February 19, 2025 • 54 mins
Chris Moore and Lucas Coelho, two-thirds of the co-founding team of venture studio Ryft, join the podcast to talk through how the model works.
They discuss the benefits and challenges of being an all-in-one startup shop, why they are focusing on artificial intelligence, and give their highly informed takes on the current state of the AI landscape.
Plus, what to look for in Spark's upcoming results and we address some listener feedback on the Being AI episode.
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
This week on the Business of Tech powered by two
Degrees Business the rise of the venture studio.
Speaker 2 (00:05):
Here we've counted at least a dozen New Zealand companies
pursuing this relatively new model, or new for New Zealand anyway,
This sort of model of creating and funding startups called
the venture studio. Unlike venture capital firmed, venture studios play
a much more hands on role in the entire life
cycle off a startup, from ideation to execution.
Speaker 1 (00:27):
We talk to the co founders of Auckland based venture
studio Rift about its approach to operating a venture studio,
with a particular focus on artificial intelligence.
Speaker 3 (00:38):
So we have all the ingredients, all those assets that
you need to bring together, and then it's about taking
what can feel a bit boring, sometimes a formulaic approach,
and yet a lot of the times you can identify
great opportunities, assemble the team, assemble the financing and then
go and tackle it.
Speaker 1 (00:57):
I'm Peter Griffin, I'm Ben More and a look at
the venture studio model of building startups coming up with
two of Rift's co founders, Lucas Coelo and Chris Moore.
Speaker 2 (01:07):
But first, then the first half results of interdex listed
spark are announced Tomorrow, that's Friday, they'll release their financial update.
Sparks had a pretty rough run in the last year
or so, a sagging sheer price, major cost cutting going on,
a number of senior people have gone from the organization,
(01:29):
and a move to divers parts of the business. What
are you expecting to see tomorrow?
Speaker 1 (01:34):
What I'm expecting is an update, first of all on
the restructuring. So that's something they promised at their and
your general meeting last year, was that they would release
more details on the restructuring, including kind of how it's
going so far, what they've managed to save. You may
recall their goal was to say fifty million on labor costs,
(01:55):
which is about ten percent of their total labor spend.
It was recently came out that they were cutting their
management staff as part of that restructure, and how they're
changing the portfolios that different managers look after, so probably
a reasonable chunk there as well. So that's what they
(02:16):
have said is that there will be more information about
the restructuring at this update, but they've been pretty reticent
about giving too much detail anyway, so very curious to
see exactly what they do release and what they don't
about that situation. So that's kind of the main thing
(02:38):
I think most people are on tender hooks about.
Speaker 2 (02:40):
Yeah, they sort of foreshadowed this last year. They reduced
their ebit dye guidance somewhat not a huge hit, but
that's down. They reduced their dividend guidance to twenty five
cents per share from twenty seven point five cents, so
they're basically softening up the market for lower expectations, and
(03:03):
you know that's been reflected in the share price. I
think it's down like thirty six percent of something over
the last year, even including dividends, so it's it's been tough.
And then there was all angst about them being delisted
from that big sort of index Overseas, and I was
expecting a big dip after that, but that was clearly
sort of priced into it beforehand, so that didn't really
(03:28):
take a massive hit. But still, you know, this is
a company that still has a big market shriff to
telecommunications industry, you know, forty something percent of mobile and
fixed line broadband, so it is you know, still you know,
the big player along with one end Z in that market,
but sort of really facing a bit of an existential crisis.
Speaker 1 (03:51):
Absolutely, especially in like broadband and mobile as well. Two
Degrees nipping at their heels along with some of the
Mbenno operator that are starting to collectively take share away
as they kind of use broadband as a lost leader
to get people onto their other services. So yeah, that's
an area where they're struggling. Interestingly, on their share price,
(04:14):
analysts are kind of a little bit mixed at the moment,
with some saying, you know, still not worth buying at
this point. Others are saying, look, this is actually a
depressed price for Spark. It's likely to go up in
the future. Now might be a good time to buy.
What to me that says is that there is a
little bit of uncertainty about what's actually going to come
(04:36):
for Spark, that there's obviously going to be a huge
effort from Spark to reassess their position, try and capture
back some of their core market and also try and
capture new market with data centers. That's been a big discussion,
so I expect we'll probably hear a bit more about
their data center strategy as well, including some efforts to
raise capital to co fund development of new data center.
Speaker 2 (05:00):
Yeah, it's been reviewing it's non core assets. So it
still owns a bit of Connects, the mobile tower business,
so it might divest the rest of that. And it's
got quite a big IT division as well, and a
few sort of units there, health, sort of tech unit, IoT,
interneted things business, it's curious data business. So do you
(05:25):
think there may be moves signal to carve off some
of these units.
Speaker 1 (05:31):
Yeah, we know Connectsha, they're staking Connection, they're looking to sell,
so that's a definite. They haven't really said yes or
no on some of the other areas, but I know
there is definitely some rumblings about what they're planning to
actually divest. We might hear more about that at this update.
(05:52):
They might not update on that, because sometimes what will
happen is they will just get their earnings out of
the way, update on some more of the directly related
fiscal stuff, and then update later on some of the
other plans around investment. That is, unless they've already made
the deals and they're announcing that they have sold x,
(06:14):
Y or z. That might happen at this earnings. I
think what's really interesting is if you look at Deleeperfonseca's
column from last year that was on Business Desk, and
he talks about Spark trying to kind of figure out
what it is now. Is it a tech company or
is it an infrastructure company, because those are two quite
(06:37):
different things, and how does it balance those in these
times of difficulty, So well worth a background read if
you're interested.
Speaker 2 (06:45):
Yeah, and it's interesting that a lot of the restructuring
is in this sort of enterprise and government sections of
the business. And on the face of it, it looks like
Spark had been laying the groundwork for really smart, DIVERSEI
vacation in recent years, with the likes of Curious, the
sort of AI and data division. It's got its IT
(07:07):
division as well, which is doing more sort of cloud
and digital transformation stuff, so you'd think like and they
published that report which we had Jolie Hodson on the
podcast about last year looking at productivity, the productivity gap
in New Zealand, the need to embrace advanced technology, so
(07:28):
sort of doing that with a focus on IoT and
health tech and the like. It just hasn't translated. And
I don't know if that's a that they got the
market wrong or if it's a just bad execution, but
more than any other talco. They were sort of saying,
we want to be about more than supplying speeds and
feeds which have low margins. There's lots of competition, we
(07:50):
want to do more, and they've sort of been not,
you know, slapped back on that, which is sort of disappointing.
I don't know if it's them or if it's us.
Speaker 1 (07:58):
Yeah, yeah, it's a big question, I mean, and that's
that was one of the things that analysts have been
talking about a lot, is like is this just the
ups and downs of the macro environment or is this
a fundamental shift in how things are operating.
Speaker 2 (08:13):
So check out Business Desk on Friday, there'll be full
coverage off the Spark results. Just before we get into
our featured interview, an update on last week's story about
being AI Ben. You went in depth into this listed
tech company and the travails it's been experiencing. Some really
interesting discussion resulted from the podcast, including on LinkedIn what
(08:37):
were people saying.
Speaker 1 (08:38):
So on your posting of that episode. The CEO of
Callahan Innovation is Stephane Korn. He is an AI expert
from way back, and he posted talking and specifically about
their project Treehouse and this idea of APIs as a
way for AI agents to enter operate, which I I thought, well,
(09:01):
you know my perspective, I thought, it's a much more
efficient and effective way for machines to talk to each
other rather than open Ay's operator approach where it tries
to navigate websites that are designed for humans. Now, Stephane
Cortan had an interesting take on that where he basically said,
one of the reasons that APIs don't really work is
(09:23):
for competitive reasons, which was something I didn't really think about.
So what was your take on some of the reasons
that they might.
Speaker 2 (09:30):
Not Well, sitting behind the website, there's a database, and
for a travel company, it could be all their dynamic
pricing and all that sort of thing. So they've got
to be really careful that they don't give away this
sort of the secret source to their business model, so
they hold that information very closely to them. I don't
think it precludes them from issuing API access. For instance,
(09:53):
an Uber would really like the idea of being integrated
into a messaging app if you can pull in information
from Uber saying there are drivers available, it's only going
to be five minutes. You know that's possible via an API.
So I think there's some level that they could do.
But I think Stephann is right. The reason why we
(10:13):
just haven't seen open ai really go big on its
operator platform this is the AI agents it's now building
with an API sort of ecosystem for that, is that
a lot of companies are sitting back going do I
want to give you that level of access? So open
ai has just forged ahead and they're basically screenscraping all
of that information off websites. It means they can move fast,
(10:36):
and they're hoping that everyone will go, I need to
have operator because it's so good, my customers want it.
So let's do something a bit more substantial here with
with APIs. So I think that will come, but there's
definitely a competitive tension there in giving a lot more
fundamental access to these websites.
Speaker 1 (10:56):
Yeah, and I think one of the concerns might be
if you look at how deep Seak used distillation, for example,
to train its model off open ais model, there's actually
if you have AI running things really rapidly, and you
could do a whole bunch of API calls to travel company.
You could start to see how they're pricing actually works,
(11:17):
and kind of use AI agents that might not be
exactly what you might want to be accessing your API.
So yeah, you know there is a possibility there all
that could potentially be mitigated through careful you know, management
of the agents and what they're able to do and
things like that in the future. But that's why I
kind of said in the last episode as well. It
(11:38):
has potential, but there's a lot of things that need
to fall into place in order to make it work.
It's not impossible that the network effect might turn out
to make this API marketplace somewhere that companies need to
be in order to get their business out there into
the world. But on the other hand, yeah, there is
(11:59):
a lot of trains, a lot of changes that will
need to occur in order for it to find that
market for Yeah, so we've gone.
Speaker 2 (12:06):
From bigless incumbents like Spike to the earliest stage startups,
and we're focusing in this episode Ben on the idea
of a venture studio. I'd heard the concept before, really
in the Silicon Valley context, but it's the first I
really heard of a local venture studio. Was the Auckland
Company previously unavailable. It had some really big success with
(12:28):
the startup track Suit, which is like a brand tracking startup.
It raised around twenty two million dollars from venture capital
firms in early twenty twenty four in a series A
fundraising round that valued it at over one hundred and
fifty million dollars. But it started with a small amount
of seed funding and the expertise of the people at
(12:50):
previously unavailable.
Speaker 1 (12:52):
And then it turns out a number of local companies
are now pursuing this model, no surprise after a success
like that, among them a New and Improved Ventures, super
Bowld Ministry of Awesome, Paloma Ventures, Edition Group, and Rift,
who we are talking to on the podcast today.
