Audi is the car brand being mentioned as part of Cruise’s early self-driving car testing. The hosts are talking about which cars Cruise used at the beginning.
This is the behind-the-scenes software that helps engineers build and ship products. It’s the toolkit that makes development easier and more organized.
This means making self-driving systems work for lots of vehicles or robots, not just one test car. It’s about building the tools needed to roll it out widely.
Self-driving means a vehicle can drive itself without a person controlling every move. The speaker says the same tools could help with other robots too.
Concept
AI
AI is computer software that can make decisions or recognize patterns. The speaker is saying it has gotten much better in recent years.
Cruise is a company that tried to build self-driving cars you could ride in without a human driver. The speaker is talking about how big it got and how things later changed.
AV means autonomous vehicle, or a car that can drive itself. The speaker is talking about the whole wave of companies trying to make self-driving cars work.
These are self-driving vehicles. They use cameras, radar, and software to figure out where to go and how to avoid crashes without a person doing all the driving.
This means something has to work correctly because if it fails, people could get hurt. In a self-driving car, the computer has to make very careful decisions all the time.
This means a car can drive itself instead of needing a person to do everything. The speaker is saying this was one of the first big places where AI really worked in the real world.
These are companies trying to make vehicles drive themselves. The discussion is about a newer group of companies that are using AI from the beginning instead of adding it later.
It means the company was built with AI at the center from day one. The speaker is saying these newer self-driving companies are designed that way instead of trying to retrofit AI later.
This means machines that can do tasks on their own in the real world, not just in software. The speaker is talking about self-driving tech spreading to other kinds of machines too.
This is a company making machines that can work by themselves on construction sites. The speaker is using it as an example of self-driving technology outside cars.
These are construction machines that can do work on their own or with very little help from a person. The speaker is saying self-driving technology is spreading into construction equipment too.
These are robots regular people might buy or use at home. The speaker is saying self-driving-style technology is also showing up in robots for consumers.
This means military or security systems that can do some tasks by themselves instead of needing a person to control every move. It can include drones and other unmanned machines.
This means the system keeps getting better because it learns from more and more real-world examples. The more it drives, the more it can improve itself.
This means the unusual stuff that doesn’t happen every day, like odd road situations or rare mistakes. Self-driving systems need to learn from those rare cases too.
This is like a computer-made practice world where the car’s software can be tested safely. It helps engineers see how the system would react before trying it in real life.
Safety systems are the features that help catch problems early. The speaker means the robot already has tools that notice when something might be wrong.
Data mining means digging through a lot of information to find useful patterns. The speaker is saying they want to search past events to find similar ones.
This means digging through lots of information to find the useful bits. Here, it’s about finding unusual driving moments that can help teach the software.
This means looking through a lot of data to find the exact examples you need. Instead of inventing something new, you’re trying to find the right past cases.
This means the behind-the-scenes setup for running lots of vehicles at once. It includes things like keeping track of cars, sending them where they need to go, and managing them efficiently.
This means starting with the basics and figuring things out from scratch. Instead of following a recipe, you build the solution by understanding the core problem first.
Company
Zoo
This is probably Zoox, a self-driving car company, though the transcript sounds a little off. The speaker is naming companies that work on autonomous vehicles.
This is a software tool that shows different kinds of information together in one place. For self-driving cars, it helps people look at camera views, sensor data, and driving decisions at the same time.
This means how hard it is for a new company to get started in a business. The speaker is saying self-driving cars are too expensive if only giant companies can afford to do it.
It’s basically a special kind of file that saves a record of what a system did. People use it later to figure out problems or study how something worked.
This means a company made its own special way to save data instead of using a common standard. That can make it harder for other people or software to read and use it.
They mean a single file type that lots of different robot tools can open, like how PDFs work for documents. It's a way to make robot data easier to share and read.
MCAP is a file format for robot data. It helps different robot systems save and share information in the same way, instead of everyone using a different format.
SaaS is the kind of software you use online and usually pay for regularly. They're comparing the robot business to that software business to make a point about standardization.
Company
Fox self
This is the name of the company or project the speaker is talking about. They’re saying it helps build the tools underneath robotics products.
This means the physical parts may become more common and less special, so they cost less and are easier to buy. The speaker is saying the important part may end up being the software and system around the hardware, not the hardware alone.
This is just the way computer data gets saved so it can be read later. The speaker is saying that building that kind of plumbing is less exciting than solving the actual customer problem.
This is the part of a tech trend where people stop being overly excited and a lot of companies give up or fail. After that, the useful ideas and stronger businesses are the ones left standing.
These are AI programs that can look at pictures and understand words, and the public can use and modify them. People are teaching them to help robots decide what to do.
Aurora is another self-driving company. The speaker is saying people from companies like Aurora are now working on all kinds of machines, not just cars.
They’re talking about how hard it is to grow a self-driving taxi service. It’s not just about the cars driving themselves, but also making, charging, cleaning, and moving them around.
This is about the business side of running a fleet, not just the driving part. It means making, moving, charging, and cleaning the cars so the service actually works in real life.
