Near Future Podcast

Tom and Jonny sit down with Randy Hunt, Head of Design at Notion and chair of MFA Design at SVA. They dig into why speed and quality were never a trade-off, unpack the weekend build that turned into a two-week code review, and go behind the scenes of how a graduate design programme is rewriting itself around AI.
Chapters
  • 0:04 — Two institutions, one calendar: how Randy keeps both plates spinning
  • 3:00 — Agents that absorb the busy work so he can stay close to the craft
  • 5:29 — Dogfooding vs. knowing the competition: you have to do the work twice
  • 8:22 — One place for everything, or a walled garden? Jonny re-buys Notion
  • 11:10 — Battle-tested vs. multi-model: why living in one suite stops working
  • 13:22 — Good, fast, cheap: Randy refuses to pick two
  • 15:17 — Quality isn't the screen being right, it's how many people hit it wrong
  • 19:05 — The handoff is a telephone game: why passing the baton kills the work
  • 22:52 — Designers were shipping production code before the coding agents arrived
  • 25:42 — He built it in a weekend. The PR took two weeks to review
  • 29:35 — Writing code got cheap. Writing good code didn't
  • 30:46 — Emergent chat vs. written-down skills: making the system legible to everyone
  • 33:07 — Working in the open so you can see how your colleagues actually prompt
  • 35:42 — Build it yourself or wait for someone to sell it: micro tool making
  • 39:04 — The barbell curriculum: what endures vs. what has to stay malleable
  • 43:20 — Directing a photographer vs. directing a model: same skill, different studio
  • 44:57 — Where's the punk rock student who refuses to use AI?
  • 47:58 — He went home for the summer and built an ERP for his dad's warehouse
  • 49:33 — Quick fire: AGI takes over, so what do you actually do with the time?
  • 51:37 — Wrap up, and a promise to come back for the music conversation
Mentioned in the episode
Randy
Companies and past roles
Tools and products discussed
Events

Creators and Guests

Host
Jonny Burch
Co-founder of Near Future
Host
Tom Harman
Co-founder of Near Future

What is Near Future Podcast?

A regular podcast covering design and AI from the founders of Near Future, a boutique AI consultancy focused on teams that care about craft. We cover both what we're seeing on the ground and industry trends, ways of working and occasional guests from the design world.

Tom: Welcome to the Near

Randy: Welcome.

Tom: where Jonny and I chat with
design leaders at the edge of

what's possible today with ai.

Today we have Randy Hunt, veteran
design leader of New York.

you've spent time at
Etsy Artsy in New York.

You also spent some time in Singapore
working at Grab as VP of Design there.

Um, and most recently you've been
at Leading Design at Notion, and you

also balance this alongside work at
the S at SVA where you are currently

the chair of the MFA in design.

amazing things.

Welcome to the show.

It's great to have you here.

Randy: Great to be here.

Thanks for having me.

Yeah.

Tom: Um, so just to, to kick things
off, uh, a little bit, and maybe this

is a weird time to do this, given
the summer, and, um, this, maybe

that's not as true in New York, but
definitely in Europe it is true.

But curious about your day to day.

You've got a lot of things going on,
a lot of plates, you're spinning.

Um, could you tell us a little bit
about what your day to day looks like?

Randy: Yeah, sure.

I mean, it's, it's interesting.

It's, um.

I do have a lot of things going on.

I would say I manage at a high level.

I like manage what's going on
with, uh, a support system.

You know, there's kinda like the
human support system and there's like

the systems planning support system.

You know, I think my, uh, my wife
and partner and many things is

a really helpful like sort of
collaborator and planner around stuff.

I think we do a lot to orient our,
our lives to op, allow each to have

the sort of optimal times we need
when we kind of are doing the need

to work the way that we need to work.

You know?

Um, that's extremely, extremely helpful.

And also to kinda like come together
and, and support and support each other.

Um, she's involved a lot in
supporting the students at SVA and

their thesis work, which is cool.

Um, so it's fun to get to
work together, um, on that.

The other is, um, you know, there's
support staff in places like

Notion and places like SVA that
help things really run and move.

And so, uh, I think being, you
know, having developed good habits

and skills around delegation and
um, how to kinda like operate with

a team, uh, is certainly helpful.

And then the other part is maybe
gets closer at some of this topic

of like working with ai, which is
really just kind of scaled personal

systems and also finding ways to, um,
use those systems to allow me to get

or remain like close to work itself.

You know, that really kind of like in
the weeds, not just be kind of, you

know, delegating and orchestrating,
but also be doing like the craft

of whatever it is I'm working on.

Um,

Tom: Do you have

Randy: and to be able to.

Tom: any, do you have
any examples of that?

I would love to hear
that looks like for you.

Randy: Yeah, I mean, there's
interesting ones around kind of

just operational things around
communication and stuff like that.

You know, I've got email agents and task
agents and things that, uh, in a way kind

of, yeah, they've mirror some of my, like,
behaviours and decision making in the

past, but I've kind of documented them.

And so the way that things get sort of
triaged, categorised, prioritised, and

can like, acted upon, um, in a way follow
a personal process I've probably been

doing for like, I don't know, 20 years
at this point or something like that.

Um, and that's really helpful because
I, the busy work of kind of managing

all of that is dramatically reduced.

That's just like time saved.

Well, I still have a good degree of
confidence that the things that I wish to

treat as important are in front of me when
I'm choosing to spend time doing whatever

that sort of class or category of work is.

Um, the other is that in some parts
of work, really like establishing, um.

What are effective with that custom
agents to do more specialised

workflows, um, around things that
would be more labour intensive.

You think about kind of collecting,
collecting resources for communications,

whether that's, you know, in the kind
of industry professional context, you

might look at things like updates on
progress that's been made either for me

to receive and understand or to like,
share in the education context actually

looks a lot like, um, kind of collecting
events in a way you might imagine

like, here's accomplishments of alumni,
here's accomplishments of students,

here's kind of things of faculty.

