]
So I really kind of, you know,
[00:16:50]
you reach into your toolbox and at the
[00:16:52]
same time, you like,
[00:16:54]
one of the things that I always say
[00:16:56]
is like the most valuable lesson that I
[00:16:58]
learned from my like amazing four year
[00:17:01]
university degree that I was fortunate
[00:17:02]
enough to get was not my like computer
[00:17:05]
science degree.
[00:17:05]
It was, I learned how to learn things.
[00:17:10]
And I think that like,
[00:17:12]
that couldn’t be more true and more
[00:17:13]
valuable right now than like it was,
[00:17:16]
I don’t know,
[00:17:17]
fifteen years ago of like learning how to
[00:17:19]
learn things is actually extremely
[00:17:22]
important.
[00:17:22]
And when we had like our,
[00:17:25]
like whenever we had exams or projects or
[00:17:28]
anything, it was all like open book,
[00:17:30]
open online resources, open friends,
[00:17:33]
open whatever.
[00:17:34]
Like it was use the tools that are
[00:17:35]
available to get to like the desired
[00:17:39]
outcome within some constraints.
[00:17:41]
And so it was nice to have some
[00:17:43]
constraints, but at the same time,
[00:17:44]
it was nice to be able to reach
[00:17:46]
into my toolbox as well as say,
[00:17:47]
what things can I add to my toolbox
[00:17:49]
to make my toolbox more valuable?
[00:17:51]
And that’s where I really kind of just
[00:17:53]
leaned heavily on cloud code.
[00:17:55]
I leaned heavily on skills within cloud
[00:17:58]
code.
[00:17:59]
I leaned heavily on MCP servers to be
[00:18:03]
able to kind of use AI as this
[00:18:06]
like thought partner.
[00:18:07]
And then it was like a constant back
[00:18:10]
and forth of,
[00:18:12]
me kind of doing some work,
[00:18:13]
handing it off, it doing some work,
[00:18:16]
handing it back off to me,
[00:18:18]
me doing some work, handing it off.
[00:18:19]
And so I think like there’s still a
[00:18:21]
lot of value in the human aspect and
[00:18:23]
the human involved in this process.
[00:18:25]
So yeah, I don’t know.
[00:18:27]
Cool.
[00:18:29]
Cool.
[00:18:29]
Any other topics or questions before we
[00:18:32]
move on to some of the stuff you’ve
[00:18:34]
been working on?
[00:18:35]
I had an urban planning analogy.
[00:18:38]
At some point,
[00:18:39]
you reminded me of my first sort of
[00:18:41]
foray into the world of AI was with
[00:18:44]
self-driving cars.
[00:18:46]
And urban planners were sort of tasked
[00:18:49]
with thinking of how does this affect how
[00:18:51]
we write rules about what you can do
[00:18:53]
in a city and where do cars go?
[00:18:55]
And at one point,
[00:18:57]
I don’t know what the general consensus is
[00:18:59]
anymore, but we were like,
[00:19:01]
well you still got to park the car
[00:19:02]
somewhere you can’t eliminate the concept
[00:19:05]
of garages unless you build them like the
[00:19:08]
perpetual train from snowpiercer they have
[00:19:10]
to stop at some point and um the
[00:19:15]
interplay between Airtable and ai
[00:19:18]
there’s always going to be value in things
[00:19:20]
like Airtable of I just need to look
[00:19:22]
at a list of stuff really quick.
[00:19:24]
And then Airtable does a lot of other
[00:19:26]
things nowadays,
[00:19:27]
especially with interfaces.
[00:19:29]
We’re going to see some that Rob built
[00:19:31]
completely custom using AI to help.
[00:19:33]
But
[00:19:35]
no matter what you’re building with AI,
[00:19:36]
you’re probably going to have some data
[00:19:38]
that you need to store somewhere.
[00:19:40]
And it’s helpful if your database is
[00:19:43]
already AI enabled.
[00:19:44]
So you get the leg up there of
[00:19:46]
not having to get several tools to sort
[00:19:50]
of jam together to make the thing that
[00:19:53]
you’re making.
[00:19:54]
Airtable is pretty well suited for certain
[00:19:56]
use cases to do most of what you
[00:19:58]
need off the top.
[00:20:00]
Yeah.
[00:20:04]
Yeah.
[00:20:05]
Yeah, I’m excited for what,
[00:20:06]
what comes next.
[00:20:07]
I mean,
[00:20:07]
the deep match thing is super exciting.
[00:20:11]
Like I kind of, I just,
[00:20:13]
I can’t wait until I can have it
[00:20:15]
like make decisions based off of the whole
[00:20:18]
data set, like actionable decisions.
[00:20:21]
Um, but yeah, I could go on.
[00:20:25]
I’m excited to see what you’ve got to
[00:20:27]
show Rob.
[00:20:29]
Yeah.
[00:20:29]
So we could, so let’s see,
[00:20:32]
why don’t we start with a hyper agent?
[00:20:35]
Let’s start with HyperAgent if you wanna
[00:20:37]
get that ready.
[00:20:38]
So HyperAgent,
[00:20:39]
we talked about a week or two ago,
[00:20:42]
recently released.
[00:20:44]
Well, not released,
[00:20:45]
it’s still in private beta.
[00:20:46]
So this might be a peak.
[00:20:48]
It’ll be the first time I’ve seen it.
[00:20:50]
I signed up for the wait list and
[00:20:52]
I’m not yet added.
[00:20:53]
So I’m excited to see this.
[00:20:55]
So Rob, if you wanna share your screen.
[00:20:58]
Get it on there.
[00:21:01]
And so, yeah, so this is,
[00:21:03]
this one we talked about,
[00:21:05]
this one is under the name of Airtable
[00:21:08]
and it says it’s from Airtable where
[00:21:10]
SuperAgent is kind of under its own name.
[00:21:12]
It didn’t have the Airtable owner on it.
[00:21:17]
So HyperAgent.
[00:21:18]
Yeah.
[00:21:18]
Yeah.
[00:21:21]
Okay, here it is.
[00:21:22]
I was gonna try and like pull open
[00:21:24]
the like website for it.
[00:21:27]
I could probably just log out real quick.
[00:21:32]
There’s a distinct difference between
[00:21:35]
hyper-agent and super-agent.
[00:21:37]
A hyper-agent
[00:21:39]
Let’s actually talk about Super Agent real
[00:21:41]
quick.
[00:21:41]
Super Agent is this idea of essentially
[00:21:46]
this research assistance.
[00:21:48]
A lot of what’s actually been put into
[00:21:52]
Airtable’s built-in AI functionality and
[00:21:55]
features in terms of the field agents,
[00:21:58]
in terms of those individual research
[00:22:01]
components of Airtable is actually a lot
[00:22:04]
of what Super Agent is in the background.
[00:22:07]
And Super Agent was
[00:22:09]
deep sky,
[00:22:10]
which was a company that they acquired.