Speaker 2 (13:10):
The three co founders are familiar with building products and
have a little cash to play with after they sold
a pretty successful IT consulting firm.
Speaker 1 (13:18):
That's right. So Chris Moore, Lucas Coelo and Ben morew
All ran Rome Digital together, which sold for around forty
five million in twenty twenty one, and after a bit
of a break, they decided to put their expertise and
money to use building and seeding companies of their own.
Speaker 2 (13:34):
And so Rift was born. Let's go to Ben's interview
with Rifts Lucas Coelo and Chris Moore talking about the
venture studio model, their company Rift and air approach to
building AI products.
Speaker 1 (13:52):
Hello, and welcome to the Business of Tech. Thank you
so much for joining us. We're here to talk about
Rift AI. That's your guy project that you've been working
on for a while. Now you're out here pushing out
this studio venture studio model in New Zealand, which is
a reasonably well known model now internationally and is starting
(14:13):
to pick up pace in New Zealand as well. So
why didn't you introduce yourselves to start with?
Speaker 4 (14:19):
Yeah? Cool, Lucas co founded a Rift Ventures before that,
Ravens in about eight years ago, and before that it
was heavily involved in the startup scene in Brazil, the
startup in design and technology.
Speaker 3 (14:31):
Seeing Chris so I was a co founder in Rome Digital,
which was a product design studio. Actually the missing third
party here is Ben who's also part of Rift. So
we've pulled together kind of the same team that we
had previously. Then we exited that in twenty twenty one
(14:53):
to a ten billion dollar US listed company, took a
bit of time off and took a bit of the
cash and then got the old team back together and
kicked off rift of the inches go.
Speaker 1 (15:05):
And that story we talked about, we've talked about before.
I've got a story up on business Desk that listeners
can read if they want a bit more background. How
does it work with a venture studio? Are you guys
developing the products? Are you investing in the products? What
does it actually mean?
Speaker 4 (15:21):
In summary, there are a couple of entities within sort
of the studio model. We have the studio itself that
produces the ideas. So we have a backlog of ideas.
Were always fishing for ideas, even within our immediate sort
of team. Our team now it's composed at about eleven people
with doing engineering, design, growth, marketing, and we always sort
(15:44):
of have a large backlog of ideas. Ideas are free
right that we keep adding ideas in there for us
to take a look at. And one thing that we
have is kind of like sort of like a playbook
on a few of the criterias of how we evaluate
ideas in order to turn them into opportunities, move them
up onto the pipeline. And that's when we start doing
(16:05):
a little bit of a desk research, understanding more of
the vertical that that idea plays in the industry, sort
of the product can look like, and then we start prototyping,
testing and see what sort of products or or companies
Mike can come out of that. So, for example, last
year our backlog of ideas had about one hundred or
(16:28):
something ideas. It's ideas that from the time of Rome
ideas that were sall online from other companies doing in overseas.
We also did some vertical sort of days or weeks
that we're going to speak with people within a certain vertical,
for example health tech to understand sort of their problems
and what things could we tackle, problems could we solve
(16:50):
there with with sort of the heavy sort of II mindset.
And then from quite a few of those we we
did a first sort of like a hot or no
quick rating on those ideas, thinking about what is the
competition locally? Can we build this in New Zealand? Is
there an entry way here that we can build this
fast in order to expand with the product later on?
(17:14):
The future sets are later on, especially because of a
little limited resource of of inentro studio. Do you want
to grab something that is too grandios or too expensive
to bute that requires millions and millions of dollars to
but so it needs to be something that we can
uh that can we can buy a big chunk off
with our with our limited capability at least at the beginning.
(17:37):
And then from that we tested it about. We prototype
like design and researched about ten of them a little
bit deeper, and that resulted into a few of the
ones that we have today. So after we take the
products through the through the validation process, and that might
be do we get customers, do we get some early
(17:59):
custom there's interest, what is the the is it feasible
to build? After we do that process, we take them.
We also have a fund inside Rift, so Riff the
studio and Rift the fund. So we have our first
fund called Rift zero. Now where did the first close
last year? And we take those opportunities into the fund
(18:24):
evaluate with a pretty brecorous sort of criteria, evaluate like
if someone pitching to us and if it passes or not.
We also invest in all those opportunities as well. That's
why we say that they graduated from the studio and
we spin them off as portfolio companies instead of being
(18:44):
just experiments. So it's it's both.
Speaker 1 (18:48):
Yeah, it's it's quite interesting because obviously the traditional startup
method is you raise a fund and then you go
and look for talent or startups to invest into. And
my fund might have a specific focus on a vertical
or a stage or whatever you like. So how is
(19:09):
that discussion with kind of LPs to say, not only
are you trusting us to do what's right with your money,
you're trusting us to develop the tech that's going to
bring you the return on investments. So you're starting, you're
pitching startups, and you're pitching a fund at the same time.
How does that conversation go.
Speaker 3 (19:28):
Yeah, I mean it's a more complex conversation than a
standard VC. You know, when a VC is raising funds,
they just say, look, we're going to invest in these areas.
We don't know what the companies will be, we believe
that we'll be able to find them as a studio
and as a fund. Rift zero also has an investment thesis.
It has four different areas that it's looking at. We
(19:50):
are mainly an AI driven fund, looking at how AI
is used in the companies as well as the products themselves.
So looking at how we can stretch the But those
four areas are across productivity, security, compliance, health and finance
as for getting people to sort of understand like does
(20:11):
Rift have that capability and skill. You know, we spent
the last ten years building companies and products for a
lot of startups and corporates. So we built companies for
X fifteen, which is CBA's accelerator program. We built them
for Westpac Ventures, we built them for ASB, B and Z,
(20:34):
the Warehouse Group, Mercury Energy, sky City. You know, we've
been doing this type of product building for ten years
as a group and before that. We also have that
personal experience of building companies in Silicon Valley and in
the UK, in South America. You know, for us, the
building and the technology and all those parts aren't really
(20:56):
the risk. Adventure Studio is there to de risk things
around repeatable process which we have from time at Rome
skilled entrepreneurs that we have within the studio as well
as within our wider network. You know, we would have
had about two hundred and fifty employees come through Rome
at the time when we sold it, we had one
(21:17):
hundred and fifty, So it's a large network across Australia
and New Zealand, Singapore as well as reach into the US.
So we have all the ingredients, all those assets that
you need to bring together, and then it's about taking
what can feel a bit boring, sometimes a formulaic approach,
and yet a lot of the times you can identify
(21:40):
great opportunities, assemble the team, assemble the financing, and then
go and tackle it. It's actually more similar to how
Silicon Valley startups are funded. Someone sees an idea or
an opportunity, they assemble the team, they assemble the capital,
and then they go and target and tackle the market.
In the past have seen a lot of New Zealand startups.
(22:02):
It's kind of they've been in the market for a while,
scraping along a little bit of friends and family funding,
and they just happen to hit an inflection point where
the market needs the product they have. This style is
quite different, I mean.
Speaker 1 (22:18):
And that's one of the questions as well that I
was interested in, which is, you know, you have the capability,
and you have the processes, and you have the experience
and the talent, but the X factor, the key thing
is finding the right place to apply those and that's
something that's talked about significantly in New Zealand. We have
limited resources in this country, we have limited capital, and
(22:42):
so we need to be quite specific about how we
are targeting our solutions. So how do you think through that?
Speaker 4 (22:52):
Yeah, I think it's the what I was saying before.
Like when we're looking at the vertical just for example,
a product that we have like chain, we start thinking
about the first idea, remember that was there on the
backlog was AI power job management software tool. There are
(23:15):
a few of them, a few of them are very successful.
Every once in a while they get acquired by larger
companies that which means that sometimes the customer service is
not what it used to be. And therefore, especially on
the SMBs, they start looking for all right, so the
prices went up. What else can I start using it?
It might be interesting to have something that, especially with
(23:37):
vertical AI businesses, you're always trying to look into, let's
look at automating things that they're not properly automated on
sort of the traditional job management tools, and let's start
let's focus on the product on that. But then while
we were taking that idea through the validation process and
we had a prototype of the tool, we had a
(23:58):
couple of sort of design screens and starts check with
so traders, with builders, with treadees, we're always trying to
think about an angle, that is, what is the what
is the line of this experience that will deliver them
instance of r O I that will solve an immediate
problem or or something that they want out to make today.
(24:20):
What can we build that will be fast to build,
that we can solve that problem. They will they will
pay for us to solve that problem, and then we
can build the rest of the capability afterwards, after the
proof that that piece of that piece of the of
the of the solution. And that's when, for example, we
start testing the in this case, the I think we
(24:42):
call a builder boost the the job management software lending page.
We have a couple of prototype screens, and after every
interview they will say, yeah, it's really cool, it looks
really cool, but not what my problem is? I hate
doing this, I doing the admin work. I hate all
of the tools that already exists. Yeah, this looks nice,
(25:04):
but I just want to I just can't pick up
my phone while am I work and I lose jobs
because of it. Can you solve someone? Do you know
anyone that can do that and then it qualify the
leads for us. And that's when Jane came about.
Speaker 2 (25:17):
Right.
Speaker 4 (25:17):
They were like, yeah, that's something that we can solve,
and we can solve reasonably fast.
Speaker 1 (25:22):
And Jane, is your l M powered super intelligent the
voicemail that can understand the context of what's being spoken
verbally through the phone.
Speaker 4 (25:33):
Yeah. Yeah, we're looking at now at the vision of
being this working towards being on the vision being this
Oh encompassy our encompassing sort of inbound lead qualification for
Assouli traders. Right, so not only qualifying leads that come
through the phone, it also can schedule sort of appointments
(25:54):
through the phone. It's on your Google calendar and in
your tool of choice and also integrating with WhatsApp. Business
noticed that it's a lot of small businesses that their
value proposition is helping some traders set up their sas
tools and we're like, okay, so this is a this
(26:17):
is really a gap in the market, Like people are
trying to solve this by being a consultancy to to
this is how that's how hard it is and how
uh that that it shows the friction that exists with
the things that exist today and how without without even
(26:38):
everyone that we spoke with they're like, yeah, we don't.
We want to spend less time doing that so I
can spend more time just doing the work getting paid.
Speaker 1 (26:49):
Interesting. Yeah, you're a New Zealand based business, you have
international ties, so I'm assuming and the question isn't will
New Zealand does pay for this, but it's will people
somewhere pay for this? And where is that? So that
must be another consideration as well.