You wanna like compile these things
together and communicate them back out.

This is kinda the ecosystem of, um,
kind of credibility and accomplishments

and like the long tail of the effect,
the, you know, the positive effect of,

you know, uh, an education programme.

It's important for us to kind of un
understand those as much as possible,

organise them and share them back.

So there's a set of like workflows there.

There's a kind of what I would used
to be like manual knowledge work and

it's like essentially automated now.

And I automate those things almost ent
entire, uh, almost entirely within Notion.

I, I, I mean I do most of my like,
agent orchestration inside Notion

and stuff, which is one, you know,
because that's something we built

and it's something I'm very close to.

That's my preferred way
to kinda work on things.

But there are other ways to
accomplish similar, yeah.

Jonny: Randy, I was gonna
ask, did you go full Notion?

I, but in in past roles, I've, I, I cut
my teeth at Deliveroo and I was, and there

was a leaderboard of who'd spent the most
money internally, uh, on Deliveroo, which

is Europe's answer to Uber Eats for, for

Randy: Yep.

Jonny: other Americans.

Um, and uh, I was really
low down the chart.

I barely used the service, uh, proudly.

So I dunno why, just, you
know, careful with money.

I dunno.

Um, it's probably more a question for
a therapist than anything else, but,

um, I'm just interested like how you,
are you, are you running all of this

operational load through and like
as you've been building agents, have

you been, you know, full dog food.

Randy: Yeah, yeah.

Uh, absolutely, absolutely.

So the, I love dog fooding in general,
and, and I think that it's such a powerful

way to help understand and improve a,
a piece of software and experience.

Um, and yeah, that was true of my
time at Etsy, so we didn't have like a

leaderboard, but there's moments where I'd
go and, you know, do some research myself,

just kind of understand and for sure I
was pretty like a top-tier Etsy shopper.

I mean, I mean at one point we bought this
old house from like the 1880s and, uh, we

were kind of like, did a gut renovation.

I mean every like lighting fixture
in there and, you know, everything

was from Etsy, like all of the major
kind of like pieces of furniture and

lots of small decor pieces stuff.

So I, I use that as an excuse
to kind of dogfood the product.

Um, and so feel similarly, so
yeah, the, here's maybe the hard

part about something like this.

Uh, so like full in on Notion, both
in like my Notion life, I was full

in Also I got the, uh, graduate
design department at SVA kind of

fully running in Notion as well.

And that's like a, you know, we're
a paid customer there out of none of

the, you know, none of the development
features, none of the tests, none

of the next generation stuff.

I've used it very much like a
regular paying customer there.

And I like having both those
experiences to kind of appreciate that.

But the other thing that's happening,
we're working on sort of technology

is I'm trying to, you also wanna
understand the other methods.

You wanna like be a real user of like
competitive products and all this stuff.

You spend a lot of time
using tools and trying stuff

multiple ways to understand it.

You're like, okay, it's pretty
laborious when I describe it that way.

Then you're like, all that other
operational efficiency stuff I

described is, has got to happen.

Otherwise it'd be like impossible
to spend time, you know, duplicating

an activity only using Cowork

Jonny: Right.

Randy: just to feel what that experience
is like and then go do it again in

Notion because you want it to do it
the way you would prefer to do it.

Yeah.

Yeah.

Jonny: went through the process
recently of, of re-buying Notion.

Tom and I, we, we had to sort of
make a decision about six months

ago, okay, where do we put our stuff?

We were sort of messy Google Docs
and I think we'd finally decided

to like buy a whole new G Suite,
know, organisation for ourselves.

You know, really splurging on that.

But do you really want to use Google
Docs for all of your collaboration?

No, not really.

So then therefore, you know,
going back through the process

again of being like, Notion is
actually great a, a load of stuff.

And um, so it was nice actually because
I, I sort of feel like I, I was a

such an early user of Notion that
I have it free forever, seemingly,

I dunno, maybe they'll change it.

Um, but uh, and loved it and
then was like, eh, I don't know.

now sort of coming back,
I think it's really

Notion is speaking to me quite a
lot in that I sort of just want one

thing and I like that I can plug in
agents and like extend it as I need,

but also it can be really simple.

Um, so anyway, it's just,
uh, yeah, it's nice.

It's, it's actually really nice to
have you on on here 'cause we are

using Notion so, so heavily now
again, and we've built our own CRM

inside of Notion and stuff like that.

But

Randy: Yeah, great.

I mean, that's the, yeah.

Music to my ears.

No, it's a real pleasure to hear that.

I think that, right, that kind of like
say scalability, I feel so buzzwordy,

but the flexibility of the software to
kind of serve very simple needs, but also

quite advanced needs, uh, in one place.

And while you kind of flex between
those, and it's certainly one of

its points of power also, if you
were think you kind of connect your

external agents as we've done that.

I think there's a, we don't use
this kind of like open source

or even like walled garden kind
of terminology, much externally.

But there's definitely a spirit
of like Notion is, uh, it's

like not a walled garden.

You can get your content in and out,
you can connect other things to it.

It's kind of like AI knowledge
work's Switzerland is a little

bit of a idea, you know?

Mm-hmm.

Um, and I think that pairs well with what
has historically been a strong emphasis

in the, in the product, even pre AI to
be sort of malleable and flexible to a

user's needs, um, and pot, sort of the
possibilities of what you might wanna do.

Um, which has also, I mean, it's a
great advantage and also has some

challenges in user experience and
onboarding product communication.

You know, it's kind of
a natural tension there.

But overall, this idea that you can
be extremely flexible and do lots

of different things and then that,
uh, you can bring things to it and

bring things out of it, I think
is actually a really, um, a really

durable, like, competitive position.

And also just makes for like sort of
a nice product to use at this point.

I

mean you can I mean to your

point, you can, be in the, in the
Google Suite and be like quite happy

with the number of tools there.

I mean, a lot of the stuff's
very battle tested, right?