[00:22:12]
And so, you know,
[00:22:14]
super agent in its pieces is in air
[00:22:16]
table and different pieces,
[00:22:17]
but super agent as a tool and as
[00:22:19]
a product is like all of those things
[00:22:23]
put together.
[00:22:24]
And so you can create like slide decks
[00:22:26]
and you can create like reports and things
[00:22:28]
like that.
[00:22:28]
Right.
[00:22:28]
So it’s like,
[00:22:29]
it’s more of this like research agent for
[00:22:32]
you.
[00:22:33]
Whereas hyper agents are,
[00:22:36]
Hyperagent is really this idea and this
[00:22:40]
concept of the orchestration layer of AI
[00:22:45]
agents at an
[00:22:50]
enterprise level and so it’s it’s
[00:22:53]
something that you know that howie is
[00:22:55]
working on personally and some of his
[00:22:57]
other um team members on his like is
[00:23:00]
the ai team at Airtable and so
[00:23:03]
if we log into this you’ll see um
[00:23:06]
like a lot of a lot of the
[00:23:07]
commentary on hyper agent that they’re
[00:23:10]
saying is okay it’s great to have all
[00:23:12]
these ai agents but if you don’t have
[00:23:15]
this orchestration layer behind all these
[00:23:17]
AI agents to actually know and control
[00:23:20]
what’s going on,
[00:23:21]
then it becomes
[00:23:23]
you know, noisy and hectic essentially.
[00:23:26]
And so they’ve, you know,
[00:23:27]
kind of come up with some like recommended
[00:23:29]
agents up at the top here.
[00:23:31]
And essentially what’s going on in the
[00:23:33]
background of hyper agents is it’s linked
[00:23:35]
to a tool called Composio.
[00:23:38]
And Composio is essentially this like tool
[00:23:41]
that allows you to connect your agents to
[00:23:43]
a whole bunch of different things.
[00:23:44]
It’s like a whole bunch of different MCPs.
[00:23:47]
So it’s kind of like an MCP as
[00:23:49]
a service.
[00:23:50]
And so
[00:23:51]
when you hop in here you can see
[00:23:53]
this integrations table here and you can
[00:23:55]
see like all of these integrations so i
[00:23:58]
can browse like a thousand integrations um
[00:24:01]
sorry a thousand integrations here so you
[00:24:03]
can see like perplexity google docs
[00:24:05]
hubspot etc and so the cool thing about
[00:24:08]
hyper agents is that similar to things
[00:24:11]
like open claw and similar to things
[00:24:13]
there’s a couple other tools out there um
[00:24:15]
like twin tasklets and even your like open
[00:24:19]
AI, your chat GPTs and your clods,
[00:24:21]
they do have some ability to connect to
[00:24:24]
MCP servers,
[00:24:26]
to be able to kind of do AI
[00:24:27]
things with integrations and whatnot.
[00:24:31]
But HyperAgent is cool in that it’s this
[00:24:34]
like layer
[00:24:37]
on top of all that to just do
[00:24:39]
agentic things.
[00:24:40]
So I actually just like,
[00:24:41]
I created an agent here,
[00:24:43]
which is a BuiltOnAir podcast prep
[00:24:45]
agents.
[00:24:45]
And so you can see like, it’s a,
[00:24:48]
it’s an actual age.
[00:24:49]
Like this is an actual agent.
[00:24:50]
This is an actual AI agents.
[00:24:53]
So it’s not just a chat window.
[00:24:56]
It’s not just like,
[00:24:58]
this is an actual agent and we can
[00:25:00]
see that this agent has usage, right?
[00:25:02]
So it’s actually been like,
[00:25:03]
calling certain AI endpoints.
[00:25:08]
It’s giving us the cost of those
[00:25:09]
endpoints.
[00:25:10]
And then I had it…
[00:25:13]
You can see here,
[00:25:14]
you can hook it up to Slack.
[00:25:15]
You can schedule jobs.
[00:25:17]
You can have it be hooked up to
[00:25:19]
webhooks so that whenever anything comes
[00:25:21]
into this, it’ll do something.
[00:25:23]
Again, email, SMS, like MC…
[00:25:25]
You can create your own MCP server for
[00:25:28]
this agent specifically.
[00:25:30]
So it’s kind of like…
[00:25:31]
it’s kind of trying to extrapolate all
[00:25:33]
these individual pieces that are going
[00:25:35]
into AI agents and giving it back to
[00:25:38]
the user to say, hey,
[00:25:40]
you can actually create your own.
[00:25:42]
Again,
[00:25:42]
taking the analogy of you can have your
[00:25:45]
Excel spreadsheet if you want to,
[00:25:47]
but if you want to have
[00:25:49]
your data structured, always online,
[00:25:53]
you know, always secure.
[00:25:56]
You don’t have to worry about all that
[00:25:57]
stuff.
[00:25:57]
Like then use Airtable for that.
[00:25:59]
As one example,
[00:26:00]
this is their kind of answer.
[00:26:02]
This is Airtable’s answer to,
[00:26:04]
at least this is my understanding.
[00:26:05]
This is my interpretation,
[00:26:06]
just to be clear.
[00:26:07]
This is my interpretation of it as well.
[00:26:11]
And so I just did like,
[00:26:14]
one quick research agent as well.
[00:26:18]
So this is the research agent for the
[00:26:20]
BuiltOnAir like agent itself.
[00:26:22]
And it actually went through and it found
[00:26:25]
like a whole bunch of like five potential
[00:26:29]
guests.
[00:26:30]
So, you know, Chris Chang,
[00:26:31]
who’s the deep sky CEO, Alan Barter,
[00:26:34]
who’s head of AI at Airtable,
[00:26:37]
And then some builders as well,
[00:26:38]
some like builders and company,
[00:26:40]
Ben Green got put in here and Dimitri
[00:26:43]
Shapiro.
[00:26:44]
And so HyperAgent actually went out and
[00:26:47]
did all of this for me on my
[00:26:51]
behalf.
[00:26:51]
And then you can even, you know,
[00:26:53]
create websites from this.
[00:26:54]
You can create videos,
[00:26:55]
you can create images, audio, slides,
[00:26:57]
tables, et cetera.
[00:26:58]
So it’s pretty cool.
[00:27:00]
It’s very much in beta, but yeah,
[00:27:02]
happy to kind of answer any questions or
[00:27:04]
poke around if you guys want to see
[00:27:05]
anything else.
[00:27:08]
Yeah, it’s cool.
[00:27:09]
It’s like,
[00:27:10]
so it’s more of a toolkit that you
[00:27:13]
can build on top of.
[00:27:15]
Does it connect to Airtable?
[00:27:18]
Yeah, so you can connect it to Airtable.
[00:27:21]
So you can feature connection,
[00:27:22]
connect to Airtable.