This means the hard part is making the car drive on its own without mistakes. They’re saying that part may be less of the problem than actually running the service day to day.
These are the oddball situations that don’t happen very often. Self-driving systems can be great most of the time and still struggle with these rare cases.
Validation is the process of making sure a self-driving system really works and is safe. It means testing lots of situations before letting it operate on its own.
This is the part of the conversation where they shift to other industries besides cars and trucks. They’re talking about defense and aircraft-related businesses.
This is a big AI system that can be reused for lots of robot jobs. Instead of building a separate brain for every task, engineers try to make one flexible model that can handle many of them.
This is about robots being able to grab and move things carefully, almost like a human hand. It’s hard because objects are all different and the robot has to be precise.
They’re talking about robots that help in warehouses by moving boxes and packages around. These machines make shipping and storage faster and more organized.
They’re talking about robots that go through grocery stores and check what’s on the shelves. These robots help stores see what’s missing, what’s priced wrong, and what needs restocking.
They’re talking about machines that move luggage around airports by themselves. It’s a smaller, easier version of self-driving than what you’d need on city streets.
San Francisco is a city with busy streets and tricky driving conditions. The speaker is using it as an example of a place where self-driving cars have to handle a lot of challenges.
A moat is something that gives a company a strong edge over competitors. It could be special technology, unique data, or customer relationships that are hard for others to copy.
This means the company has a special edge that other companies don’t have. It might be a tool, data, or method that makes them better or harder to compete with.
Stripe is a company that helps businesses take payments online. The speaker is comparing their software to a tool that companies rely on every day and don’t want to switch away from.
Datadog is a company that helps businesses watch their computer systems and spot problems. The speaker is using it as an example of software that people keep using because it becomes part of their daily operations.
Company
Cruz
This seems to be the name of a company the speaker used to work for. They’re saying that even after it closed, people kept using the tools he had built there.
ADAS means car features that help the driver. Things like lane assist and automatic braking are part of it, and the speakers are talking about that kind of work at GM.
This is probably talking about cars that drive people around by themselves, like a taxi with no driver. The transcript seems to have misheard the word a little.
This means one company tries to make most of the important parts itself. In cars, that can mean the company designs the car, the software, and the self-driving system instead of buying those pieces from others.
OEM means the car company that actually builds the vehicle. An OEM licensing play is when a tech company sells its software to car makers instead of selling cars itself.
This means the software was built with AI at the center from the beginning. Instead of adding smart features later, the whole system is designed to use AI.
It means the car can drive itself for a while, so you don’t have to keep staring at the road the whole time. But the driver may still need to take over if the car asks.
Level 3 means the car can handle driving itself in some situations, but you still need to be ready to jump back in if it asks. It’s more advanced than cruise control or lane keeping.
It means a car company might work with another company instead of trying to build everything itself. They team up to make the technology faster or better.
It means making money by selling the tools other people need, not by selling the finished thing itself. In this case, the speaker is talking about companies that help other companies build software.
This means letting AI help build software for you, instead of writing every line yourself. The person is saying it’s becoming easy for non-programmers to make their own apps.
Databricks is a company that helps businesses work with large amounts of data. The speaker is saying companies like this are booming because AI makes data tools more important.
Supabase is a service that helps developers store data and build app backends. It’s mentioned here as one of the companies getting more popular because AI is increasing demand for software infrastructure.
Vercel is a company that helps developers build and run websites and apps. The speaker is saying companies like this are growing quickly because AI is driving more demand for software tools.
This means the business of making robots and the tools that help them work. The speaker is saying a lot more robot companies are going to start up soon.
This means a small business or individual can pay for the service with a credit card instead of going through a big sales process. It’s usually the easy, online checkout option.
AV means self-driving vehicles. The speaker is talking about the world of cars that can drive themselves, and how people in that field have tried different approaches.
This means AI that does things in the real world, like controlling robots or machines, not just answering questions on a screen. The speaker is saying that kind of AI is getting a lot of attention and money.
These are very large AI systems that can be reused for lots of different jobs. The speaker is saying they’re becoming important in robots too, not just in chatbots.
Mind Robotics is a company working on robots for factories and manufacturing. In this conversation, they’re saying it’s a new business that will first try its system with Rivian.
you name it, they're all out there building autonomy
in all of these different areas.
So one last follow up in spite of all this,
what is really striking to me
and this is on the technological side,
you still have a company like Waymo specifically I'll mention
has put so much money into this,
has is considered a leader
and it is literally still struggling
with the school bus problem in Austin.
And I'm wondering what your perspective at like what,
is this something that is going to be an issue
in perpetuity, not specifically school buses,
but like you would think that this would be solvable
at this point and yet it's going to not be.
Yeah, I mean, I think Waymo at this point
is more of an operational problem
than an autonomy problem, right?
Like there are always going to be these edge cases
that they have to work through.
But you know, the way I like to think about it
is that even if you could wave a magic wand
and say Waymo has perfectly infallible autonomy today
and then you're like, great,
go roll that out to like a hundred cities in the US.