I mean, Sheets is like,
you know, it's just

Jonny: totally.

Randy: battle tested at this point.

Yeah.

Uh, uh, and Gemini's making some
great, you know, advancements and

points, and you have moments where you
feel like that even the integration

of the product is like, pretty good.

Uh, it'll surely keep getting better,
but then you're like, oh, but I

kind of wanna use like, Claude for
this thing or something like that.

You are like, oh, if I wanna live in
like a multi-model world and not just

be in like Gemini land, I kind of need
another place to get my work done.

Jonny: Right,

Randy: That's true for all the labs, you
know, like, you like, unless you wanna

do everything with ChatGPT, you kind of
need to have something that's also not in,

Jonny: Yeah,

Randy: you know, and that e.

Jonny: a few months ago, we, with a
client, we were implementing Notion

agents and, uh, it was as it was
launching and the pricing was a bit

unclear and like, you know, that I
think, no, you guys haven't quite built

some of the controls for the, the,
organisation at large to, you know,

control spend and stuff like that.

So it was a bit of

Randy: Yeah.

Jonny: it um, I I maybe I had assumed
that it would be a walled garden and we

wouldn't be able to plug in other stuff.

So it was really just like Claude
at the MCP and, oh wait, actually we

can just build quite a lot of stuff
outside Notion and use Notion workers

for the, for the stuff that is it's
best for, and have that balance.

I think it's just such a reasonable, um,
attitude to take in this, in this world.

Like

Randy: Yeah.

Appreciate that.

Jonny: out of Notion if you don't
let them use what they want.

So it's self-defeating.

It's like short term greedy rather
than long term greedy, you know?

Randy: Yeah.

Yeah.

It's like you should use it because
you want to use it, not 'cause

you feel like trapped using it,

you

know?

And, yeah.

Yeah.

Yeah.

Jonny: Um, Randy, I'm, I'm like, I'm
being greedy with all my questions, but

I'm just gonna ask, I'm gonna, go into
sort of, uh, maybe outta inside baseball

Notion specific, but I know that you have
opinions around, uh, speed and craft.

Uh, and

Randy: Yeah.

Jonny: known as a, as a product that,
you know, uh, displays incredible craft.

I think that's sort of how Notion
broke through in my mind anyways.

Like, it just looked and felt
better than anything else.

Um, and so, and actually we are gonna be
sharing a stage in September, uh, Hatch.

Uh, and hopefully you're
gonna be talking about it.

I'm gonna watch, I'm gonna watch, uh,
your talk, um, September in Berlin,

but ahead of that, if you can tease
anything about your opinions there.

I just like, how do you, um, avoid
the, you can have two of good, fast and

cheap, uh, and, and achieve like speed
while also maintaining a quality bar.

Like what, what have
you learned, I suppose.

Randy: Yeah, sure.

Well, I kind, maybe some of
this comes from, I mean, in

mindset or like in spirit.

I, I.

I really dislike any of these setups that
say like, you can't somehow have both.

So it's like, so put two things in,
some like binary with each other

and assume that they're a trade off.

I mean, I do like that, you know, I
think balancing three things is already

closer to reality, but there's probably
also other factors, you know, like these

are simplified models to help us kind
of make sense of some, some things.

But the problem is that they, they're
such good like sense making mechanisms,

this idea of like, you know, quality,
speed and cost or something that they

become, um, they, I think, limit our
thinking too, if you like, start from

there and be like, actually this whole
endeavour is balancing these three things.

And you're like, but what?

But like, what if you don't have to

Jonny: not

Randy: this right now?

Yeah,

Jonny: if the first

Randy: yeah

Jonny: is not true?

Yeah.

Randy: yeah, yeah.

What do we have right now?

Um, and so I, I feel the same about
this kind of the speed, speed,

quality, tension, uh, and, and
here's, so maybe that's the spirit is

like, oh, I want to find a way where
that's not true as one part of it.

But the other is, I think the,

what I've experienced over time is that.

One of the key factors between
something being good quality and not

good quality or being, or like falling
short of where you believe it could

be is, uh, the ability to take inputs
about what it, what is kind of falling

short and do things to close that gap.

Right.

And, uh, there's others which is like
the belief that that is a good thing

to do and like the desirability to do
it, you know, but, but this particular

part I think could best be summarised
as really like a learning cycle or

like a learning loop, which if I use
a bunch of like industry jargony stuff

over the last, what, three decades or
something, it's all approaches like agile

or like cycles or you know, whatever.

But it's basically like if I could
do some little loop of like learning

about a thing, you know, like sort of
realising something, looking at it,

have some insight, could be as minor
as like these things should be aligned,

you know, how I fix it and deploy a
change could be zoomed out a bit more.

This behaviour should be this way,
this language, this concept and

this flow, whatever it might be,
uh, is confusing for this reason.

I mean, the most obvious ones are just
when it's like broken, you're like, okay.

Uh, and lots of tiny, small, broken
or confusing things add up to making

something feel pretty poor quality.

Jonny: Hmm.

Randy: Uh, it's like if I have the
opportunity to do those things and

then fix it there, I'm closing that
gap between the thing not being as

good as it could be and being as
good as we would like for it to be.

And so then the extent to which you
can recognise those things quickly

Jonny: Hmm.

Randy: fix them quickly.

The earlier, the quality gets better,
which means that, like for, for

your users, assume, let's assume
you have like a flat or growing

user base or usage of your product.

The sooner you fix those things, the
better the quality is experienced

by users, like in aggregate, right?

They have more time of experience and
better, and likely you have more users

experiencing it better because you
have more people using it more often.

And so it's not about like the,
the screen being right or like

that little product being right,
like it was bad and now it's good.

It's actually about the number
of times an end user experience

is the thing being bad or good.

And like that, I find that
justification far more compelling.

Jonny: Yeah.

Randy: You're like, okay, there's
like 10,000 people using it this week.

There's gonna be like 11,000
people using it next week.

12,000 people using it
the week after that.