[00:27:23]
So you can connect it to Airtable,
[00:27:24]
but you can also connect it to like
[00:27:26]
Databricks and Snowflake and some other,
[00:27:29]
yeah,
[00:27:31]
some other like data sources as well.
[00:27:33]
Linear.
[00:27:35]
Yeah, linear, go linear.
[00:27:36]
Connected, there it is.
[00:27:40]
If you had a particular agent and you
[00:27:42]
wanted it to be focused on a particular
[00:27:45]
data set, that integration,
[00:27:47]
can you be like, for this agent,
[00:27:50]
use this set of data?
[00:27:52]
Exactly, yeah.
[00:27:53]
So you can see this agent has access
[00:27:56]
to do all of these things.
[00:27:58]
You can see Snowflake and Databricks down
[00:28:00]
here as well.
[00:28:01]
And then…
[00:28:04]
There is a way I haven’t poked around
[00:28:06]
and played as much as I,
[00:28:07]
as I should.
[00:28:09]
But yeah,
[00:28:10]
there is a world where you then can
[00:28:12]
say this agent, at least I would like,
[00:28:14]
that would make sense.
[00:28:15]
Like that’s kind of one of the biggest,
[00:28:17]
the bigger concepts that I’ve been,
[00:28:19]
admittedly like struggling to handle and
[00:28:22]
manage is all these different projects
[00:28:24]
that I’m working on.
[00:28:25]
I want to make sure that the individual
[00:28:27]
projects have guard rails to them in that
[00:28:30]
they only have access to certain MCP
[00:28:32]
servers, certain API endpoints,
[00:28:35]
certain tokens.
[00:28:37]
Um,
[00:28:38]
and like there isn’t cross pollination.
[00:28:40]
between all that.
[00:28:41]
And that’s definitely something that I’ve
[00:28:42]
been seeing as a huge risk for people
[00:28:45]
that are just jumping right into things
[00:28:47]
like OpenClaw or Cloud Code or whatnot.
[00:28:50]
And they’re just kind of feeding it,
[00:28:52]
you know,
[00:28:53]
access tokens and they’re feeding it,
[00:28:55]
you know, all these different things.
[00:28:57]
And if you’re not
[00:29:00]
If you’re not neat and organized,
[00:29:03]
you can very easily not realize that your
[00:29:06]
AI actually has access to something that
[00:29:09]
you don’t want it to have access to.
[00:29:11]
And then that’s kind of where it gets
[00:29:13]
back to the trust and the guardrails of
[00:29:15]
this whole thing that I’m really excited
[00:29:19]
to dive into and try and do my
[00:29:21]
part to help out with that.
[00:29:23]
Yeah,
[00:29:24]
the principles of app design haven’t
[00:29:26]
really changed whether you build it
[00:29:28]
yourself or you use AI to help you.
[00:29:30]
You can’t just be giving out API tokens
[00:29:34]
and things like that willy-nilly.
[00:29:36]
You absolutely cannot just grant an agent
[00:29:41]
or a person, hey, yeah,
[00:29:42]
look at all of my databases.
[00:29:44]
Get full right access to everything I own.
[00:29:48]
Don’t do that.
[00:29:50]
Recipe for disaster.
[00:29:52]
Oh my goodness.
[00:29:54]
Did you see the thing about a guy
[00:29:55]
that like tried to like vibe code himself
[00:29:58]
a little app to control his robot vacuum
[00:30:01]
and he ended up with control of like
[00:30:03]
all the robot vacuums on the planet or
[00:30:05]
something?
[00:30:06]
Yeah.
[00:30:07]
From this what company?
[00:30:11]
Yep.
[00:30:11]
Was that news correct?
[00:30:12]
Did I miss?
[00:30:13]
Did OpenAI buy OpenClaw?
[00:30:16]
Oh, yeah.
[00:30:17]
So this is a daily brief.
[00:30:18]
Yeah, go ahead.
[00:30:19]
Go ahead.
[00:30:21]
I was saying I thought they did.
[00:30:21]
Yeah, I thought I read that.
[00:30:23]
I didn’t hear that.
[00:30:25]
They bought him.
[00:30:26]
They bought Peter Steinberger,
[00:30:27]
essentially.
[00:30:29]
OpenClaw itself is going to be owned by
[00:30:32]
a foundation.
[00:30:35]
So it’s going to stay open source.
[00:30:38]
But OpenClaw is going to be owned by
[00:30:41]
a foundation specifically for OpenClaw.
[00:30:45]
Peter now works in some fashion for OpenAI
[00:30:49]
to help push their personal agents kind of
[00:30:53]
agenda.
[00:30:54]
but yeah,
[00:30:55]
this is a daily agent brief that like
[00:30:57]
happened yesterday.
[00:30:58]
Um, you know,
[00:30:59]
and some of this stuff wasn’t breaking.
[00:31:01]
I like this happened like a week or
[00:31:03]
two ago.
[00:31:04]
Um, this happened last week.
[00:31:06]
Um, but it is very like,
[00:31:08]
this is a,
[00:31:09]
an agent brief that I told it to
[00:31:12]
make for me about AI agents in general.
[00:31:16]
Um, and this is very accurate.
[00:31:17]
Like Manus is top of like the headlines
[00:31:19]
right now.
[00:31:20]
Perplexity is top of the headlines right
[00:31:22]
now.
[00:31:23]
Um,
[00:31:24]
it’s doing like a comparison on a whole
[00:31:26]
bunch of them as well,
[00:31:27]
which is really helpful for me.
[00:31:29]
So this is actually something that like is
[00:31:31]
useful.
[00:31:32]
And I think it’s actually a website like,
[00:31:35]
or no, it’s yeah.
[00:31:36]
So I could turn this into a website,
[00:31:37]
but yeah,
[00:31:37]
there’s like a whole brief here that I’ve
[00:31:39]
given it to build this for me.
[00:31:44]
You’ll be curious how,
[00:31:45]
cause it seems like super agent
[00:31:48]
functionality is built into hyper agent.
[00:31:51]
There’s, you know,
[00:31:52]
like you could almost kind of do what
[00:31:55]
super agent does inside of here.
[00:31:58]
It’d be interesting to see how they
[00:31:59]
support both those products.
[00:32:02]
Totally.
[00:32:03]
Yeah.
[00:32:04]
Totally.
[00:32:05]
A lot of people are talking about command
[00:32:06]
centers as well with AI agents right now.
[00:32:08]
Like, how do you,
[00:32:10]
how do you keep track of all of
[00:32:11]
your agents?
[00:32:11]
Like,
[00:32:11]
so there’s like a running list of like
[00:32:13]
all of your agents here.
[00:32:14]
What things like need your attention is
[00:32:18]
like super important.
[00:32:19]
Yeah.
[00:32:22]
Let’s see what this says here.
[00:32:26]
HyperAgent is actually really cool in that
[00:32:28]
as you start to talk to it,
[00:32:30]
it starts to suggest some things.
[00:32:33]
Should we have a guest brief?