Speaker 3 (27:05):
Yeah, you know, New Zealand market is far too small,
so if anything would have to be a minimum of
a regional market Australia and New Zealand, say Asia, so
apax but predominantly of course we're targeting the US, but
you have to start somewhere where it's easy to get
some initial feedback. But you don't want to stay with
that story of New Zealand being the trial market and
(27:28):
just two years later it's still your trial market. It's
really just used during the validation phase. Then once the
portfolio company is really to be launched, you're straight into
the US.
Speaker 4 (27:38):
Yeah. I find the New Zealand market, especially coming from overseas,
a very interesting market for tech in comparison to what
I'm used to in the US or in Brazil. I
think in the US, of course, depending on a little
bit of the state, but same as in Brazil. But
people tend to swarm more into new things, and in
(28:01):
New Zealand, I feel it's a little bit the majority
of the customers they wait someone to test it first
before they jump into it. So I tend to think
that a few of these things, especially on the products,
if you can sell it here to New Zealanders and
can convince a convince a very traditional sort of trade
(28:22):
to adopt that into its workflow, that's a good signal
that you can probably get some more adoption overseas.
Speaker 1 (28:29):
That's really interesting. Yeah, and I guess because you aren't
thinking of New Zealand as your primary market. You know,
I've heard stories before of startups who have had great
ideas that could have great ROI for trade's but they
just weren't able to convince the trades to actually adopt
it at a scale that was able to provide some
(28:51):
assurance to investors before unfortunately the companies collapsed. Not because
you guys have the venture model, because you aren't seeing
New Zealand as your prime market, but as a kind
of like trial market. That allows you to then go, well,
look now we have the set, we can throw it
somewhere that's a bit more swarmy and see and see
what happens there.
Speaker 4 (29:11):
Yeah.
Speaker 3 (29:12):
The venture studio model also allows you to play with
the different leavers of engineering marketing spend across all of
your portfolio companies. So if you have a traditional startup,
you hire an engineering team or you've got some co
founders as engineers, and that's great as you get a
bit of scale. Then every month you're paying for the
engineering team and you just got to keep feeding them
(29:34):
with work to be done. Now, some of the time
it's actually you don't need more features. What you need
to do is be selling what you have and get
the feedback from market. I mean, I think you know,
there's there's a lot of stats on I think it's
like forty percent of the features you build aren't even
the ones that customers adopt. And I've seen it myself
now with a venture studio, because you have a portfolio
(29:55):
of these, you can pull back some of the engineering
talent from one of the portfolio companies should into the
second one. While you're looking for the right market signal
waiting for market adoption. AI is one that you know,
in some cases is fast and other cases is slow.
It's so trying to get the timing right is I
believe easier with the venture studio where you can move
(30:17):
the burn around across the portfolio companies.
Speaker 1 (30:20):
This is something I wanted to talk to you guys
about as well, because we have in the I guess
really end of last year beginning of this year, we've
started to see this kind of counter rhetoric towards AI
and generative AI. And you know that this this idea
that open AI and it's ken who have been funneling
(30:46):
billions and billions of dollars into the development of the
technology are ultimately going to fail to see any actual
return on that investment just because of the scale of it.
And so you know, with the recent release of deep
Seek as well, that has simplified and made those models
maybe a little bit more affordable to actually run, that
(31:10):
kind of proves that there's going to be a difficulty
in getting a return on the investment in the long
run because they can't justify the expense when there's going
to be cheaper alternatives. What's your thoughts around that, and
how for companies like yourselves that are building on these
AI platforms that may represent a risk around the continuation
(31:32):
of the technology into the future.
Speaker 3 (31:34):
Yeah, if I sort of think back to two thousand
when you had you know, the Internet bubble and boom,
and you had this huge buildout of infrastructure and companies
like you know, Cisco, Juniper Networks, etc. We're riding high,
huge investments to get people more bandwidth. And back then
you had a ADSL link you know, one point five
(31:56):
megabit or something, and everyone was like, yeah, that's that's amazing.
I can now watch some standard DEF video with twenty
frames per second. What would I need more for? And
you know, you're fast forward now and you've got five
G networks where you can do video streaming, you know,
on the go in your car, you've got five to
(32:16):
your house with one hundred to three hundred or eight
hundred megabit down Like, people will find a way to
consume bandwidth. It's going to be the same with AI compute.
Like you know, deep seat coming out as great as
open source, it's already been ripped apart in bits of
being put into existing models, ignoring the supposed cost because
(32:39):
there's a lot of issues around. Well, actually they did
have access to fifty thousand in video GPUs and they
did spend more money than the kind of the sticker price.
You know, that's that's great. You need to bring the
costs down. We're already seeing that with open Ai. Like
they're the three mini models like ninety three percent cheaper
(33:00):
than their at last one. It's faster, less latency, which
means you can do more of these real time voice,
real time communications, real time robotics. It's does really feel
like we're at the beginning of this expansion. We do
see it with some of the AI startups, you know,
we you know, we we review a lot to see
(33:22):
where they are, what's the state of the market, you know,
what's real, what isn't And a lot of times you'll
click through and it will be coming soon, you know,
sign up for a better access. But we also know
because we're using it and building products that are in
market that customers are using, and it's today solving a
problem that it does work for you know a fair
(33:44):
number of use cases that are still there to be
to be tackled. And yeah, it's the worst it's going
to be at the moment and it just keeps getting better.
Runway Sourer for image generation mid you know, for video
generation mid journey, like the disruption is massive.
Speaker 1 (34:03):
Mm.
Speaker 4 (34:04):
I still think there's a lot of case that the
use cases that we at you they're the low hanging
through use cases for l lambs that people are finding
out and then now we're figuring out that, oh wait
a second, we might not need other lambs for everything.
We can use a couple of the smaller models there
will be much more useful and choose some specific verticals
(34:25):
or for some specific sort of tasks or even a
sort of agenta task and that's much eeaper to run.
So mind you start using that and also thinking AI
researcher open ai pulls it on on Twitter x formerly
Twitter about that more money now is being spent on
(34:48):
inference then on pre training, So more money even spending
on what can this do and can we learn things
that there's no data for it before, which will become
more useful. So even the CAPEC spend is going silver structure, yeah,
and the infrastructure is going a lot. They're spending a
lot of money on building the infrastructure for that. I
(35:11):
do think that if the market is expecting an immediate
return for this investment. That might be especially for this
companies they are spending a lot.
Speaker 1 (35:23):
You mentioned kind of in video, and you mentioned Cisco,
and one of the observations was with Cisco. Obviously in
the early two thousands they had that massive wipeout of
a lot of their value and there was a similar blip,
but in video probably was it was not obviously of
the same scale. But the argument was that these companies
(35:45):
that are being continually inflated, the value is being inflated
by the the the pouring of billion dollars into R
and D by these companies, that we will start to
see a similar popping off that value bubble, not necessarily
around the quality of the tech, but the value bubble
(36:05):
in itself. Does that concern you guys as having both
the products and the fund.
Speaker 3 (36:12):
I think on the end Video one, what's quite interesting
is you know that they've managed to ride three waves.
They've managed to ride the GPU wave, when you know
graphics move from being calculated on CPU offload. Then they've
managed to ride the cryptomning wave where it moved from
CPU to GPU and then GPU to ASK. And now
(36:32):
they've managed to ride this third wave with AI, you know, phenomenal,
But there's spent a lot of time making sure that
there's software. It's it's not just the chip, it's the
software that you use around it. Kuda, which is the
language they use for programming those GPUs. But you are
seeing you know, Amazon have their own processor. I think
(36:54):
it's called Trainium. You know, like Google have got their
tensor processor unit, the TPU. People are coming for that crown.
Of course, it's so rich. The good news is, you know, yes,
Cisco wore that bubble and that bubble burst, but when
you look at the major players, you know, Google and
(37:15):
Facebook and all the social media Instagram, they all came
out after that collapse, after the infrastructure burst. Also, there
was a lot of costs that the US government was
selling Spectrum at the time to the telpots, so that
sucked a lot of their cap becks up as well. So, yeah,
if we were in the chip space, if we're in
(37:37):
the model space, I think that's going to be cut throat.
You're going to end up with open source, and you're
going to end up with Microsoft, which is open AI.
You're going to end up with Amazon, which is anthropic,
and you're going to end up with Google, with with
Gemini and you know, and you know under open source
of course you got Lama with Meta and all these others.
Like I wouldn't want to be in that space, application space,
(37:59):
you know, being able to look at what tasks administration
work you can replace. Yeah, that's that's the place to play.
Speaker 1 (38:06):
Yeah, because you're not concerned about being caught in the
inner potential bubble burth.
Speaker 3 (38:14):
If your companies are making money and we're not building
companies that are growth at all costs its balance of
growth and profitability as a venture studio, that's okay.
Speaker 4 (38:26):
Yeah, because we're we're consuming more the commodity that they
provide in terms of the models than being caught up
on developing the models ourselves. And depending on who wins
this race, it's an open source with the excellent models
that that exist, and we're building with a with a
(38:47):
paradigm that is, we're building our capabilities in our future
is based on on the available sort of capabilities of
the of this commodity that exists and where it might
be in a fewure based on the trend of development
that this exists. But we can consume from any player. Yeah,
that it's out there, so we can consume the open
(39:08):
Eye one the Anthropic one, because they're probably going to
be very converged. They are converge to be very similar,
at least that it looks today. I'm not sure tomorrow,
Open a I might release something that we look and say,
oh my god, this is ten years ahead of everyone else,
as it was with the Chat GPT, although some might
(39:29):
argue that other companies had what the first GPT was,
they just didn't want to release to the public, so
Opening I was not that ahead as others might think.
But they seem to be very ethnical.
Speaker 3 (39:41):
Nick and Nick.
Speaker 4 (39:42):
It just continuously and there are some that are For example,
even in a few other products, we don't use one
model we use. We might use multiple models they're better,
or smaller models they are better in each task. There's
Claude is excellent for coding for for some reason, and
I'm not sure what they added into the the find
(40:04):
the tuning or on the model, but it's I find
it much better for coding than GPT. So you usually
want I'm use in cursor, I want to look into
coding or use Claude instead of using Chat GIPT. I
think it's just better, especially with Python. I might see
that other companies that are consuming a few of these
(40:26):
models my pick and choice and do the same because
I think the secret is still on. The secret selves
of companies still be the services they provide. They are
solving a problem. How is the experienced layer, how is
that sort of application layer more than the commodity itself?
Chatting with someone that is like, as long as there's
(40:48):
coffee to be bought, I want to keep building Starbucks
shops or nice coffee shops. I don't want to build
a farm to start producing coffee because that would that
would probably be a RaSE to zero or very very soon,
if not that already started. And as you're saying, with
a deep seek, wiping out quite a few, quite a
(41:09):
bit of the value of these companies.