The sooner I make it better,
the aggregate effect of all

of those quality improvements,

Jonny: Right.

Randy: you know, adds up for all the,
every touch touchpoint from here forward.

And so the speed of improvement is like,
has a positive correlation with the

quality experience of the, of the product.

Uh, so it's not about like
rushing to finish something.

It's like a, I think like
a bad old idea somehow.

It's like the factory industrial model
of making the thing or something.

You're like, if I can just shove a
project through the pipeline fast

and be done, then like speed, yes.

Quality.

No,

Jonny: right.

Randy: but actually it's like, it's a
bunch of tiny like micro decisions and

learnings and fixes and you're like,
the faster I can learn those things

and then make good quality decisions
and have the results of those decisions

make their way into customer's hands,
the better quality the product is.

And sooner is, sooner is better.

I can't imagine a world where you'd
be like, you know what, that's a

good set of quality improvements.

Let's wait.

You know?

Like,

Jonny: right.

Tom: I think that, um, that definitely
makes a tonne of sense in the but

it raises some interesting questions
at the, in the tension there of,

historically in product development,
there has been maybe a designer using

their cycles to craft the experience
and then handing off to an engineer

who's then shipping to production.

Um, I think baked into what you were
just describing, there either some

handoffs or, or designers pushing
all the way through to production

or engineers starting with design.

you have any thoughts on that, how
that relationship changes in this world

where we move quickly at high quality?

Randy: Yeah.

Yes.

Uh, well, one, I think that's been true.

Pre AI is like a non AI answer.

I think that the mental
model of like handoffs.

In, I mean, certainly in software
development, this is probably true in

lots of kinds of projects where there
is truly, uh, like a passing of the

baton and no further inter, like no
pre interaction, no further interaction

is just a terrible way to work.

Awful.

Like it just, it, it rarely does it
work well, it's like playing a telephone

game, but with like far more, far more
variables, you know, like I feel bad

for the end user or the customer at
the other end of this thing because

like, you know, they're just the,
they're like the victim of, or like

decision making, poor operations, bad
leadership, you know, like, sucks on you.

Give, give your money, you know,
like, I don't, it's just awful.

Like that's bad, bad, bad.

Thankfully, I think I've been, I mean
even probably my earliest kinda like

graphic designy print production days
or something, it was just like in

environments where that was not so much
the way, you know, you would go and you'd

go and talk to your printer at the press
about the work you were about to do.

You know what I mean?

They were like brought in early
on, you know, they kind of,

they'd give you some guidance.

They'd be like, if you make this
decision, like they'd watch costs

on that one, you know, all these
interesting things would emerge.

You'd be like, cool.

They're like a partner
for you along the way.

Um, you know, you get some feedback
on some stuff, you know, they're

giving you technical, you know, if
you cut and score it this way, it's

gonna cost you a bunch of money.

But if you do it that way,
it's gonna do this, this way.

I can save time.

I've already got these
materials in the warehouse.

There's all this interesting stuff.

Right?

Uh, and then you're there at the end,
like on the day it's on, on press, you're

there like with your own eyes to like see
it roll off and be like, is that right?

You know?

Uh, I think that same mental model
applies in software, like, for sure.

And so I think you want to,
the ideal stage, to your

point, Tom, is that like yeah.

Engineering's there from
the beginning and Yes.

Design's there at the end.

Yep.

And I think the, you know, the teams I've
like built and led or been lucky enough

to be a part of where that's been true,
has been where we've just achieved higher

quality more consistently over time.

For sure.

And it's just more pleasant
to do the work, honestly.

It's like, everything
about this is better.

I, like, it's hard for me to think of
a, a down downside to this, except that

maybe you can't, then it's more human.

And so you build systems around and stuff.

You, uh, you can't treat stages of the
processes interchangeable as easily.

If you start to think this way because
it's more organic, like collaboration

matters, meaning you can't be like,
uh, we just ramp up quickly, just like,

here's some outsourced engineering.

Like if you approach your whole
thing as baton passing, then you can

interchange some parts, which maybe
at a high optimisation level can work.

But the mistakes of when you let like
high optimisation level kind of like, I

don't know, executive resourcing things
impact the actual delivered product.

Part of the job.

I think it's any kinda leadership
role is to make both of these things

be true without having this one suck.

You know, like, you gotta
make this one great.

Um, the other is that like, yes, the,
like here's maybe what's changing

some, I mean, it's definitely changing
is like, oh, the designer can do the

engineering work more easily now.

Um, also the engineer can maybe do the
design work more easily now somebody else

can do the design and engineering work.

Like that's maybe the interesting,
an interesting thing to address.

But I think that, like, I think
about the Notion design team.

I think about the Etsy design team.

Like we were shipping, like
designers, were shipping production

code changes, pre AI tools.

This is not a function of do
I have a coding agent or not.

There's a group of people who are
software designers and so they're

like hired in that profile managing
that profile expectation, that

pro the skillset in that profile.

And so this profile of designer, like
loves building software and I think

if you love, if you love building
software, other people also make software

that don't have the same profile.

And so of it's great.

Like it turns out fine.

So I don't think there's only one
way, but our way is like, well if

I love this thing, I for sure am
interested in the material it's made of.

And so I understand how software
is built and I try to get close to

that and I enjoy being close to it.

And that's not some trade off between
like the design and engineering, but

that's just like how software's made.

And you're like, of course.

Like this is all one, it's all one thing.

Tom: It makes me think of your, the book
that you wrote back in the day many years

ago, Desi, I forget the exact title of
the book now, but, um, built designing for

the web, designing products for the web,

Randy: It's called Product Design
for the Web, which is a funny,

yeah, it's like a funny name.

Even that feels like it falls in and out
of relevance, like actually right now.

That name kind of made me feel a little
relevant again, but there's a moment

where like for the web feels like a
funny thing to say when like everything

we're making is a native app, you know?

Tom: Yeah, I feel like that that
really has come back around.