[00:32:35]
Yeah, that sounds cool.
[00:32:36]
Let’s take that.
[00:32:37]
Should we do a, what else?
[00:32:42]
Yeah, that’s cool.
[00:32:43]
Let’s dismiss that.
[00:32:44]
Again,
[00:32:45]
it’s just this layer into your AI agents.
[00:32:50]
like doesn’t make it as like a theory
[00:32:53]
and like theoretical, it’s much more like,
[00:32:56]
okay,
[00:32:56]
this is actually what’s fueling my agents
[00:32:59]
right now.
[00:33:01]
Um,
[00:33:02]
which is exciting to like have that
[00:33:04]
visibility.
[00:33:05]
Um,
[00:33:05]
and then it gets into things like rubrics
[00:33:07]
and evals,
[00:33:08]
which is like when you talk about LLMs
[00:33:10]
in general and training LLMs and a lot
[00:33:12]
of people are talking about, well,
[00:33:14]
I could use mini max or some of
[00:33:16]
these other, um,
[00:33:18]
like open source or local models, uh,
[00:33:21]
versus using open AI or anthropics models.
[00:33:25]
And you can certainly do that.
[00:33:27]
And then you’re,
[00:33:27]
you aren’t necessarily burning cloud
[00:33:29]
tokens to, to, to, to do that.
[00:33:31]
But, um, you know, you,
[00:33:35]
if you’re running your own local models,
[00:33:38]
it’s very important for you to train your
[00:33:39]
own local models, because if it’s local,
[00:33:42]
it means no one has access to it.
[00:33:44]
There’s,
[00:33:44]
there’s training that has gone into that
[00:33:46]
model in, in general,
[00:33:47]
but
[00:33:48]
it’s not,
[00:33:49]
it’s only being trained on what you’re
[00:33:51]
training it on.
[00:33:52]
And so AI in general is like,
[00:33:55]
it’s only,
[00:33:55]
it only gets better if it knows what’s
[00:33:58]
right and what’s wrong.
[00:34:00]
And so that comes, you know, that,
[00:34:01]
that goes into the idea of like these
[00:34:02]
evaluations and rubrics and things.
[00:34:04]
So it’s,
[00:34:05]
it’s cool that like a lot of these
[00:34:06]
core fundamental concepts of AI are built
[00:34:10]
into hyper agents to kind of
[00:34:12]
learn about AI in general without having
[00:34:14]
to go out and just do research.
[00:34:16]
And you can kind of learn by clicking
[00:34:17]
buttons, which is fun.
[00:34:20]
This kind of reminds me of like the
[00:34:22]
more like analog, so to speak,
[00:34:25]
parallel in my brain is like,
[00:34:28]
and just that page in particular,
[00:34:29]
the rubric thing.
[00:34:31]
And like, you know,
[00:34:33]
we’ve done like an email parser,
[00:34:34]
like a Zapier parser.
[00:34:36]
And like,
[00:34:36]
you can go in and just like keep
[00:34:37]
redoing the template over and over again
[00:34:39]
to like train it to be like, no,
[00:34:41]
no, no.
[00:34:41]
Like it’s this line,
[00:34:42]
like over and over again,
[00:34:44]
like it keeps getting it wrong.
[00:34:46]
You have to keep highlighting the token,
[00:34:47]
like.
[00:34:48]
I don’t know.
[00:34:49]
It kind of reminds me of that.
[00:34:50]
It’s like nudging it in that direction.
[00:34:54]
Definitely.
[00:34:55]
Have they announced pricing for this or is
[00:34:58]
it free during beta?
[00:35:00]
Right now it’s free.
[00:35:01]
I see you spent one dollar and forty
[00:35:03]
six cents.
[00:35:03]
Yeah.
[00:35:06]
Where’s my one dollar and forty six cents?
[00:35:07]
Yeah.
[00:35:08]
I believe it’s free during beta right now.
[00:35:11]
Where’s the pricing?
[00:35:12]
There was a billing
[00:35:16]
early adopter, free to use hyper agents.
[00:35:19]
Appreciate your feedback.
[00:35:20]
There’s a pricing details for you right
[00:35:21]
there.
[00:35:21]
So kind of similar standard.
[00:35:24]
So pretty,
[00:35:24]
pretty standard AI price points.
[00:35:30]
Yeah.
[00:35:30]
They’ve got to pay as you go as
[00:35:31]
well.
[00:35:31]
Interesting.
[00:35:34]
Cool.
[00:35:35]
Yeah.
[00:35:35]
This will be interesting.
[00:35:36]
Cool.
[00:35:36]
Yeah.
[00:35:36]
I’m excited to check it out.
[00:35:38]
I don’t know how they’re deciding who gets
[00:35:41]
access.
[00:35:41]
I know in our community, people are like,
[00:35:43]
I’m in and others are like, I’m not.
[00:35:45]
So I don’t know if it’s random or
[00:35:48]
Rob’s on the, on the VIP list.
[00:35:53]
I’m on the, I’m, I’m begging.
[00:35:55]
I’m at the front door, like knocking,
[00:35:57]
you know, not leaving.
[00:36:00]
Yeah.
[00:36:02]
Awesome.
[00:36:02]
Well, let’s move on.
[00:36:03]
Okay.
[00:36:04]
I know let’s look at some stuff you’ve
[00:36:06]
built.
[00:36:06]
So you built some cool stuff.
[00:36:08]
I think it might be worthwhile to explain,
[00:36:12]
you know, the technical aspects.
[00:36:14]
Yeah.
[00:36:14]
Why don’t we start here?
[00:36:15]
Talk about, we mentioned it,
[00:36:17]
but show us what you built.
[00:36:19]
Okay, cool.
[00:36:21]
Yeah, so what you’re looking at here,
[00:36:23]
if I just launch this just so it
[00:36:24]
looks as pretty as possible,
[00:36:26]
because that’s important,
[00:36:28]
is this like Airtable Fantasy Olympics
[00:36:32]
custom interface.
[00:36:34]
And this is live online.
[00:36:37]
One of the cool things about interface
[00:36:39]
pages in general is that some interface
[00:36:41]
pages have the ability to be published to
[00:36:43]
the web and you don’t have to have
[00:36:44]
Airtable access to it,
[00:36:45]
which is super cool.
[00:36:47]
Um,
[00:36:47]
custom interfaces is one of those
[00:36:48]
interfaces that you can do that.
[00:36:50]
Um,
[00:36:50]
and so everything you’re seeing here was
[00:36:53]
all, as you guys already talked about,
[00:36:54]
like it was all vibe coded.
[00:36:56]
Um, and it has, uh, you know, the,
[00:36:59]
you know, upcoming, um,
[00:37:00]
like the actual games themselves, uh,
[00:37:03]
the countries themselves, uh,
[00:37:05]
what medals they’ve, you know, they’ve,
[00:37:07]
they’ve won as well as on the right-hand
[00:37:09]
side here, we can see there was, um,
[00:37:12]
like this opportunity to make your picks
[00:37:15]
to see who you thought were going to
[00:37:18]
win what medals and everything.