Speaker 1 (41:11):
Interesting, I do enjoy that you said Starbucks or nice
coffee shops. I think that's that's very to be honest.
Speaker 4 (41:19):
Yeah, I don't mind star Yeah.
Speaker 1 (41:24):
That's great. I mean, I know we're probably over is that,
But I just one more question I want to ask
around that idea as well. Was there's also a commentary
that we might be hitting a ceiling in terms of
capability around L l ms, talking in particular around hallucinations.
You know, it's a feature, it's not something we can
(41:45):
get rid of talking about the cognitive in quotes capabilities
of these lms that the rate of increase is now
so minimal that actually it looks like we might have
hit the ceiling on that, and so building with the
idea that there's going to be something better in the
(42:05):
future may be a mistake, if that makes sense. What's
your take on that perspective.
Speaker 3 (42:12):
I guess one thing is the companies we're building at
the moment take advantage of the current state of the art.
We're very cautious about building for the crossover point of
some capability that might come in eighteen twenty four months.
So I think, yeah, if you're building companies, you need
to make sure you're building with what can be achieved today.
(42:34):
That's why we see a lot that are hidden behind
sign ups and join the wait lists. They were also
waiting for a tech shift. So that's pretty critical. And
it's not to say that the evolution of the product
cannot look to future capabilities, but your sellable kind of
wedge in today should be able, well not should, it
has to actually work. That's one of the things as
(42:57):
far as have we hit the limits, I think what's
interesting is, you know you have something like deep Seat
come out and it's got a whole bunch of these
kind of clever optimizations you know in ways of doing
things like people are going to continually find these. So
if you can use it to train your model at
a tenth the price, well, now you can train a
(43:20):
model that's ten times as large. Can you find more data?
Like I said, that's one of the big sources. So
then you start looking at synthetic data careful of hallucinations.
There's a lot of smart people, there's a lot of money.
I just don't see us hitting a wall in the
next next few years. The same as bandwidth. Again it's
(43:40):
a weird analogy, but remember thinking, you know, one point
five megabit was so fast because you had a twenty
eight K modem before that, and you know you used
to get like a say in New Zealand, a T
one or it's the US T one or an E
one line two megabit, and it would cost you ten
thousand dollars a month to connect Auckland to Wellington. It
take thirty five calls, and that just seems so quaint,
(44:04):
but at the time it would have been well, that's
the fastest that we can get it across the network,
and all of these other reasons where there's demand, you know,
money and innovation follows. I don't think we've been hitting
limits soon.
Speaker 4 (44:17):
And then's the bit about we're talked about building with
the future capability in mind, with CRUs Is saying as well,
is that there's a lot of the not a lot,
but they are parts of the workflows of a few
of the products that we're working with that they're just
ultimately like that llms are not capable of doing right now.
And then you what it do is that you do
(44:38):
that part of workflow with your standard automation sort of
standard code there. But it generates is that it starts
when people start using it generates a very specific sort
of the amount of data that you can use a
small model to be very good at that very specific
sort of workflow, which is something that starts building different
(45:00):
right It'll be very hard to build differentiation if you're
just wrapping your company around whatever is the large the
data set the lms are trained on, because then anyone
that just pluts into the data set will be able
to to roughly perform the same task. But when you
viewed something that solves a very specific sort of touch
(45:21):
point and you start getting your model into it and
then using that data to train your capability, I think
then you start creating differentiation and something that it's more defensible,
especially in the in a world where anyone can go
to cursor or go to what's the zero from A
(45:43):
and build a SaaS company in a in a week
or so, it would be very interesting to see how
it goes.
Speaker 1 (45:52):
It will be it will be very interesting.
Speaker 4 (45:54):
Yeah, very interesting. In the same moment that said that
the first billion dollar company of one person should should
should arrive very soon.
Speaker 1 (46:10):
Some helpments ad a lot of things.
Speaker 2 (46:19):
We'll put a couple of articles in the show notes,
some really interesting articles, including from New Zealand companies sort
of explaining the pros, mainly the pros of the venture
studio model. You know, in my mind, really I think
the big advantage here and I think tracks who really
illustrates this. You have, you know, an advertising guru like
(46:42):
James Herman previously unavailable, Simon Pound, people like that who
really understand marketing, advertising branding. They come across some entrepreneurs
who want to set up something in that space, and
they have deep integral knowledge of that and and what
works and what doesn't. So you invite them in, maybe
(47:03):
give them a little bit of seed funding, but mainly
take equity and return for nurturing this company, giving it
your expertise. That seems to be the venture studio model,
and it then allows a company to get to a
point where it's really attractive to venture capital companies. And
I think that's a really good model for New Zealand,
(47:25):
where we do have these areas of deep expertise, but
we don't necessarily have a lot of seed funding. So
you're seeing these people who maybe have created successful companies
exited them, have a little bit of capital and are
willing to back these teams and do numerous teams at once,
so not just put everything into one company. You have
(47:47):
half a dozen of these companies under your roof, maybe
even hot desking or sharing space, looking over the shoulder,
helping them, mentor them, introduce them to your networks and
business context. It seems to be quite a good model
for New Zealand.
Speaker 4 (48:03):
Yeah.
Speaker 1 (48:04):
Absolutely, And also it does help to slightly address the
talent issue as well. Because if you have it, like
they say, they have a group of engineers within Rift
that they can then choose how to where to kind
of direct their talent. So two different different companies say
they've got one thing over here, and they're like, oh,
we really need a new feature on that asap. Let's
(48:26):
move sixty percent of our engineering over to that. For
the meantime, you don't have to suddenly scale up and
get a bunch of contractors and figure, you know, get
them to get on board with the systems and learn
new things. It's just this kind of this core at
rift that can do things as they need them to do.
So another really great way of addressing some of New
(48:47):
Zealand's shortfalls.
Speaker 2 (48:48):
Yeah, I guess looking at the potential cons of this model.
So clearly, you know, the people who are running these
venture studios are spreading their risk but spreading their attention
across a number of ventures. So you know, you may
be diluting the attention that you can pay to a
(49:10):
particular company. And I've seen a couple of podcasts where
people who have been through venture studios and sort of say, well,
it's you know, sort of feel like I'm competing with
other companies in this venture studio for the attention of
the people who are funding this and supporting us, So
I guess that's the downside. But having said that, we're
pretty collegial in New Zealand, so you're more likely to
(49:33):
actually be able to leverage off what other companies sort
of in the venture studio stable are actually doing, to
try and learn from them and maybe even partner on
business ventures.
Speaker 1 (49:44):
Yeah, exactly. And what I think is interesting about ref
as well as that particular focus on artificial intelligence because
it is such an evolving sector and it kind of
does link into that idea of talent because the leadership
team they spend a lot of time I'm looking at
different research papers about AI, about generative AI and its applications,
(50:06):
and then they can disseminate that information to their ideas
for startups, rather than you know, having to kind of
each individual company follow along as best they can and
split that time. So it's another benefit as well. But
it's also interesting they're focused on AI, especially in this
kind of very AI heavy new world that we're in
(50:29):
the conversation about whether it's a bubble. It seems to
me from a conversation with them with the founders that
they are quite focused in their approach. They're wanting to
kind of really really focused on what products are actually
finding product market fit and value and go from there,
(50:50):
rather than trying to kind of have a product and
get a bunch of hype around it and pump a
bunch of VC in it and then see where you
go from there. Maybe that's a little I don't know.
Speaker 2 (51:01):
Yeah, they're clearly using AI and it was really an
interesting that discussion about their reflections on Deep Seek and
that move from sort of training to inference, which is
sort of changing the priorities in the market at the moment,
and also their view of the AI landscape. It seems
as though they're expecting, maybe a bit belatedly, a wave
(51:23):
of AI startups to sort of emerriage. I mean, they
are emerging, but it hasn't been as pronounced as say
in the Australian market, so it's good to hear that
that's coming. But you know what they're doing. I think
with the couple of ventures that have got traction so far,
it sort of harkens back to what we were talking
to Rowan Simpson about last week, is not necessarily being
(51:44):
first but trying to be the best, you know, so
the startup they were talking about that basically makes it
easy to take the admin out of being a small business,
you know, for tradees and that sort of thing. Sure
we've got if I have done that others we seem
to have a timely to another one. I guess we
seem to have some expertise in this area in New Zealand,
(52:08):
but they're taking a slightly different approach to it, applying
AI to it where appropriate. That's not necessarily a revolutionary idea,
but it's all about executing that well absolutely.
Speaker 1 (52:19):
And they also finished raising their first fund as well,
so as well as using internal money, they are raising
a fund that they can feed money into their own
studios as well. So that obviously shows some confidence in
the market in what Rift is doing.
Speaker 2 (52:37):
I think they've put five million into the zero Fund
Zero Fund, and that seems to be quite a bit
for a venture studio. Although I was looking at New
and Improved Ventures, which really the same people behind that
were previously unavailable. They've sort of set up a new
venture studio I think with the likes of ice House Ventures.
They've raised six million dollars to back three for new firms.
(53:01):
That's what The Herald reported late last year. So there's
definitely some good seed money starting to flow into these
and the reason is, I think statistics I've seen is
that they're actually more successful. Your chance of getting to
the next stage is greater out of these venture studios,
which have actually been around since the mid nineties in
(53:22):
Silicon Valley. So they've nailed that sort of process, and
according to some sources that I've seen online as well,
that they allow you to get there quicker. So it's
more rapid way to go from ideation to execution, which
obviously that's what is all about. Has been quick to market.
Speaker 1 (53:44):
Yeah, and they're not necessarily looking for unicorns as well, Like, yes,
there is some sense of it would be nice to
have one, of course, but at the end of the day,
a flow of one hundred to two hundred million dollar
companies that they can sell and then reinvat, or they
can split off and can go and have lives of
their own. That's also something that you actually do need.
(54:06):
We've talked about it before, that middle layer of companies
that are continuing to grow, be sold or change or
whatever else it is, and feeding talent and money back
into the market as well. It's not all about the unicorns.
Speaker 2 (54:19):
Sounds good power to them.
Speaker 1 (54:22):
So a big thanks to Lucas Coelo and Chris Moore
for coming on, and Ben Morrow as well for helping
set that up.
Speaker 2 (54:29):
More on New Zealand's growing band of venture studios in
the show notes. Head to the podcast section at Business
Desk dot co dot d.
Speaker 1 (54:36):
You can stream the podcast there or on iHeartRadio, and
of course it's available on your podcast platform of choice.