I feel like there's a big resurgence
of web design as a, an area that does

Randy: Yeah.

Yeah.

Tom: product and web world very neatly.

Um, but hearing you talk about this
sounds very reminiscent of, of the

ideas you've held for a long time.

I'm curious with AI specifically coming in
now or the sort of the current era of AI

for teams, is there anything specifically
that you think is, uh, of risk?

Like, there's obviously sort of advantages
to efficiency and speed and, and you have

to move more quickly and be more fluent
in the languages of your teammates.

But uh, are there any downsides to that?

Anything you've seen with
the Notion team perhaps?

Randy: Yeah, I mean, I think
there's always some kind of

pendulum swingy reaction risk.

I think we're seeing actually, you see a
bunch of them in kind of, I'll describe

maybe a concrete example or two, but see
'em in smaller operational ways or seeing

isolated ways and then we correct them.

You know, I don't, maybe after this we can
pause and back up and be like, oh, they're

any, like macro risks or something.

But I think here's an
example of something.

I just showed one of our designers,
this had to have been nine months

ago at this point or something.

So it's funny, like these, you know,
where these chapters are moving quickly.

But, uh, the particular designer is
actually working with an engineering

team that I would say was working in a
little more of like a linear fashion, a

little more d their, their muscle memory.

The way they'd worked historically
was a little more transactional.

Part of this has to do with like
geographic separation between some

of the collaborators, like with big
time zone differences that sometimes

almost necessitates a little more
of like a handoff kind of behaviour

unless you have other operating
models to help get around that.

Um, long story short, they were used
to working in a way where a designer

would to kinda like deliver screens
and then they'd be kind of wired up.

And now of course there's a version
of like, here's Figma file screens.

There's a version of like, here's,
you know, here's a bunch of like

front end, you know, um, we get
closer to that, a bunch of, you

know, front-end dev versions.

But basically the designer, uh,
was pretty unsatisfied with the

quality of the final execution.

And also the time the team set it
would take, you know, like, oh, this

will probably take us like four weeks.

And designer was like, that
can't possibly be, you know.

Um, and so he just worked all weekend.

I mean, he was kind of like,
you know, it was early days-ish,

using like Claude Code.

Uh, he just like built it all at himself.

I mean, all the micro interactions,
like all the, like, he just

spent a long weekend doing it,
not because anyone asked him to.

He just like wanted to, he
was like, I can't believe this

would take that many weeks.

I said, there's no way.

Um, and it was awesome.

You're like, oh man, this is incredible.

Both the love and passion for making
the thing, the quality of the details.

You're like, this is rich.

And it was, look at the
amount of time it took.

Um, but basically it was
almost like a funny story.

If you would, you could kind of
maybe predict what this is going.

Uh, it was like one giant PR.

And the engineers were like, um,
so this is now gonna take us like

two weeks to review, you know?

And he is like, wait, what?

Like, that can't possibly be, it's
like an interesting, it's like running

into each other and you're like, oh,
what are some of the principles of

like, designing complex software at
scale with like hundreds or thousands

of people, like touching the code and
you're like, oh, there's like a best

practice here of kind of breaking things
down into small chain change sets and,

you know, the kind of way in which
you might go through a code review.

And so we're like, okay, we need
to put some both cultural learnings

in place, but maybe also some
checks in place and the tooling to

not have things like this happen.

Uh, and you can imagine that lots
of people could make that same.

I think the, you know, it becomes quite
easy to be prolific with the making, but

then you still gotta like review and merge
and like all these things around like,

like security, performance, et cetera,
like, you know, are real considerations.

But I think the way to address these
things today isn't to go like, oh,

and then we need to like, make sure
the human can fix this actually, like

the design scaled systems that most
of the time are making the kinds of

decisions a human would make so that
you can like get these things, you

know, kind of reviewed, approved,
merged in, and it can be, you know,

autonomous or quite nearly autonomous.

Yeah.

Yeah.

Jonny: It's yeah, interesting that, that
learning that, um, it turns out, if you

don't have to worry about the quality of
your code, writing code is easy and cheap.

Um, it's when the code has to be good, uh,
that, that's when it gets more expensive.

Um,

Randy: That's right.

And then you have to use all your
meta thinking skills again to be like,

okay, what are the decisions I made?

You know?

Then you get,

Jonny: completely.

Randy: tools and skills, you share
those skills with your teammate.

You know what I mean?

So everybody understands
how the system work.

You know, you're just constantly
in like systems design mode and

Jonny: Right.

Randy: kinda like meta design thing,

you know

to keep the machine evolving.

Yeah.

Yeah.

Jonny: I'm interested, Randy, if, um,
you've mentioned the, the building a

system, uh, presumably a Notion but
maybe elsewhere, that that encourages

speed and encourages these tight loops
and this, this learning loop to happen.

speaking, what is involved in that Is
that skills, is that, you know, ways

of working or, or training or you know,
what goes into building that system of

humans and complex, unpredictable people
that, that can move at speed predictably.

Randy: Yeah.

Yeah, there's a, I think
there's a really, okay.

It's a really organic part that
looks pretty like chatty, like with

humans, which you could imagine just
like slack channels and, or like

comments and Notion docs and things.

Right.

I'll come back to like why that is,
but I think there's always some kind

of like emergent stuff bubbling up.

It's like that's how you first find out
that this person on the team wrote this

skill or like you're both talking about
something like that should be a skill.

This kind of stuff kind of like
comes in moments like that.

Then there's like a habit and I think it
was sort of like learned across the team.

We don't really train on it, I
think you sort of like learn through

osmosis and exposure of like, maybe
we should train about it more.

I don't know, but you, where you.

Uh, how would I say?

You kind of make shared resources of
common decisions for other people to make.

You know, the simplest version of this
is like, oh, I've got some guidelines.

I mean, that predates AI or, so, you
know, like there's just some rules

or some ways of working or some
common, you know, how we do stuff.