[00:37:20]
And so it’s kind of cool to see
[00:37:22]
that the community has actually been like
[00:37:24]
engaging here.
[00:37:25]
I probably should vibe code this to like
[00:37:28]
say the Olympics are over.
[00:37:31]
But as of right now,
[00:37:33]
it’s all kind of just live like this.
[00:37:35]
And then it was cool.
[00:37:36]
The idea behind this whole,
[00:37:37]
this whole thing was not only to like,
[00:37:39]
create an inspiration of like, Holy cow.
[00:37:42]
Like this is all built on top of
[00:37:44]
Airtable,
[00:37:44]
but also that it’s possible to build this
[00:37:47]
on Airtable and it’s possible to, um,
[00:37:52]
Whenever we’re building in general,
[00:37:54]
it all starts in the data and having
[00:37:56]
clean data is always a luxury.
[00:37:59]
And so what we ended up doing at
[00:38:01]
the bottom here is giving all these data
[00:38:06]
sets out for anyone to use.
[00:38:09]
So you could open up and sync any
[00:38:11]
of the tables that…
[00:38:14]
eye builds um and are keeping up to
[00:38:16]
date and build something like build your
[00:38:19]
own custom interface with it so it was
[00:38:21]
kind of cool to see um like someone
[00:38:23]
created like a comic strip uh generator
[00:38:25]
and we should be able to actually open
[00:38:26]
it up as well um and these are
[00:38:28]
like comic strips of all the um athletes
[00:38:31]
which is really cool um let’s see if
[00:38:33]
it opens yeah we showed this on a
[00:38:36]
podcast we saw
[00:38:38]
Yeah.
[00:38:40]
No, it’s fine.
[00:38:41]
So some people made some games,
[00:38:42]
which was cool.
[00:38:44]
There was this globe of all the countries,
[00:38:47]
which was cool.
[00:38:48]
This guy here made this,
[00:38:51]
which was really cool.
[00:38:53]
Someone made a Tinder app as well,
[00:38:55]
where you could swipe through the athletes
[00:38:57]
Tinder style, which was kind of funny.
[00:39:00]
We didn’t necessarily put that in, but…
[00:39:04]
Yeah,
[00:39:04]
so everything here was all built with
[00:39:06]
Claude Code as my kind of thought partner
[00:39:11]
in the background.
[00:39:12]
And I really used Claude as well as
[00:39:17]
some specific skills to be able to help
[00:39:20]
me go through the brainstorming process
[00:39:23]
and go through the like
[00:39:26]
prd process and go through the feature
[00:39:28]
process of this all the way through to
[00:39:31]
like actually pushing things to github as
[00:39:33]
well so if we like you can even
[00:39:35]
open it up on on github and see
[00:39:37]
on github like this is the actual code
[00:39:40]
base for what i just showed and so
[00:39:43]
you can see like the two contributors
[00:39:45]
claude and myself claude being the primary
[00:39:47]
contributor
[00:39:50]
But yeah,
[00:39:51]
I definitely just felt like it was a
[00:39:53]
great opportunity to use AI to just help
[00:39:55]
me move faster and to keep things
[00:39:57]
organized and make sure that things were
[00:40:01]
easy to read, for example.
[00:40:04]
And so yeah,
[00:40:05]
I don’t know if you guys have any
[00:40:09]
questions about that so far?
[00:40:10]
Yeah,
[00:40:10]
I’ll put a link to that in here
[00:40:18]
as well.
[00:40:22]
One thing I’ll just highlight is the
[00:40:23]
Airtable MCP server.
[00:40:25]
There’s an Airtable MCP server that this
[00:40:29]
guy here, Dom, I’m assuming his name is.
[00:40:32]
Dom Dom Egg.
[00:40:33]
Adam Jones.
[00:40:34]
There you go.
[00:40:35]
So he works for Anthropic.
[00:40:37]
And so he’s put out an MCP server
[00:40:40]
for Airtable, which is great.
[00:40:42]
And he keeps it pretty up to date.
[00:40:43]
Five days ago, it was released.
[00:40:46]
So you can pretty much do anything that
[00:40:47]
the API can do,
[00:40:50]
this MCP server can do,
[00:40:52]
which is really cool.
[00:40:53]
Explain MCP.
[00:40:55]
We’ve talked about it in the past,
[00:40:56]
but give your definition of what it is.
[00:40:59]
So the couple of analogies I’ve heard is
[00:41:01]
like,
[00:41:02]
it’s the USB-C for APIs and AIs in
[00:41:08]
general.
[00:41:08]
So there’s two core differences between an
[00:41:13]
MCP.
[00:41:15]
There’s core differences between MCPs and
[00:41:18]
APIs.
[00:41:19]
APIs, if we just look at the…
[00:41:23]
APIs in general,
[00:41:24]
there’s documentation for APIs,
[00:41:26]
and then there’s the actual like written
[00:41:28]
API of the code base in the backend.
[00:41:31]
And you have to,
[00:41:32]
if you wanna know what an API can
[00:41:34]
do,
[00:41:35]
You go to the documentation.
[00:41:37]
You look through the documentation.
[00:41:38]
You look through sample payloads and stuff
[00:41:40]
like that to learn about what it can
[00:41:42]
do.
[00:41:44]
And so I can just quickly open up
[00:41:46]
the API documentation for this specific
[00:41:50]
base.
[00:41:52]
But you have to know how to work
[00:41:54]
your way around API documentation.
[00:41:59]
and then if you if you if you
[00:42:00]
do you can make api calls to your
[00:42:04]
data sets or to this tool to this
[00:42:06]
application and get information in and out
[00:42:08]
of it but mcp essentially there is no
[00:42:14]
documentation the documentation is built
[00:42:16]
into the mcp and so if you want
[00:42:19]
to know what an mcp can do
[00:42:21]
the MCP has a specific set of tools
[00:42:25]
that it can do.
[00:42:25]
So you can point your AI to an
[00:42:28]
MCP and you don’t have to worry about
[00:42:30]
pointing it to documentation and matching
[00:42:32]
documentation or anything like that.
[00:42:34]
So it’s like,
[00:42:35]
it’s cool because it happens in real time
[00:42:38]
as opposed to making,
[00:42:39]
needing to make sure that the
[00:42:40]
documentation is up to date to the API.
[00:42:42]
And I’m sure like we all have experienced
[00:42:44]
it when the documentation is not actually
[00:42:47]
what the API actually returns.
[00:42:50]
and how frustrating that is.
[00:42:51]
You have to like manage and keep track
[00:42:53]
of both.
[00:42:56]
So yeah, is that a decent?
[00:43:00]
Yeah,
[00:43:01]
you’re gonna show us the project using it.
[00:43:05]
You wanna show us that?