Speaker 2 (54:43):
Get in touch with your feedback and topic suggestions. We're
on LinkedIn and blue Sky these days.
Speaker 1 (54:48):
Catch us again for the next episode next Thursday.
Speaker 2 (54:51):
See you then,
This week on the Business of Tech powered by two
Degrees Business the rise of the venture studio.
Speaker 2 (00:05):
Here we've counted at least a dozen New Zealand companies
pursuing this relatively new model, or new for New Zealand anyway,
This sort of model of creating and funding startups called
the venture studio. Unlike venture capital firmed, venture studios play
a much more hands on role in the entire life
cycle off a startup, from ideation to execution.
Speaker 1 (00:27):
We talk to the co founders of Auckland based venture
studio Rift about its approach to operating a venture studio,
with a particular focus on artificial intelligence.
Speaker 3 (00:38):
So we have all the ingredients, all those assets that
you need to bring together, and then it's about taking
what can feel a bit boring, sometimes a formulaic approach,
and yet a lot of the times you can identify
great opportunities, assemble the team, assemble the financing and then
go and tackle it.
Speaker 1 (00:57):
I'm Peter Griffin, I'm Ben More and a look at
the venture studio model of building startups coming up with
two of Rift's co founders, Lucas Coelo and Chris Moore.
Speaker 2 (01:07):
But first, then the first half results of interdex listed
spark are announced Tomorrow, that's Friday, they'll release their financial update.
Sparks had a pretty rough run in the last year
or so, a sagging sheer price, major cost cutting going on,
a number of senior people have gone from the organization,
(01:29):
and a move to divers parts of the business. What
are you expecting to see tomorrow?
Speaker 1 (01:34):
What I'm expecting is an update, first of all on
the restructuring. So that's something they promised at their and
your general meeting last year, was that they would release
more details on the restructuring, including kind of how it's
going so far, what they've managed to save. You may
recall their goal was to say fifty million on labor costs,
(01:55):
which is about ten percent of their total labor spend.
It was recently came out that they were cutting their
management staff as part of that restructure, and how they're
changing the portfolios that different managers look after, so probably
a reasonable chunk there as well. So that's what they
(02:16):
have said is that there will be more information about
the restructuring at this update, but they've been pretty reticent
about giving too much detail anyway, so very curious to
see exactly what they do release and what they don't
about that situation. So that's kind of the main thing
(02:38):
I think most people are on tender hooks about.
Speaker 2 (02:40):
Yeah, they sort of foreshadowed this last year. They reduced
their ebit dye guidance somewhat not a huge hit, but
that's down. They reduced their dividend guidance to twenty five
cents per share from twenty seven point five cents, so
they're basically softening up the market for lower expectations, and
(03:03):
you know that's been reflected in the share price. I
think it's down like thirty six percent of something over
the last year, even including dividends, so it's it's been tough.
And then there was all angst about them being delisted
from that big sort of index Overseas, and I was
expecting a big dip after that, but that was clearly
sort of priced into it beforehand, so that didn't really
(03:28):
take a massive hit. But still, you know, this is
a company that still has a big market shriff to
telecommunications industry, you know, forty something percent of mobile and
fixed line broadband, so it is you know, still you know,
the big player along with one end Z in that market,
but sort of really facing a bit of an existential crisis.
Speaker 1 (03:51):
Absolutely, especially in like broadband and mobile as well. Two
Degrees nipping at their heels along with some of the
Mbenno operator that are starting to collectively take share away
as they kind of use broadband as a lost leader
to get people onto their other services. So yeah, that's
an area where they're struggling. Interestingly, on their share price,
(04:14):
analysts are kind of a little bit mixed at the moment,
with some saying, you know, still not worth buying at
this point. Others are saying, look, this is actually a
depressed price for Spark. It's likely to go up in
the future. Now might be a good time to buy.
What to me that says is that there is a
little bit of uncertainty about what's actually going to come
(04:36):
for Spark, that there's obviously going to be a huge
effort from Spark to reassess their position, try and capture
back some of their core market and also try and
capture new market with data centers. That's been a big discussion,
so I expect we'll probably hear a bit more about
their data center strategy as well, including some efforts to
raise capital to co fund development of new data center.
Speaker 2 (05:00):
Yeah, it's been reviewing it's non core assets. So it
still owns a bit of Connects, the mobile tower business,
so it might divest the rest of that. And it's
got quite a big IT division as well, and a
few sort of units there, health, sort of tech unit, IoT,
interneted things business, it's curious data business. So do you
(05:25):
think there may be moves signal to carve off some
of these units.
Speaker 1 (05:31):
Yeah, we know Connectsha, they're staking Connection, they're looking to sell,
so that's a definite. They haven't really said yes or
no on some of the other areas, but I know
there is definitely some rumblings about what they're planning to
actually divest. We might hear more about that at this update.
(05:52):
They might not update on that, because sometimes what will
happen is they will just get their earnings out of
the way, update on some more of the directly related
fiscal stuff, and then update later on some of the
other plans around investment. That is, unless they've already made
the deals and they're announcing that they have sold x,
(06:14):
Y or z. That might happen at this earnings. I
think what's really interesting is if you look at Deleeperfonseca's
column from last year that was on Business Desk, and
he talks about Spark trying to kind of figure out
what it is now. Is it a tech company or
is it an infrastructure company, because those are two quite
(06:37):
different things, and how does it balance those in these
times of difficulty, So well worth a background read if
you're interested.
Speaker 2 (06:45):
Yeah, and it's interesting that a lot of the restructuring
is in this sort of enterprise and government sections of
the business. And on the face of it, it looks like
Spark had been laying the groundwork for really smart, DIVERSEI
vacation in recent years, with the likes of Curious, the
sort of AI and data division. It's got its IT
(07:07):
division as well, which is doing more sort of cloud
and digital transformation stuff, so you'd think like and they
published that report which we had Jolie Hodson on the
podcast about last year looking at productivity, the productivity gap
in New Zealand, the need to embrace advanced technology, so
(07:28):
sort of doing that with a focus on IoT and
health tech and the like. It just hasn't translated. And
I don't know if that's a that they got the
market wrong or if it's a just bad execution, but
more than any other talco. They were sort of saying,
we want to be about more than supplying speeds and
feeds which have low margins. There's lots of competition, we
(07:50):
want to do more, and they've sort of been not,
you know, slapped back on that, which is sort of disappointing.
I don't know if it's them or if it's us.
Speaker 1 (07:58):
Yeah, yeah, it's a big question, I mean, and that's
that was one of the things that analysts have been
talking about a lot, is like is this just the
ups and downs of the macro environment or is this
a fundamental shift in how things are operating.
Speaker 2 (08:13):
So check out Business Desk on Friday, there'll be full
coverage off the Spark results. Just before we get into
our featured interview, an update on last week's story about
being AI Ben. You went in depth into this listed
tech company and the travails it's been experiencing. Some really
interesting discussion resulted from the podcast, including on LinkedIn what
(08:37):
were people saying.
Speaker 1 (08:38):
So on your posting of that episode. The CEO of
Callahan Innovation is Stephane Korn. He is an AI expert
from way back, and he posted talking and specifically about
their project Treehouse and this idea of APIs as a
way for AI agents to enter operate, which I I thought, well,
(09:01):
you know my perspective, I thought, it's a much more
efficient and effective way for machines to talk to each
other rather than open Ay's operator approach where it tries
to navigate websites that are designed for humans. Now, Stephane
Cortan had an interesting take on that where he basically said,
one of the reasons that APIs don't really work is
(09:23):
for competitive reasons, which was something I didn't really think about.
So what was your take on some of the reasons
that they might.
Speaker 2 (09:30):
Not Well, sitting behind the website, there's a database, and
for a travel company, it could be all their dynamic
pricing and all that sort of thing. So they've got
to be really careful that they don't give away this
sort of the secret source to their business model, so
they hold that information very closely to them. I don't
think it precludes them from issuing API access. For instance,
(09:53):
an Uber would really like the idea of being integrated
into a messaging app if you can pull in information
from Uber saying there are drivers available, it's only going
to be five minutes. You know that's possible via an API.
So I think there's some level that they could do.
But I think Stephann is right. The reason why we
(10:13):
just haven't seen open ai really go big on its
operator platform this is the AI agents it's now building
with an API sort of ecosystem for that, is that
a lot of companies are sitting back going do I
want to give you that level of access? So open
ai has just forged ahead and they're basically screenscraping all
of that information off websites. It means they can move fast,
(10:36):
and they're hoping that everyone will go, I need to
have operator because it's so good, my customers want it.
So let's do something a bit more substantial here with
with APIs. So I think that will come, but there's
definitely a competitive tension there in giving a lot more
fundamental access to these websites.
Speaker 1 (10:56):
Yeah, and I think one of the concerns might be
if you look at how deep Seak used distillation, for example,
to train its model off open ais model, there's actually
if you have AI running things really rapidly, and you
could do a whole bunch of API calls to travel company.
You could start to see how they're pricing actually works,
(11:17):
and kind of use AI agents that might not be
exactly what you might want to be accessing your API.
So yeah, you know there is a possibility there all
that could potentially be mitigated through careful you know, management
of the agents and what they're able to do and
things like that in the future. But that's why I
kind of said in the last episode as well. It
(11:38):
has potential, but there's a lot of things that need
to fall into place in order to make it work.
It's not impossible that the network effect might turn out
to make this API marketplace somewhere that companies need to
be in order to get their business out there into
the world. But on the other hand, yeah, there is
(11:59):
a lot of trains, a lot of changes that will
need to occur in order for it to find that
market for Yeah, so we've gone.
Speaker 2 (12:06):
From bigless incumbents like Spike to the earliest stage startups,
and we're focusing in this episode Ben on the idea
of a venture studio. I'd heard the concept before, really
in the Silicon Valley context, but it's the first I
really heard of a local venture studio. Was the Auckland
Company previously unavailable. It had some really big success with
(12:28):
the startup track Suit, which is like a brand tracking startup.
It raised around twenty two million dollars from venture capital
firms in early twenty twenty four in a series A
fundraising round that valued it at over one hundred and
fifty million dollars. But it started with a small amount
of seed funding and the expertise of the people at
(12:50):
previously unavailable.
Speaker 1 (12:52):
And then it turns out a number of local companies
are now pursuing this model, no surprise after a success
like that, among them a New and Improved Ventures, super
Bowld Ministry of Awesome, Paloma Ventures, Edition Group, and Rift,
who we are talking to on the podcast today.
Speaker 2 (13:10):
The three co founders are familiar with building products and
have a little cash to play with after they sold
a pretty successful IT consulting firm.