But you start to take the how we do
stuff to be like instructions for

agents or for other systems right now in
Notion, well, this is a Notion document.

Jonny: Yeah.

Randy: right?

Like, uh, and they're like, oh,
I've got some skill basically.

And like we put 'em in Notion, um, you
know, fun, fun things to come to help

make that like, even, even, you know, e
easier and have kind of the right, um,

kinda like structuring taxonomy around it.

Um, but basically for us, like
everything's in a Notion doc.

Um, same thing as saying like everything's
in a markdown file, except that in

Notion it's like, you know, uh, now that
means it's kind of, it's it's version

controlled, it's collaborative, it's
part of the same like permissions and

access control, your system of record.

So that's convenient, um,
uh, for working with teams.

Uh, so you're like, okay,
I'm doing something.

I need to like add it
to the shared system.

Add to the shared system.

I kind of make like people know
about it or learn about it because

there's this chattery stuff going on.

And then also there's the results of
the output of that cycling back around.

So then part of that kinda like quality
learning moment thing too is this like

stream of continuous sharing of like what
we and or the agents are doing with us,

you

know and the, and the
feedback we're giving to it.

So like you might imagine that, you
know, comments on a Notion document

or like orchestrating across like
several agents or even like a Slack

thread, uh, we tend to do all of this.

This is like cultural part of it, um, as
much as possible, like out in the open.

These are like public
channels, public docs.

So you can, I can see how
you interact with an agent.

So like when you mention Claude in
a comment of a Notion document and

ask it to maybe evaluate what's there
and like spin up a process to start

like executing answer to that, I can
see like how you did it, you know?

And so I kind of learned just by it being
there part of the like public record.

But of course I can also ask and be like,
oh, what are all the ways that, like

Brian, you know, what are some of the
commands Brian uses most on the thing?

They're like, oh, cool, cool
learning, you know, uh, they're

like, oh, it's these three things.

Maybe we should make a document on that
so that everybody can see it, you know?

And then you show up and kind of like
design share next week and you're like,

Hey everybody, I made this doc of like
all the stuff like Brian writes, you know?

Um, so I think that's where you
get this kinda like dance between

like the, the, there's kinda
like the human emergent part.

There's the documentation part.

And the documentation.

I think, I mean that in terms of like
just straight normal like documentation,

but also skills, tools, et cetera.

Uh, and that thing being viewed
as quite organic, meaning

like we keep blowing into it.

It can and should be like constantly
improving and changing both by

our own like directed means.

And sometimes like by
an agent that, you know,

Jonny: Right.

Randy: have a UX writing, you know,
kind of agent that kinda self-heal.

And then what, like our, our writer
would, you know, kind of approve,

you know, like, um, not in the like
approve it, but actually sort of

like observe it and if it's not good,
revert it kind of thing, you know?

So it's like a roll, roll forward approach
of being kinda like self-learning.

Uh, and then there's just like a, an
actual tooling part of that constellation.

So there's meaning, uh, well, the way
we might make a custom agent to do some

things and then be like, Hey everybody,
here's what the custom agent does.

Uh, or the way we've done some things
to have, you know, a production

environment in a more like prototype
environment, you know, this is like net

new tooling we've made for ourselves.

Um, but, but a lot of the tooling is
fairly like simple and not to down, I,

not to downplay, I mean, it, it adds
up into a quite like sophisticated

system and you connect it to these
behaviours and things that we have.

Um, but none of these tools are like, you
know, that radical and in fact, there's

stuff we've made ourselves, you know,
you're like bumping up against something.

I think lots of teams are doing this,
but we don't talk about it enough.

You're like, you're bumping up against
something and you're like, oh, I

kind of, I wish I could just like
point at this thing in the browser.

And you kind of make like a
low-fi version of what ends

up being like Pencil or Paper.

I'm, I keep confusing those two products.

Um, uh,

but you know uh, you kind of get
at this because it feels like

intuitively a little thing that would
be nice to have in your toolkit and

then you kind of make it yourself.

And then I don't know if some other
thing comes along, somebody has an

open source tool is some commercial
product, you might use that one instead.

Um, but yeah, that kind of tool
making part feels very of this moment.

People always did that, but
it's much easier to do that now.

Like, here's the little specialised
tool we need along the way.

Yeah.

Tom: I'm curious

on the tool making front, if you, if
you've, how you foresee that trend

evolving over the next year or two?

Like is, is that of this moment
right now or is that sort of a

discipline that's here to stay like
the, the makers of tools for internal

use and internal team operations.

Randy: I think it will continue.

Yeah, it's interesting.

I think that, I mean, you think
about something like the product

promise of something like Notion
overall, not just like design

team, but you kind of zoom out.

Part of the premise is actually that
um, there are lots of kind of like

bespoke needs and organisation may
have, and it's helpful to have this

kind of malleable surface that lots
of people can manipulate and kind of

build tools in some automated workflows,
some custom agents and different

organising methods for things, um, that
can help them get knowledge work done.

You know, that you can't kind of buy
a point solution or get some off the

shelf thing and accomplish that as
easily or as cheaply or as, as well as

when you can kind of do it yourself.

'cause lots of organisations
have like a bunch of smart stuff.

They might have a bunch of smart stuff
happen here and then some legacy system

that is just, just is the way it is inside
this organisation or some weird cultural

thing where like actually the permissions
from that to that or they just are.

And so I think this idea of like homegrown
tools and the kinda like bespoke internal

tools will we'll see, continue to
grow actually, when I think about them

from a design tool standpoint, well
maybe this is true of all tools too.

At the same time I think you'll see we're
trying a bunch of stuff and some things

are just gonna kind of like normalise
a little bit, you know, whatever.

Uh, and there's like shaders and I'm
like, I built my custom shader tool,

or I used the shader tool in Figma,
or I, you know, I, I don't know, it'll

probably normalise around some stuff.

Maybe in the same way, like, you
know, for 20 years we were all using

some of the same like open source
packages, you know, we like stand on

the shoulders of things and like at
some point, not everybody needs to make

our like home brew web type ning tool.