[00:43:09]
Using the MCPs, using the table?
[00:43:12]
Yeah, so let me,
[00:43:13]
so I can open up,
[00:43:15]
let me do this.
[00:43:17]
Let me just open up a warp terminal
[00:43:19]
real quick.
[00:43:22]
Actually, I can do it here.
[00:43:22]
It’s totally fine.
[00:43:25]
Can you see my warp,
[00:43:26]
or are you just seeing my browser?
[00:43:27]
We’re seeing browser.
[00:43:30]
My fault.
[00:43:31]
One second.
[00:43:32]
Whoa, whoa, whoa, whoa.
[00:43:36]
Hello.
[00:43:36]
Hello, crazy.
[00:43:39]
Let’s share my screen.
[00:43:39]
Here we go.
[00:43:45]
Okay, can you see Warp here?
[00:43:48]
Yeah.
[00:43:49]
Okay, cool.
[00:43:50]
So as you can see,
[00:43:51]
like I actually use Cloud Code to just
[00:43:54]
kind of like identify some talking points
[00:43:57]
for this specific podcast,
[00:43:59]
which was pretty cool.
[00:44:00]
What you’re looking at is Warp,
[00:44:01]
which is a terminal that allows you to
[00:44:04]
interact with things like Cloud Code.
[00:44:07]
And so if I just type forward slash
[00:44:09]
clear,
[00:44:10]
you can see like now we’re in Cloud
[00:44:12]
Code.
[00:44:14]
The nice thing about Warp is that you
[00:44:15]
have all of your documents on the
[00:44:17]
left-hand side here.
[00:44:18]
This is a folder that I’m in right
[00:44:20]
now on my computer.
[00:44:21]
This is the actual terminal.
[00:44:24]
I can say what
[00:44:30]
In Cloud, if I do forward slash MCP,
[00:44:33]
you can see here that these are all
[00:44:35]
of the MCPs that are either installed on
[00:44:39]
this project at a user level or at
[00:44:42]
a Cloud level themselves.
[00:44:44]
You can see I have the Airtable MCP
[00:44:46]
that’s installed and connected on this
[00:44:49]
specific project.
[00:44:51]
I can escape out of this and say,
[00:44:53]
show me my tables.
[00:44:56]
And it should, everything happens,
[00:44:59]
everything should work perfectly.
[00:45:03]
But this should essentially do a call to
[00:45:06]
the Airtable MCP.
[00:45:08]
So you can see it loaded context here.
[00:45:10]
Okay,
[00:45:10]
so it’s running a list tables MCP call
[00:45:14]
on my specific table.
[00:45:16]
This is my base ID that’s saved within
[00:45:18]
the context of MCP.
[00:45:21]
this specific Olympics app.
[00:45:23]
And you can see here like, OK,
[00:45:25]
here’s fourteen tables that are within the
[00:45:27]
Winter Olympics fantasy base.
[00:45:29]
And so all I had to do was
[00:45:31]
say, show me my tables.
[00:45:32]
And it knew all the context that it
[00:45:34]
needed to know to know that what tables,
[00:45:37]
what base I was even talking about.
[00:45:39]
this specific MCP server only has a
[00:45:42]
personal access token to only this
[00:45:45]
specific base as well.
[00:45:47]
So if I were to give it a
[00:45:49]
base ID of some other base ID and
[00:45:52]
say, what tables are in this base?
[00:45:54]
It should error out completely and say,
[00:45:56]
I don’t have access to that base,
[00:45:58]
which is kind of a good general best
[00:46:00]
practice.
[00:46:00]
So from here,
[00:46:02]
you can do so many things.
[00:46:04]
I think you can say,
[00:46:06]
what can you do with the…
[00:46:09]
MCP from Airtable, but yeah.
[00:46:16]
Any questions about the Airtable MCP?
[00:46:19]
This is just King’s Quest.
[00:46:23]
Exactly.
[00:46:24]
The dopamine of like playing games and the
[00:46:27]
dopamine of using AI and like Claude is
[00:46:30]
exactly the same.
[00:46:31]
It’s just like never ending.
[00:46:32]
Like, oh my gosh, I can keep going.
[00:46:33]
I can keep going.
[00:46:34]
I can keep going.
[00:46:38]
which you can’t do.
[00:46:38]
That’s cool.
[00:46:39]
Yeah.
[00:46:43]
That’s cool.
[00:46:44]
So it’s like an interesting highlight to
[00:46:45]
point out here with this MCP is that
[00:46:48]
it can do… There’s this MCP,
[00:46:51]
Airtable themselves have an MCP server.
[00:46:54]
And then there’s like Omni,
[00:46:56]
which is Airtable’s…
[00:46:58]
AI and they all do different things.
[00:47:01]
This MCP server can do pretty much
[00:47:03]
anything that the actual public API can
[00:47:06]
do.
[00:47:06]
It can do things like create tables,
[00:47:08]
it can update tables, update fields,
[00:47:10]
all that stuff,
[00:47:11]
but it can’t create a new base because
[00:47:13]
the public API,
[00:47:14]
you can’t create a new base with the
[00:47:15]
API, you have to create that base.
[00:47:19]
This MCP is the most comfortable to me
[00:47:23]
because I’m the most comfortable with the
[00:47:25]
public API.
[00:47:28]
The Airtable MCP server,
[00:47:29]
which is the one that’s actually built
[00:47:31]
into ChatGPT.
[00:47:33]
And for example,
[00:47:35]
like you can load up ChatGPT and link
[00:47:36]
up your Airtable.
[00:47:37]
That MCP can only read from your Airtable,
[00:47:42]
I think.
[00:47:42]
It can’t actually write anything.
[00:47:44]
It can’t create tables.
[00:47:45]
So it’s kind of like a more locked
[00:47:47]
down MCP that’s in the Airtable one.
[00:47:51]
And then the beauty about Omni is that
[00:47:53]
Omni can do a whole bunch of things.
[00:47:56]
that neither MCPs can do and the public
[00:48:00]
API can’t do.
[00:48:02]
And so as I was going through this
[00:48:04]
process of building this Olympics app,
[00:48:08]
there were times where I would say,
[00:48:10]
and I can say like,
[00:48:12]
show me a plan for Omni…
[00:48:16]
that I created that AI created that works.
[00:48:18]
It’s funny how easy I’m like typing AI
[00:48:20]
and I at the same time.
[00:48:24]
But there were times where I had Claude
[00:48:28]
actually come up with an Omni prompt for
[00:48:31]
me.
[00:48:31]
because Omni can do things that these
[00:48:34]
things can’t do.
[00:48:35]
And so things like building interfaces,
[00:48:37]
building automations,
[00:48:40]
building certain fields that aren’t
[00:48:42]
possible to be built with the API was
[00:48:47]
super, super helpful.
[00:48:48]
So I kind of like used Claude for
[00:48:51]
some things, I use Omni for some things,
[00:48:53]
and then I use just like domain experience
[00:48:56]
and like domain expertise for other
[00:48:58]
things.