Speaker 1 (13:18):
That's right. So Chris Moore, Lucas Coelo and Ben morew
All ran Rome Digital together, which sold for around forty
five million in twenty twenty one, and after a bit
of a break, they decided to put their expertise and
money to use building and seeding companies of their own.
Speaker 2 (13:34):
And so Rift was born. Let's go to Ben's interview
with Rifts Lucas Coelo and Chris Moore talking about the
venture studio model, their company Rift and air approach to
building AI products.
Speaker 1 (13:52):
Hello, and welcome to the Business of Tech. Thank you
so much for joining us. We're here to talk about
Rift AI. That's your guy project that you've been working
on for a while. Now you're out here pushing out
this studio venture studio model in New Zealand, which is
a reasonably well known model now internationally and is starting
(14:13):
to pick up pace in New Zealand as well. So
why didn't you introduce yourselves to start with?
Speaker 4 (14:19):
Yeah? Cool, Lucas co founded a Rift Ventures before that,
Ravens in about eight years ago, and before that it
was heavily involved in the startup scene in Brazil, the
startup in design and technology.
Speaker 3 (14:31):
Seeing Chris so I was a co founder in Rome Digital,
which was a product design studio. Actually the missing third
party here is Ben who's also part of Rift. So
we've pulled together kind of the same team that we
had previously. Then we exited that in twenty twenty one
(14:53):
to a ten billion dollar US listed company, took a
bit of time off and took a bit of the
cash and then got the old team back together and
kicked off rift of the inches go.
Speaker 1 (15:05):
And that story we talked about, we've talked about before.
I've got a story up on business Desk that listeners
can read if they want a bit more background. How
does it work with a venture studio? Are you guys
developing the products? Are you investing in the products? What
does it actually mean?
Speaker 4 (15:21):
In summary, there are a couple of entities within sort
of the studio model. We have the studio itself that
produces the ideas. So we have a backlog of ideas.
Were always fishing for ideas, even within our immediate sort
of team. Our team now it's composed at about eleven people
with doing engineering, design, growth, marketing, and we always sort
(15:44):
of have a large backlog of ideas. Ideas are free
right that we keep adding ideas in there for us
to take a look at. And one thing that we
have is kind of like sort of like a playbook
on a few of the criterias of how we evaluate
ideas in order to turn them into opportunities, move them
up onto the pipeline. And that's when we start doing
(16:05):
a little bit of a desk research, understanding more of
the vertical that that idea plays in the industry, sort
of the product can look like, and then we start prototyping,
testing and see what sort of products or or companies
Mike can come out of that. So, for example, last
year our backlog of ideas had about one hundred or
(16:28):
something ideas. It's ideas that from the time of Rome
ideas that were sall online from other companies doing in overseas.
We also did some vertical sort of days or weeks
that we're going to speak with people within a certain vertical,
for example health tech to understand sort of their problems
and what things could we tackle, problems could we solve
(16:50):
there with with sort of the heavy sort of II mindset.
And then from quite a few of those we we
did a first sort of like a hot or no
quick rating on those ideas, thinking about what is the
competition locally? Can we build this in New Zealand? Is
there an entry way here that we can build this
fast in order to expand with the product later on?
(17:14):
The future sets are later on, especially because of a
little limited resource of of inentro studio. Do you want
to grab something that is too grandios or too expensive
to bute that requires millions and millions of dollars to
but so it needs to be something that we can
uh that can we can buy a big chunk off
with our with our limited capability at least at the beginning.
(17:37):
And then from that we tested it about. We prototype
like design and researched about ten of them a little
bit deeper, and that resulted into a few of the
ones that we have today. So after we take the
products through the through the validation process, and that might
be do we get customers, do we get some early
(17:59):
custom there's interest, what is the the is it feasible
to build? After we do that process, we take them.
We also have a fund inside Rift, so Riff the
studio and Rift the fund. So we have our first
fund called Rift zero. Now where did the first close
last year? And we take those opportunities into the fund
(18:24):
evaluate with a pretty brecorous sort of criteria, evaluate like
if someone pitching to us and if it passes or not.
We also invest in all those opportunities as well. That's
why we say that they graduated from the studio and
we spin them off as portfolio companies instead of being
(18:44):
just experiments. So it's it's both.
Speaker 1 (18:48):
Yeah, it's it's quite interesting because obviously the traditional startup
method is you raise a fund and then you go
and look for talent or startups to invest into. And
my fund might have a specific focus on a vertical
or a stage or whatever you like. So how is
(19:09):
that discussion with kind of LPs to say, not only
are you trusting us to do what's right with your money,
you're trusting us to develop the tech that's going to
bring you the return on investments. So you're starting, you're
pitching startups, and you're pitching a fund at the same time.
How does that conversation go.
Speaker 3 (19:28):
Yeah, I mean it's a more complex conversation than a
standard VC. You know, when a VC is raising funds,
they just say, look, we're going to invest in these areas.
We don't know what the companies will be, we believe
that we'll be able to find them as a studio
and as a fund. Rift zero also has an investment thesis.
It has four different areas that it's looking at. We
(19:50):
are mainly an AI driven fund, looking at how AI
is used in the companies as well as the products themselves.
So looking at how we can stretch the But those
four areas are across productivity, security, compliance, health and finance
as for getting people to sort of understand like does
(20:11):
Rift have that capability and skill. You know, we spent
the last ten years building companies and products for a
lot of startups and corporates. So we built companies for
X fifteen, which is CBA's accelerator program. We built them
for Westpac Ventures, we built them for ASB, B and Z,
(20:34):
the Warehouse Group, Mercury Energy, sky City. You know, we've
been doing this type of product building for ten years
as a group and before that. We also have that
personal experience of building companies in Silicon Valley and in
the UK, in South America. You know, for us, the
building and the technology and all those parts aren't really
(20:56):
the risk. Adventure Studio is there to de risk things
around repeatable process which we have from time at Rome
skilled entrepreneurs that we have within the studio as well
as within our wider network. You know, we would have
had about two hundred and fifty employees come through Rome
at the time when we sold it, we had one
(21:17):
hundred and fifty, So it's a large network across Australia
and New Zealand, Singapore as well as reach into the US.
So we have all the ingredients, all those assets that
you need to bring together, and then it's about taking
what can feel a bit boring, sometimes a formulaic approach,
and yet a lot of the times you can identify
(21:40):
great opportunities, assemble the team, assemble the financing, and then
go and tackle it. It's actually more similar to how
Silicon Valley startups are funded. Someone sees an idea or
an opportunity, they assemble the team, they assemble the capital,
and then they go and target and tackle the market.
In the past have seen a lot of New Zealand startups.
(22:02):
It's kind of they've been in the market for a while,
scraping along a little bit of friends and family funding,
and they just happen to hit an inflection point where
the market needs the product they have. This style is
quite different, I mean.
Speaker 1 (22:18):
And that's one of the questions as well that I
was interested in, which is, you know, you have the capability,
and you have the processes, and you have the experience
and the talent, but the X factor, the key thing
is finding the right place to apply those and that's
something that's talked about significantly in New Zealand. We have
limited resources in this country, we have limited capital, and
(22:42):
so we need to be quite specific about how we
are targeting our solutions. So how do you think through that?
Speaker 4 (22:52):
Yeah, I think it's the what I was saying before.
Like when we're looking at the vertical just for example,
a product that we have like chain, we start thinking
about the first idea, remember that was there on the
backlog was AI power job management software tool. There are
(23:15):
a few of them, a few of them are very successful.
Every once in a while they get acquired by larger
companies that which means that sometimes the customer service is
not what it used to be. And therefore, especially on
the SMBs, they start looking for all right, so the
prices went up. What else can I start using it?
It might be interesting to have something that, especially with
(23:37):
vertical AI businesses, you're always trying to look into, let's
look at automating things that they're not properly automated on
sort of the traditional job management tools, and let's start
let's focus on the product on that. But then while
we were taking that idea through the validation process and
we had a prototype of the tool, we had a
(23:58):
couple of sort of design screens and starts check with
so traders, with builders, with treadees, we're always trying to
think about an angle, that is, what is the what
is the line of this experience that will deliver them
instance of r O I that will solve an immediate
problem or or something that they want out to make today.
(24:20):
What can we build that will be fast to build,
that we can solve that problem. They will they will
pay for us to solve that problem, and then we
can build the rest of the capability afterwards, after the
proof that that piece of that piece of the of
the of the solution. And that's when, for example, we
start testing the in this case, the I think we
(24:42):
call a builder boost the the job management software lending page.
We have a couple of prototype screens, and after every
interview they will say, yeah, it's really cool, it looks
really cool, but not what my problem is? I hate
doing this, I doing the admin work. I hate all
of the tools that already exists. Yeah, this looks nice,
(25:04):
but I just want to I just can't pick up
my phone while am I work and I lose jobs
because of it. Can you solve someone? Do you know
anyone that can do that and then it qualify the
leads for us. And that's when Jane came about.
Speaker 2 (25:17):
Right.
Speaker 4 (25:17):
They were like, yeah, that's something that we can solve,
and we can solve reasonably fast.
Speaker 1 (25:22):
And Jane, is your l M powered super intelligent the
voicemail that can understand the context of what's being spoken
verbally through the phone.
Speaker 4 (25:33):
Yeah. Yeah, we're looking at now at the vision of
being this working towards being on the vision being this
Oh encompassy our encompassing sort of inbound lead qualification for
Assouli traders. Right, so not only qualifying leads that come
through the phone, it also can schedule sort of appointments
(25:54):
through the phone. It's on your Google calendar and in
your tool of choice and also integrating with WhatsApp. Business
noticed that it's a lot of small businesses that their
value proposition is helping some traders set up their sas
tools and we're like, okay, so this is a this
(26:17):
is really a gap in the market, Like people are
trying to solve this by being a consultancy to to
this is how that's how hard it is and how
uh that that it shows the friction that exists with
the things that exist today and how without without even
(26:38):
everyone that we spoke with they're like, yeah, we don't.
We want to spend less time doing that so I
can spend more time just doing the work getting paid.
Speaker 1 (26:49):
Interesting. Yeah, you're a New Zealand based business, you have
international ties, so I'm assuming and the question isn't will
New Zealand does pay for this, but it's will people
somewhere pay for this? And where is that? So that
must be another consideration as well.
Speaker 3 (27:05):
Yeah, you know, New Zealand market is far too small,
so if anything would have to be a minimum of
a regional market Australia and New Zealand, say Asia, so
apax but predominantly of course we're targeting the US, but
you have to start somewhere where it's easy to get
some initial feedback. But you don't want to stay with
that story of New Zealand being the trial market and
(27:28):
just two years later it's still your trial market. It's
really just used during the validation phase. Then once the
portfolio company is really to be launched, you're straight into
the US.