There might just be a couple that we are
all like, yep, that gets the job done.

But at the moment it's almost like
a, maybe it's a little, it's fun,

maybe it's a little bit of a flex.

Maybe it's a little bit of
like, I gotta, I'm gonna do an

interview, I need to show people.

I'm like AI savvy.

So like look, I built this tool, you know?

Um, it's like a proliferation of
like micro personal tool making.

Yeah.

Tom: Yeah.

Super interesting.

I really wanna switch gears and ask you
about SVA because you, uh, we talked a lot

about Notion so far and, and general ways
of working, which are very interesting.

But you, I feel like you sit in
this very unique position in the

world right now as the chair of
the MFA design programme at SVA.

And I'm curious if you could, uh, speak
a little bit to the, like this next

generation of designers that's coming
through and how, uh, what designers are

thinking about what they're worrying
about, but also as a, as the chair

of that programme, how are you, um,
intentionally, uh, constructing a

curriculum or a, or a programme that
allows designers to do well in this future

that they're arriving into, which looks
probably quite different from graduating

from master's programmes in the past.

Randy: Yeah.

Yeah.

There's a, well, I mean there's kind
of a meta version of this, which

is, I my view it, the curriculum.

It's almost like this kinda
like, I don't know, barbell kind

of like strategy or something.

There's some, uh.

Fundamental, kind of
like foundational things.

Um, I mean, at a graduate level that we
believe are important and kind of endure.

And you want to get like, real solid about
this is like ethics and design practice.

We talk about like intellectual property.

Um, I mean it's 'cause we're like a kind
of very industry and business and almost

like entrepreneurial oriented programme.

There's a bunch of things here
that are kind of universally true.

Also, some aspects of just kinda
like brand and how psychology kind

of connects to design and like
creating desire and differentiation.

All this is kind of true.

Uh, there's a bunch of the middle
stuff where you usually like don't

teach like tools and mean, you know,
like watch YouTube, but that was fine.

Uh, but there's this other end which is
like, I think must be like malleable.

And this is like, how do you kind
of structure a curriculum as a

set of containers for those things
to be able to like evolve always.

Um, what's interesting is I think that
AI literacy right now is one of these

like, new emergent things, but like, I
think very quickly it'll become one of

these things where you're like, visual
literacy, AI literacy, media literacy.

Kinda like how, you know, it's
like how propaganda works.

You know, and, or like, I don't, I
mean, broadly speaking, you know, um,

like how you shape perception, uh, and
like shape behaviour with design, uh,

and how AI like both works and
how it sort of integrates in

your practice and actually learn.

It'll probably end up down here
like pretty quickly, but like

right now it's kind of here.

The way we've done this is.

Threefold.

I'd say one is we've introduced
a couple of, well, it's an

important thing that student, uh,
or special about the programme.

I mean, other institutions may also
do this, but I mean, it's like a very

like defining characteristic here.

Uh, all of our faculty are adjunct
faculty who are all working professionals,

and so they bring into the classroom
their own professional practice and

what's happening there often, right?

And so naturally for some of our faculty,
the questions about the relationship

between like AI and design, you know,
it like flows in that way, right?

So that's one.

Two, we've changed, we've like
introduced a couple of new courses

or a couple of new classes, um, to
specifically address some things.

And then the other is we've
encouraged, um, faculty members to.

Bring some AI experimentation
into parts of some work that

was done differently previously.

And some of these are really like
experiments along the fly and we'll

kind of like see how, for example, it
was like course around like creative

direction, which is very much like in
the photography studio, like live, you

know, not like mood board, creative
direction, although there's things

like this, but it's like, you're gonna,
there's another person, here's a very

accomplished, in this case image maker.

There's like serious pro
photographers and you're like,

we've got the studio for a day.

Like, time is money.

You need to like give them direction
to get the outcomes you want.

Like, doesn't a very, you know,
this is a skill that not often, you

know, kind of taught in this way.

Right?

In that same class, we had
them do the same exercise.

We actually, this is like before,
just before Figma acquired, like Weavy

and the Weavy founders come in, do a
workshop with the students, and then

they basically did that same exercise,
but trying to like basically direct

the AI to achieve common outcomes.

Just kinda like, see what would happen.

This wasn't like a ab
like, which is better?

I mean, you could do that if you wanted,
but it was more just like different modes

of trying to achieve a similar outcome.

Um, so that's like an example of, well,
we did another, we actually had these

kinda like synthetic personas giving
feedback on thesis topic ideas from

really kind of interesting angles.

Being like, okay, you're an, you
know, like you're an attorney.

You're trying to help me, like
minimise risk in this idea.

Like what are the, you know, what are
the things I should be reading, learning

about considering in this like direction
if I want to take into account like future

liability of the things that, you know.

So things like that which are like, would
be very difficult to do previously or

just wouldn't get the level of like depth.

So it's like this cool thought
partner like kind of challenge thing.

And then of course there's some areas
who were like, oh, I'm like, let's, you

know, sit down and leave the classroom
with an iOS app after three hours.

Like, let's go.

Kinda like vi vibe, coding e stuff,
sort of hitting it from a few angles.

Um, but it's, it's evolving
like in real time, you know?

Yeah.

Jonny: How are you finding the,
um, the students respond to ai?

Generally, it's, it is quite
well publicised that, you know,

some of the, uh, uh, the speeches
given to, to graduates, especially

in the states over the summer,

Randy: Yeah, that's funny.

Jonny: look, look at, look forward to
your AI future didn't go down super well.

So, you know.

Have you, have you

Randy: Yeah, it's interesting.

Jonny: or is it.

Randy: Here's what I've
experienced so far.

It's funny, I almost, maybe it
was my own punk rock spirit.

I wish that there was like, you know, a
student who would come in and be like,

I refuse to use AI in any assignment.

You know, uh, got kind of like
politicised about it or something.