[00:48:58]
And there’s a lot of times where,
[00:49:01]
you know okay here you go like there’s
[00:49:02]
some interesting things like a plan that
[00:49:05]
Airtables uh where is it oh the
[00:49:08]
the landing page like so there’s a plan
[00:49:10]
here an implementation guide uh like
[00:49:12]
document for the landing page specifically
[00:49:15]
and so i like i could if i
[00:49:18]
wanted to just actually pull this entire
[00:49:21]
document and give this entire document to
[00:49:23]
omni and let omni
[00:49:26]
build this entire thing,
[00:49:27]
which was pretty cool.
[00:49:31]
So.
[00:49:31]
Yeah,
[00:49:32]
I wish Omni had a public API or
[00:49:35]
MCP server.
[00:49:36]
I would hope it’s close.
[00:49:40]
Like I would hope it’s close.
[00:49:41]
That’s all I’ll say.
[00:49:42]
Okay.
[00:49:43]
Yeah.
[00:49:43]
Fair enough.
[00:49:46]
And just a correction,
[00:49:47]
you can create a base through the public
[00:49:49]
API.
[00:49:50]
So I’m surprised that the MCP doesn’t
[00:49:53]
expose that.
[00:49:57]
I forget exactly what it is.
[00:49:59]
I know this is for enterprises,
[00:50:01]
but the enterprise API,
[00:50:03]
you can create a managed app instance,
[00:50:06]
which creates a base.
[00:50:08]
But there’s got to be some other endpoint
[00:50:11]
that just give me an empty base.
[00:50:13]
It’s weird if it can’t.
[00:50:19]
that’s where those like AI agents that are
[00:50:21]
doing like the constant research every
[00:50:22]
single night is like actually been super
[00:50:24]
helpful, you know, cause things change.
[00:50:27]
Like,
[00:50:27]
like that’s actually one of the most
[00:50:28]
helpful things that I’ve found with AI is
[00:50:31]
like, I just give it some GitHub repos.
[00:50:33]
I give it some like tools that I
[00:50:35]
use and I just have it do research
[00:50:38]
for me on a routine basis.
[00:50:40]
And this isn’t something that like you
[00:50:41]
could do this in an Airtable.
[00:50:43]
Like you could do all this like web
[00:50:45]
research actually in Airtable.
[00:50:46]
It doesn’t have to be something super
[00:50:48]
complex, but,
[00:50:49]
But with so much new stuff happening,
[00:50:52]
it’s so hard to stay on top of
[00:50:54]
things and actually deliver value and
[00:50:56]
like, yeah,
[00:50:57]
like deliver value as like a human in
[00:50:59]
this world, you know,
[00:51:02]
to keep track of all this stuff gets
[00:51:04]
exhausting for sure.
[00:51:07]
So talk about the skills repo that you
[00:51:11]
have.
[00:51:11]
Like, what does that add on top?
[00:51:14]
Yeah, for sure.
[00:51:16]
So one of the things that was extremely
[00:51:18]
frustrating was the simple fact that not
[00:51:22]
everything… Admittedly,
[00:51:25]
I wish that Airtable’s MCP and or the
[00:51:30]
documentation for what Airtable can and
[00:51:32]
cannot do was in a more central location.
[00:51:37]
There’s some things that aren’t documented
[00:51:39]
at all.
[00:51:39]
There’s some things that are new.
[00:51:40]
There’s some things that have old
[00:51:41]
documentation to it.
[00:51:43]
And so there were times where like either
[00:51:46]
Claude or Omni, I think,
[00:51:49]
was it you Ali that had posted something
[00:51:50]
of like,
[00:51:51]
I just taught Omni that it couldn’t do
[00:51:53]
something that it thought it could do,
[00:51:55]
or I don’t know.
[00:51:58]
Maybe I, Oh,
[00:52:00]
that happens all the time where it’s like,
[00:52:02]
I mean,
[00:52:02]
my clients will come to me and they’ll
[00:52:03]
be like, well, chat GPT said that,
[00:52:06]
that it can work.
[00:52:07]
And I’m like, no, it can’t.
[00:52:10]
Or like, it’ll,
[00:52:11]
it’ll kind of get them halfway there,
[00:52:13]
but not quite like, it’s just, yeah.
[00:52:16]
It’s sometimes.
[00:52:17]
I’ve seen formulas that Omni builds and
[00:52:20]
I’m not impressed yet.
[00:52:21]
Yeah, exactly.
[00:52:24]
So, so what, um,
[00:52:27]
Real quick,
[00:52:29]
as I was actually building this interface,
[00:52:32]
I was also building my own Airtable skills
[00:52:37]
for Cloud Code.
[00:52:38]
There’s a handful of skills here that you
[00:52:42]
can just go and download off my GitHub
[00:52:44]
repository.
[00:52:45]
and install in your own Cloud Code.
[00:52:47]
You can see that there’s actually other
[00:52:48]
people that are contributing to it now.
[00:52:49]
Shout out to Andrew Mitchell.
[00:52:51]
He’s been contributing to these skills as
[00:52:53]
well, which is super helpful.
[00:52:56]
Ultimately,
[00:52:57]
the goal of this repository and the goal
[00:52:58]
of these skills is to
[00:53:00]
be able to go into something like Claude.
[00:53:03]
And if you do forward slash Airtable,
[00:53:06]
these are my specific Airtable skills.
[00:53:10]
So if I just wanted to do like
[00:53:11]
a field audit, for example, on this base,
[00:53:14]
I can just do my Airtable field
[00:53:16]
audit skill.
[00:53:17]
And obviously there’s some,
[00:53:19]
opinions that i have injected into these
[00:53:22]
skills that you can take and do on
[00:53:25]
your own like you know go for it
[00:53:26]
so i’m just gonna say so this is
[00:53:27]
gonna walk me through the process of like
[00:53:30]
what do you want me to field audit
[00:53:31]
so i’m gonna say like do my entire
[00:53:33]
base right like let’s do a field audit
[00:53:35]
on my entire base and so there’s a
[00:53:37]
series of prompts that this will go
[00:53:38]
through
[00:53:40]
And each individual skill has some prompts
[00:53:42]
that it goes through and it has other
[00:53:44]
skills that it knows about and everything.
[00:53:46]
And so right now it’s going through,
[00:53:48]
it’s auditing this entire base.
[00:53:50]
And so as we can see here,
[00:53:51]
I’ll just pull open the actual base itself
[00:53:55]
just so we can keep everyone honest here.
[00:53:57]
So there’s like, you know,
[00:53:59]
what’s nice is that like,
[00:54:01]
Cloud Code and the MCP has put
[00:54:05]
descriptions for every single one of my
[00:54:07]
fields.
[00:54:08]
Every single one of my tables has really
[00:54:11]
nice, neat, built-in documentation.