Speaker 4 (27:38):
Yeah. I find the New Zealand market, especially coming from overseas,
a very interesting market for tech in comparison to what
I'm used to in the US or in Brazil. I
think in the US, of course, depending on a little
bit of the state, but same as in Brazil. But
people tend to swarm more into new things, and in
(28:01):
New Zealand, I feel it's a little bit the majority
of the customers they wait someone to test it first
before they jump into it. So I tend to think
that a few of these things, especially on the products,
if you can sell it here to New Zealanders and
can convince a convince a very traditional sort of trade
(28:22):
to adopt that into its workflow, that's a good signal
that you can probably get some more adoption overseas.
Speaker 1 (28:29):
That's really interesting. Yeah, and I guess because you aren't
thinking of New Zealand as your primary market. You know,
I've heard stories before of startups who have had great
ideas that could have great ROI for trade's but they
just weren't able to convince the trades to actually adopt
it at a scale that was able to provide some
(28:51):
assurance to investors before unfortunately the companies collapsed. Not because
you guys have the venture model, because you aren't seeing
New Zealand as your prime market, but as a kind
of like trial market. That allows you to then go, well,
look now we have the set, we can throw it
somewhere that's a bit more swarmy and see and see
what happens there.
Speaker 4 (29:11):
Yeah.
Speaker 3 (29:12):
The venture studio model also allows you to play with
the different leavers of engineering marketing spend across all of
your portfolio companies. So if you have a traditional startup,
you hire an engineering team or you've got some co
founders as engineers, and that's great as you get a
bit of scale. Then every month you're paying for the
engineering team and you just got to keep feeding them
(29:34):
with work to be done. Now, some of the time
it's actually you don't need more features. What you need
to do is be selling what you have and get
the feedback from market. I mean, I think you know,
there's there's a lot of stats on I think it's
like forty percent of the features you build aren't even
the ones that customers adopt. And I've seen it myself
now with a venture studio, because you have a portfolio
(29:55):
of these, you can pull back some of the engineering
talent from one of the portfolio companies should into the
second one. While you're looking for the right market signal
waiting for market adoption. AI is one that you know,
in some cases is fast and other cases is slow.
It's so trying to get the timing right is I
believe easier with the venture studio where you can move
(30:17):
the burn around across the portfolio companies.
Speaker 1 (30:20):
This is something I wanted to talk to you guys
about as well, because we have in the I guess
really end of last year beginning of this year, we've
started to see this kind of counter rhetoric towards AI
and generative AI. And you know that this this idea
that open AI and it's ken who have been funneling
(30:46):
billions and billions of dollars into the development of the
technology are ultimately going to fail to see any actual
return on that investment just because of the scale of it.
And so you know, with the recent release of deep
Seek as well, that has simplified and made those models
maybe a little bit more affordable to actually run, that
(31:10):
kind of proves that there's going to be a difficulty
in getting a return on the investment in the long
run because they can't justify the expense when there's going
to be cheaper alternatives. What's your thoughts around that, and
how for companies like yourselves that are building on these
AI platforms that may represent a risk around the continuation
(31:32):
of the technology into the future.
Speaker 3 (31:34):
Yeah, if I sort of think back to two thousand
when you had you know, the Internet bubble and boom,
and you had this huge buildout of infrastructure and companies
like you know, Cisco, Juniper Networks, etc. We're riding high,
huge investments to get people more bandwidth. And back then
you had a ADSL link you know, one point five
(31:56):
megabit or something, and everyone was like, yeah, that's that's amazing.
I can now watch some standard DEF video with twenty
frames per second. What would I need more for? And
you know, you're fast forward now and you've got five
G networks where you can do video streaming, you know,
on the go in your car, you've got five to
(32:16):
your house with one hundred to three hundred or eight
hundred megabit down Like, people will find a way to
consume bandwidth. It's going to be the same with AI compute.
Like you know, deep seat coming out as great as
open source, it's already been ripped apart in bits of
being put into existing models, ignoring the supposed cost because
(32:39):
there's a lot of issues around. Well, actually they did
have access to fifty thousand in video GPUs and they
did spend more money than the kind of the sticker price.
You know, that's that's great. You need to bring the
costs down. We're already seeing that with open Ai. Like
they're the three mini models like ninety three percent cheaper
(33:00):
than their at last one. It's faster, less latency, which
means you can do more of these real time voice,
real time communications, real time robotics. It's does really feel
like we're at the beginning of this expansion. We do
see it with some of the AI startups, you know,
we you know, we we review a lot to see
(33:22):
where they are, what's the state of the market, you know,
what's real, what isn't And a lot of times you'll
click through and it will be coming soon, you know,
sign up for a better access. But we also know
because we're using it and building products that are in
market that customers are using, and it's today solving a
problem that it does work for you know a fair
(33:44):
number of use cases that are still there to be
to be tackled. And yeah, it's the worst it's going
to be at the moment and it just keeps getting better.
Runway Sourer for image generation mid you know, for video
generation mid journey, like the disruption is massive.
Speaker 1 (34:03):
Mm.
Speaker 4 (34:04):
I still think there's a lot of case that the
use cases that we at you they're the low hanging
through use cases for l lambs that people are finding
out and then now we're figuring out that, oh wait
a second, we might not need other lambs for everything.
We can use a couple of the smaller models there
will be much more useful and choose some specific verticals
(34:25):
or for some specific sort of tasks or even a
sort of agenta task and that's much eeaper to run.
So mind you start using that and also thinking AI
researcher open ai pulls it on on Twitter x formerly
Twitter about that more money now is being spent on
(34:48):
inference then on pre training, So more money even spending
on what can this do and can we learn things
that there's no data for it before, which will become
more useful. So even the CAPEC spend is going silver structure, yeah,
and the infrastructure is going a lot. They're spending a
lot of money on building the infrastructure for that. I
(35:11):
do think that if the market is expecting an immediate
return for this investment. That might be especially for this
companies they are spending a lot.
Speaker 1 (35:23):
You mentioned kind of in video, and you mentioned Cisco,
and one of the observations was with Cisco. Obviously in
the early two thousands they had that massive wipeout of
a lot of their value and there was a similar blip,
but in video probably was it was not obviously of
the same scale. But the argument was that these companies
(35:45):
that are being continually inflated, the value is being inflated
by the the the pouring of billion dollars into R
and D by these companies, that we will start to
see a similar popping off that value bubble, not necessarily
around the quality of the tech, but the value bubble
(36:05):
in itself. Does that concern you guys as having both
the products and the fund.
Speaker 3 (36:12):
I think on the end Video one, what's quite interesting
is you know that they've managed to ride three waves.
They've managed to ride the GPU wave, when you know
graphics move from being calculated on CPU offload. Then they've
managed to ride the cryptomning wave where it moved from
CPU to GPU and then GPU to ASK. And now
(36:32):
they've managed to ride this third wave with AI, you know, phenomenal,
But there's spent a lot of time making sure that
there's software. It's it's not just the chip, it's the
software that you use around it. Kuda, which is the
language they use for programming those GPUs. But you are
seeing you know, Amazon have their own processor. I think
(36:54):
it's called Trainium. You know, like Google have got their
tensor processor unit, the TPU. People are coming for that crown.
Of course, it's so rich. The good news is, you know, yes,
Cisco wore that bubble and that bubble burst, but when
you look at the major players, you know, Google and
(37:15):
Facebook and all the social media Instagram, they all came
out after that collapse, after the infrastructure burst. Also, there
was a lot of costs that the US government was
selling Spectrum at the time to the telpots, so that
sucked a lot of their cap becks up as well. So, yeah,
if we were in the chip space, if we're in
(37:37):
the model space, I think that's going to be cut throat.
You're going to end up with open source, and you're
going to end up with Microsoft, which is open AI.
You're going to end up with Amazon, which is anthropic,
and you're going to end up with Google, with with
Gemini and you know, and you know under open source
of course you got Lama with Meta and all these others.
Like I wouldn't want to be in that space, application space,
(37:59):
you know, being able to look at what tasks administration
work you can replace. Yeah, that's that's the place to play.
Speaker 1 (38:06):
Yeah, because you're not concerned about being caught in the
inner potential bubble burth.
Speaker 3 (38:14):
If your companies are making money and we're not building
companies that are growth at all costs its balance of
growth and profitability as a venture studio, that's okay.
Speaker 4 (38:26):
Yeah, because we're we're consuming more the commodity that they
provide in terms of the models than being caught up
on developing the models ourselves. And depending on who wins
this race, it's an open source with the excellent models
that that exist, and we're building with a with a
(38:47):
paradigm that is, we're building our capabilities in our future
is based on on the available sort of capabilities of
the of this commodity that exists and where it might
be in a fewure based on the trend of development
that this exists. But we can consume from any player. Yeah,
that it's out there, so we can consume the open
(39:08):
Eye one the Anthropic one, because they're probably going to
be very converged. They are converge to be very similar,
at least that it looks today. I'm not sure tomorrow,
Open a I might release something that we look and say,
oh my god, this is ten years ahead of everyone else,
as it was with the Chat GPT, although some might
(39:29):
argue that other companies had what the first GPT was,
they just didn't want to release to the public, so
Opening I was not that ahead as others might think.
But they seem to be very ethnical.
Speaker 3 (39:41):
Nick and Nick.
Speaker 4 (39:42):
It just continuously and there are some that are For example,
even in a few other products, we don't use one
model we use. We might use multiple models they're better,
or smaller models they are better in each task. There's
Claude is excellent for coding for for some reason, and
I'm not sure what they added into the the find
(40:04):
the tuning or on the model, but it's I find
it much better for coding than GPT. So you usually
want I'm use in cursor, I want to look into
coding or use Claude instead of using Chat GIPT. I
think it's just better, especially with Python. I might see
that other companies that are consuming a few of these
(40:26):
models my pick and choice and do the same because
I think the secret is still on. The secret selves
of companies still be the services they provide. They are
solving a problem. How is the experienced layer, how is
that sort of application layer more than the commodity itself?
Chatting with someone that is like, as long as there's
(40:48):
coffee to be bought, I want to keep building Starbucks
shops or nice coffee shops. I don't want to build
a farm to start producing coffee because that would that
would probably be a RaSE to zero or very very soon,
if not that already started. And as you're saying, with
a deep seek, wiping out quite a few, quite a
(41:09):
bit of the value of these companies.
Speaker 1 (41:11):
Interesting, I do enjoy that you said Starbucks or nice
coffee shops. I think that's that's very to be honest.