I think it was a good environment
for them to like work, work out

their relationship to that and like
try that, you know what I mean?

Uh, slightly lower risk environment
to do that than say like, you

know, going to your job tomorrow.

Jonny: Right.

Randy: but I'm not seeing that.

Actually I was seeing one of two things,
but one we did kinda like a listening tour

at the end of the previous semester with
some students and our main takeaway from

that, which was interesting, was a little
more like, yep, gotta get hip to it.

I think they're quite practical,
which was interesting.

I mean, it's the nature of this
programme, it's quite advanced, but

it is also not like experimental or
purely, it's not a conceptual programme.

We are like oriented towards
professional practice and ultimately

helping people, people be successful.

You know, uh, either like in industry,
getting jobs and things or like

sort of carving their own path,
being entrepreneurial and sort of

making so a practical orientation.

And I think the students
overall have that in common.

And so, yeah, my takeaway for this
listening tour is over overall even

for the students who are like, man,
I don't know, like it, but kind of

gotta like get hip to the tools.

'cause it seems like
that's what's gonna happen.

So there's that.

I think that's the most negative
of the sentiment I've seen was

a little bit of like, well, if I
need to, but I've also seen some

students like just totally lit up.

Lar largely, I'd say more from the.

I suppose I put this all in the, um,
here's something I was not capable of

or didn't realise I could be capable
of until encountering the technology.

Uh, we have designers who are more like
static image maker or like animation or

traditional graphic design or something.

Um, being like, oh, I can basically
prompt my way to like 3D models.

I can prompt my way to like visual effects
or I can use a combination of my like

visual inputs plus some prompting, but
like arrive at, at rich, um, time-based

media and stuff that there's some
students who are like, oh, I'm into this.

You know what I mean?

Now it's almost like a new canvas to like
use the wrong way and get really like,

creative with and others that are just
like, I don't know, I made this video.

I didn't think it'd be possible.

I could do it now.

I can like more simple edits, but
just in like an amount of time.

And so like pretty awesome.

And then the other is much
more on the software front.

We're like, oh cool, you got the bug.

One of our students like went back to
India for the summer and then made like an

ERP for his dad's like warehouse company

Jonny: Sick.

Randy: Like, oh, I can make apps now.

Like let's just do it.

Um, it's awesome.

So I think they, you see that kinda like
unlocking of like new possibilities.

Um, um, it didn't seem, maybe
it didn't even consider trying

something like that, which is funny.

That same student probably could have
done that same thing in a more manual way

Jonny: Yeah.

Randy: But the environment or like
the change of the conditions has

unlocked the like, possibility.

Be like, let's go for it.

Let's try this, let's try that.

So I see that as like a
really healthy experiment.

Jonny: makes me think of the saying.

Necessity is the mother of all invention.

You know, he, he goes back
and there's a clear need.

There's a burning problem.

Randy: Yep.

Jonny: needs an ERP builds.

It, comes back to the course
and thinks, well hold on.

Now I can do that.

What else can I

Randy: Exactly.

A hundred percent.

Yeah.

Yeah.

Jonny: this common material that
actually can go in any direction,

Randy: Totally.

Jonny: your horizons.

Randy: Yeah.

Yeah.

And then I think what's cool in an
environment like this, you know, I mean

this is true in a commercial environment
like, but you have people coming from

so many different angles with different
motivations and different intent

coming out, but they still like learn
from each other in those funny little

Jonny: Yeah.

Randy: You know, you observe someone do
something like that and you kind of twist

it and apply it to your little universe
that you're making for yourself in it.

Yeah.

Tom: Oh, drew, conscious.

We're, we're running outta time.

We have a a minute left.

Randy, do you have time for two
super quick, quick fire questions?

Randy: Okay, let's do it.

Tom: Uh, question.

I'll, just go with one.

So when AGI takes over everything,
what will you do with your free time?

Randy: Yeah, I'd probably probably
make music with other people.

Probably just like jam, you know,
a good organic jam feels real nice.

Jonny: Amazing.

Randy: Yeah.

Jonny: Love to dig into that, but this
is a quick fire round, so we'll move on.

Um, uh, what is your
favourite product right now?

Randy: This is a great question.

Well, I have to think deeply that
probably today I still, I spend more

time either like, you know, in the
in Notion or in Claude Code, um.

What is my favourite app today?

Hmm.

Actually, I'll say I probably,
my favourite app is the

new like Notion AI app.

We built like a standalone like
agent kind of orchestration app.

It's available now, like
you can go check it up.

Um, I use that constantly, um, to
kind of fire off parallel thread of

work and, um, it using my personal
life because my professional life, um,

it's serving me really well and it's
a, it's such a focused experience.

I really, it works for my brain well,
you know, it's like a nice compliment

to the, my whole universe is like
in, in the other app and this is just

like what I'm trying to do right now.

Yeah, it's pretty great.

Tom: Love

Jonny: Have

Randy: should check it up.

Yeah, yeah, yeah.

It's just called Notion Agents.

Jonny: I, I need to, I need to
dig back into, to, um, Notion's,

Randy: got the little, uh, can you see?

Yeah, yeah.

I got, it's got the little guy.

I mean, it's black on black.

Very designer hat.

But yeah, just look for the little,

Tom: can

Randy: know,

Tom: I can

Randy: Notion please.

On the f Yeah.

Tom: Love it.

Um, this has been great.

Thank you so much for
sharing your time with us.

Let's wrap up here 'cause
we're up time one minute over.

Um, but yeah, love chatting with you.

Really interesting to hear about
the, the systemic layers that you're

viewing your work at, Notion and,
and and SVA and your own practice.

Um, would love a follow up conversation
sometime in the future talking about

music because um, we just scratched the
surface there, um, right at the end.

But when AGI is here, if our podcast
still exists, we'll uh, hit you up

and learn more about your music then.

Randy: Yeah, let's do it.

That sounds great.

Tom: Thanks so much.

Randy: All right.

Take care.