[00:54:15]
And so if anything,
[00:54:16]
as you’re building Airtables with or
[00:54:20]
without AI help,
[00:54:22]
um just linking up the mcp server and
[00:54:25]
hopping into cloud code and saying like i
[00:54:27]
need help documenting like my entire air
[00:54:30]
table uh ask me a bunch of questions
[00:54:33]
and i’ll give you answers and then like
[00:54:34]
go for it you know uh it just
[00:54:37]
helps so much for those things that just
[00:54:39]
would normally take a lot of time or
[00:54:40]
quite honestly like i would just skip i
[00:54:42]
just wouldn’t do it you know yeah um
[00:54:47]
Yeah,
[00:54:47]
so skills are kind of like the next
[00:54:49]
evolution of,
[00:54:51]
and Claude’s the one that kind of defined
[00:54:53]
it.
[00:54:53]
Others are starting to adopt it.
[00:54:55]
So it’s kind of becoming the standard now.
[00:54:59]
It’s basically a way to like,
[00:55:02]
the way I view it is in theory,
[00:55:05]
you could create MCP endpoints that do
[00:55:07]
what skills do,
[00:55:09]
but skills run locally and don’t use the
[00:55:13]
credits.
[00:55:17]
as much.
[00:55:17]
Yeah.
[00:55:17]
Yep.
[00:55:18]
Yep.
[00:55:19]
And I will,
[00:55:19]
I will just plug as well.
[00:55:20]
Like Claude Academy is what it’s called.
[00:55:24]
I know that like everyone has like their
[00:55:26]
opinions about Airtable Academy and all
[00:55:28]
that stuff.
[00:55:28]
I think the Airtable Academy is
[00:55:29]
incredible,
[00:55:30]
but the Anthropic Academy just came out
[00:55:34]
like a couple of weeks ago and it’s
[00:55:36]
like a wildly,
[00:55:38]
wildly in depth actual Academy that you
[00:55:42]
can go through to learn about Claude and
[00:55:44]
from like end to end.
[00:55:47]
And so if you’re trying to learn about
[00:55:49]
Claude and you’re trying to learn about AI
[00:55:50]
and what things can happen,
[00:55:51]
and here’s the thing about agent skills,
[00:55:53]
that’s what got me thinking about this.
[00:55:54]
Like if you want to learn how to
[00:55:55]
build agent skills,
[00:55:57]
I’m not going to be able to teach
[00:55:58]
you better than this.
[00:55:59]
Cool.
[00:56:02]
So.
[00:56:03]
Yeah.
[00:56:04]
Yeah.
[00:56:05]
So that’s kind of the next evolution of
[00:56:07]
that to the point where, you know,
[00:56:10]
I was,
[00:56:12]
as I’ve been kind of processing this,
[00:56:14]
you know, skills,
[00:56:16]
it used to be you like we create
[00:56:18]
like these mini agents that had their
[00:56:20]
context,
[00:56:21]
and you’d have like lots of different,
[00:56:23]
you know,
[00:56:24]
agents that did different things.
[00:56:26]
Now, you don’t create multiple agents,
[00:56:28]
you just kind of have your one like
[00:56:30]
master agent that is exposed to different
[00:56:34]
skills, and the skills,
[00:56:36]
take it essentially like,
[00:56:38]
tell the agent what to do for that
[00:56:40]
specific use case.
[00:56:42]
exactly yeah so you can see here like
[00:56:44]
this just did a little field audit for
[00:56:46]
me it like changed some fields and that’s
[00:56:49]
cool and i could say like add my
[00:56:52]
emoji prefixes to like everything as well
[00:56:56]
and i can like so one of the
[00:56:58]
things like i use this a lot for
[00:57:00]
is to like this specific skill is for
[00:57:02]
creating formula fields or like lookup
[00:57:05]
fields and roll up fields i’ll have it
[00:57:06]
actually create the field
[00:57:08]
and then put a little emoji in the
[00:57:11]
beginning of the fields.
[00:57:12]
I’ll go into manage all my fields.
[00:57:15]
I’ll like search for that emoji.
[00:57:17]
So it only will show me fields with
[00:57:19]
that little emoji in it.
[00:57:20]
And then I can go in and I
[00:57:21]
can actually change a long text field or
[00:57:24]
a single line text field,
[00:57:25]
which is just like the default that it
[00:57:26]
gets created to.
[00:57:27]
And it’ll say like changes right here,
[00:57:30]
like formula changes to round.
[00:57:32]
So I’ll literally just go in,
[00:57:33]
copy this like that, come in here.
[00:57:36]
paste this here, as you can see,
[00:57:37]
I’ve done, and then like just hit enter.
[00:57:40]
And so if I,
[00:57:42]
and you can see like things are getting
[00:57:43]
created here right now,
[00:57:46]
cause doing a whole bunch of stuff right
[00:57:47]
here.
[00:57:48]
So, but yeah,
[00:57:49]
that’s just like some of the things that
[00:57:51]
the Airtable skills can do.
[00:57:52]
It also will tell you if it can
[00:57:54]
and cannot do some certain things,
[00:57:56]
which is super helpful as opposed to like
[00:57:58]
it get into these infinite loops of, oh,
[00:58:00]
I can do this.
[00:58:01]
It’s like,
[00:58:01]
it has a little bit of documentation in
[00:58:03]
terms of what it can and cannot actually
[00:58:04]
do.
[00:58:06]
Metadata has been the theme of the season.
[00:58:09]
I have been loving it.
[00:58:14]
Yeah.
[00:58:14]
Awesome.
[00:58:14]
Yeah, no, it’s inspired me.
[00:58:15]
I’ve been building stuff on that point as
[00:58:17]
well.
[00:58:18]
So very useful.
[00:58:20]
Awesome.
[00:58:20]
Thank you, Rob, for coming on.
[00:58:23]
Exciting stuff.
[00:58:24]
And we’ll be hearing from Rob in the
[00:58:26]
future, I’m sure,
[00:58:27]
keeping us up to date on all the
[00:58:29]
latest and greatest.
[00:58:31]
So appreciate it.
[00:58:34]
and that concludes this season uh we’ll be
[00:58:36]
back next month we might have some special
[00:58:39]
um episodes during the next month
[00:58:41]
especially with the with the community
[00:58:43]
hackathon want to promote that so
[00:58:46]
everybody check it out the links in the
[00:58:47]
notes and um submit your play i don’t
[00:58:51]
know rob you’re gonna submit your your
[00:58:53]
olympics you can put that in there let’s
[00:58:56]
let’s play full enough i think yeah we’ll
[00:58:58]
see maybe there’s i don’t know maybe
[00:59:00]
something else will percolate up uh
[00:59:04]
Very cool.
[00:59:05]
Great.
[00:59:05]
Everybody have a great week and month and
[00:59:07]
we will be back next time.
[00:59:09]
Take care.
[00:59:09]
Sounds good.
[00:59:10]
See you everyone.