Speaker 4 (41:19):
Yeah, I don't mind star Yeah.
Speaker 1 (41:24):
That's great. I mean, I know we're probably over is that,
But I just one more question I want to ask
around that idea as well. Was there's also a commentary
that we might be hitting a ceiling in terms of
capability around L l ms, talking in particular around hallucinations.
You know, it's a feature, it's not something we can
(41:45):
get rid of talking about the cognitive in quotes capabilities
of these lms that the rate of increase is now
so minimal that actually it looks like we might have
hit the ceiling on that, and so building with the
idea that there's going to be something better in the
(42:05):
future may be a mistake, if that makes sense. What's
your take on that perspective.
Speaker 3 (42:12):
I guess one thing is the companies we're building at
the moment take advantage of the current state of the art.
We're very cautious about building for the crossover point of
some capability that might come in eighteen twenty four months.
So I think, yeah, if you're building companies, you need
to make sure you're building with what can be achieved today.
(42:34):
That's why we see a lot that are hidden behind
sign ups and join the wait lists. They were also
waiting for a tech shift. So that's pretty critical. And
it's not to say that the evolution of the product
cannot look to future capabilities, but your sellable kind of
wedge in today should be able, well not should, it
has to actually work. That's one of the things as
(42:57):
far as have we hit the limits, I think what's
interesting is, you know you have something like deep Seat
come out and it's got a whole bunch of these
kind of clever optimizations you know in ways of doing
things like people are going to continually find these. So
if you can use it to train your model at
a tenth the price, well, now you can train a
(43:20):
model that's ten times as large. Can you find more data?
Like I said, that's one of the big sources. So
then you start looking at synthetic data careful of hallucinations.
There's a lot of smart people, there's a lot of money.
I just don't see us hitting a wall in the
next next few years. The same as bandwidth. Again it's
(43:40):
a weird analogy, but remember thinking, you know, one point
five megabit was so fast because you had a twenty
eight K modem before that, and you know you used
to get like a say in New Zealand, a T
one or it's the US T one or an E
one line two megabit, and it would cost you ten
thousand dollars a month to connect Auckland to Wellington. It
take thirty five calls, and that just seems so quaint,
(44:04):
but at the time it would have been well, that's
the fastest that we can get it across the network,
and all of these other reasons where there's demand, you know,
money and innovation follows. I don't think we've been hitting
limits soon.
Speaker 4 (44:17):
And then's the bit about we're talked about building with
the future capability in mind, with CRUs Is saying as well,
is that there's a lot of the not a lot,
but they are parts of the workflows of a few
of the products that we're working with that they're just
ultimately like that llms are not capable of doing right now.
And then you what it do is that you do
(44:38):
that part of workflow with your standard automation sort of
standard code there. But it generates is that it starts
when people start using it generates a very specific sort
of the amount of data that you can use a
small model to be very good at that very specific
sort of workflow, which is something that starts building different
(45:00):
right It'll be very hard to build differentiation if you're
just wrapping your company around whatever is the large the
data set the lms are trained on, because then anyone
that just pluts into the data set will be able
to to roughly perform the same task. But when you
viewed something that solves a very specific sort of touch
(45:21):
point and you start getting your model into it and
then using that data to train your capability, I think
then you start creating differentiation and something that it's more defensible,
especially in the in a world where anyone can go
to cursor or go to what's the zero from A
(45:43):
and build a SaaS company in a in a week
or so, it would be very interesting to see how
it goes.
Speaker 1 (45:52):
It will be it will be very interesting.
Speaker 4 (45:54):
Yeah, very interesting. In the same moment that said that
the first billion dollar company of one person should should
should arrive very soon.
Speaker 1 (46:10):
Some helpments ad a lot of things.
Speaker 2 (46:19):
We'll put a couple of articles in the show notes,
some really interesting articles, including from New Zealand companies sort
of explaining the pros, mainly the pros of the venture
studio model. You know, in my mind, really I think
the big advantage here and I think tracks who really
illustrates this. You have, you know, an advertising guru like
(46:42):
James Herman previously unavailable, Simon Pound, people like that who
really understand marketing, advertising branding. They come across some entrepreneurs
who want to set up something in that space, and
they have deep integral knowledge of that and and what
works and what doesn't. So you invite them in, maybe
(47:03):
give them a little bit of seed funding, but mainly
take equity and return for nurturing this company, giving it
your expertise. That seems to be the venture studio model,
and it then allows a company to get to a
point where it's really attractive to venture capital companies. And
I think that's a really good model for New Zealand,
(47:25):
where we do have these areas of deep expertise, but
we don't necessarily have a lot of seed funding. So
you're seeing these people who maybe have created successful companies
exited them, have a little bit of capital and are
willing to back these teams and do numerous teams at once,
so not just put everything into one company. You have
(47:47):
half a dozen of these companies under your roof, maybe
even hot desking or sharing space, looking over the shoulder,
helping them, mentor them, introduce them to your networks and
business context. It seems to be quite a good model
for New Zealand.
Speaker 4 (48:03):
Yeah.
Speaker 1 (48:04):
Absolutely, And also it does help to slightly address the
talent issue as well. Because if you have it, like
they say, they have a group of engineers within Rift
that they can then choose how to where to kind
of direct their talent. So two different different companies say
they've got one thing over here, and they're like, oh,
we really need a new feature on that asap. Let's
(48:26):
move sixty percent of our engineering over to that. For
the meantime, you don't have to suddenly scale up and
get a bunch of contractors and figure, you know, get
them to get on board with the systems and learn
new things. It's just this kind of this core at
rift that can do things as they need them to do.
So another really great way of addressing some of New
(48:47):
Zealand's shortfalls.
Speaker 2 (48:48):
Yeah, I guess looking at the potential cons of this model.
So clearly, you know, the people who are running these
venture studios are spreading their risk but spreading their attention
across a number of ventures. So you know, you may
be diluting the attention that you can pay to a
(49:10):
particular company. And I've seen a couple of podcasts where
people who have been through venture studios and sort of say, well,
it's you know, sort of feel like I'm competing with
other companies in this venture studio for the attention of
the people who are funding this and supporting us, So
I guess that's the downside. But having said that, we're
pretty collegial in New Zealand, so you're more likely to
(49:33):
actually be able to leverage off what other companies sort
of in the venture studio stable are actually doing, to
try and learn from them and maybe even partner on
business ventures.
Speaker 1 (49:44):
Yeah, exactly. And what I think is interesting about ref
as well as that particular focus on artificial intelligence because
it is such an evolving sector and it kind of
does link into that idea of talent because the leadership
team they spend a lot of time I'm looking at
different research papers about AI, about generative AI and its applications,
(50:06):
and then they can disseminate that information to their ideas
for startups, rather than you know, having to kind of
each individual company follow along as best they can and
split that time. So it's another benefit as well. But
it's also interesting they're focused on AI, especially in this
kind of very AI heavy new world that we're in
(50:29):
the conversation about whether it's a bubble. It seems to
me from a conversation with them with the founders that
they are quite focused in their approach. They're wanting to
kind of really really focused on what products are actually
finding product market fit and value and go from there,
(50:50):
rather than trying to kind of have a product and
get a bunch of hype around it and pump a
bunch of VC in it and then see where you
go from there. Maybe that's a little I don't know.
Speaker 2 (51:01):
Yeah, they're clearly using AI and it was really an
interesting that discussion about their reflections on Deep Seek and
that move from sort of training to inference, which is
sort of changing the priorities in the market at the moment,
and also their view of the AI landscape. It seems
as though they're expecting, maybe a bit belatedly, a wave
(51:23):
of AI startups to sort of emerriage. I mean, they
are emerging, but it hasn't been as pronounced as say
in the Australian market, so it's good to hear that
that's coming. But you know what they're doing. I think
with the couple of ventures that have got traction so far,
it sort of harkens back to what we were talking
to Rowan Simpson about last week, is not necessarily being
(51:44):
first but trying to be the best, you know, so
the startup they were talking about that basically makes it
easy to take the admin out of being a small business,
you know, for tradees and that sort of thing. Sure
we've got if I have done that others we seem
to have a timely to another one. I guess we
seem to have some expertise in this area in New Zealand,
(52:08):
but they're taking a slightly different approach to it, applying
AI to it where appropriate. That's not necessarily a revolutionary idea,
but it's all about executing that well absolutely.
Speaker 1 (52:19):
And they also finished raising their first fund as well,
so as well as using internal money, they are raising
a fund that they can feed money into their own
studios as well. So that obviously shows some confidence in
the market in what Rift is doing.
Speaker 2 (52:37):
I think they've put five million into the zero Fund
Zero Fund, and that seems to be quite a bit
for a venture studio. Although I was looking at New
and Improved Ventures, which really the same people behind that
were previously unavailable. They've sort of set up a new
venture studio I think with the likes of ice House Ventures.
They've raised six million dollars to back three for new firms.
(53:01):
That's what The Herald reported late last year. So there's
definitely some good seed money starting to flow into these
and the reason is, I think statistics I've seen is
that they're actually more successful. Your chance of getting to
the next stage is greater out of these venture studios,
which have actually been around since the mid nineties in
(53:22):
Silicon Valley. So they've nailed that sort of process, and
according to some sources that I've seen online as well,
that they allow you to get there quicker. So it's
more rapid way to go from ideation to execution, which
obviously that's what is all about. Has been quick to market.
Speaker 1 (53:44):
Yeah, and they're not necessarily looking for unicorns as well, Like, yes,
there is some sense of it would be nice to
have one, of course, but at the end of the day,
a flow of one hundred to two hundred million dollar
companies that they can sell and then reinvat, or they
can split off and can go and have lives of
their own. That's also something that you actually do need.
(54:06):
We've talked about it before, that middle layer of companies
that are continuing to grow, be sold or change or
whatever else it is, and feeding talent and money back
into the market as well. It's not all about the unicorns.
Speaker 2 (54:19):
Sounds good power to them.
Speaker 1 (54:22):
So a big thanks to Lucas Coelo and Chris Moore
for coming on, and Ben Morrow as well for helping
set that up.
Speaker 2 (54:29):
More on New Zealand's growing band of venture studios in
the show notes. Head to the podcast section at Business
Desk dot co dot d.
Speaker 1 (54:36):
You can stream the podcast there or on iHeartRadio, and
of course it's available on your podcast platform of choice.
Speaker 2 (54:43):
Get in touch with your feedback and topic suggestions. We're
on LinkedIn and blue Sky these days.
Speaker 1 (54:48):
Catch us again for the next episode next Thursday.
Speaker 2 (54:51):
See you then,
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com