You've still got this enormous challenge of,
got to manufacture the vehicles,
you got to get them to the cities,
you've got to run operation,
you've got to charge them,
you've got to clean them,
you've got to, there is just like a massive operate
and that's why you're starting to see,
companies like Waymo are like partner
with people to manage operations.
That alone, you know,
there's tens of millions of vehicles in the US
and like they're not all going to get replaced
by robotaxis overnight.
And so that alone is like a huge challenge.
But I think, you know,
they're getting there and they're working through that
and that's just really a problem
of sort of capital and operations at this point.
And it's more of an inevitability.
When you go to other industries,
there are many problems like self driving tractor,
for example, where it's driving
up and down an empty field.
And like, you don't have to deal with like
weird school bus issues.
There are for sure,
like there's still safety and validation cases
and you need to be like,
what if a child is lying down in front of the tractor
in the field or something?
Like, yeah, you do need to worry about these cases
and make sure you're going through the validation
but it's not anywhere near the level of complexity
of autonomy that you need for like driving
around San Francisco and New York City
or something like this.
So there are many applications of physical AI
that are way easier to solve
than driving and that are being solved today.
So you've raised a series A and a series B, is that correct?
Correct, yeah, we've raised about 60 million in total now.
So I'm looking at your client list
and it's some really cool companies
but none of them, I see only one is a self driving company
or wait, sorry, there's Wave, Wabi and Gatic, so three.
So where's the biggest area of growth
for your platform product?
Yeah, we have more by the way,
like not every company we're allowed to list
on the website, much to my dismay.
So there are more in the self driving
and especially self driving trucking, we see a lot.
There's a lot of companies in that space too.
But yeah, the biggest one, the biggest growth area
is I would say AV is a big one.
Defense and aerospace broadly
is just like a huge amount of investment happening there
and so you see a lot of, you know,
everything from drones to boats to ground vehicles
offered a lot of work happening in that space.
And then we see a lot of work, you know,
I'd say broadly in sort of humanized
and robot foundation models and dexterous manipulation,
there's a lot of people making this push into
how can we really use AI models to be able to like
pick up and grasp and manipulate complex objects
and work through these kind of long horizon tasks,
that's probably the other like biggest area for growth.
And then you just see this long tail
of productive companies doing super interesting stuff,
right?
I mean, there was like, you know,
there are ones that you think about,
like obviously in the warehouse,
you think about a lot of robots
that are just moving packages around warehouses
and things like that.
You know, you've got ones like Symbi Robotics
where they're driving up and down aisles
in a grocery store and looking at stock
and making sure that like everything's in the right place
and does it have the right price attached
and are we out of stock or anything?
And they can do that, you know, daily instead of monthly.
There are companies that you don't even think about.
Like I just talked to one recently, Aeravik
and they're doing, you know,
the luggage things that come up
beside your airplane while you're boarding the plane
and they've got to drive your luggage
from like hundreds of these vehicles.
I think they said that like, you know,
a large airport would have like a thousand plus vehicles
that are driving luggage from the airport
to the plane every single day.
And so like just that is like a huge autonomy problem
that is nowhere near again.
You know, there are a lot of these challenges
that are very solvable or a lot easier
than driving around, you know, city,
but they still look very similar to self-driving.
And that luggage, airport
and luggage autonomy movement issue,
is that going to get solved?
Yeah, how close should we go?
Oh yeah, yeah, these guys are doing it.
Yeah, I mean, yeah, it's just such an endless long tail
of things like that, right?
Where you don't even think about them
and then you look and it's like,
oh yeah, there's a person driving that
and you know, and why did we build self-driving cars
that can drive around San Francisco
when we haven't solved the luggage thing
coming around the airport, right?
This is all the tractor in the field.
So what is the moat for a company like yours?
And by the, I don't invest in software
so I would pretend to know.
Is it velocity?
Is it contract duration?
Is there proprietary advantage
baked into your platform?
Like what is?
Yeah, I mean, the advantage that we have
is a very strong team of people
who a lot of our team have come from the A.V. industry.
So we've seen this at scale
and a lot of the companies that we're selling to
either they've also come from the A.V. industry
and they know the tools that they had in that space
but they don't want to like go rebuild them.
They know that they spent years and years
kind of building the life from scratch.
Like why do I have to do this again?
So like our biggest advantage
is just really deeply understanding these workflows
better than anyone, building a really good platform.
And infrastructure products tend to be really sticky.
Like again, I draw the analogy to the SAS world.
In SAS, you don't, you know,
you use off the shelf databases,
you use off the shelf hosting,
you use platforms like Stripe to process your payments,
you use platforms like Datadog to monitor
your infrastructure and, you know,
you get on a platform and it's good
as long as it continues to be good,
you keep using it.
It's kind of a lot of work to shift these things.
So, you know, that's how we're like,
we're here to build a really good product
to help people move faster.
And at the end of the day, you know,
they like us, they keep using us.
So when you left Cruz,
they kept using the tooling that you were building there.
When Cruz shut down,
did you acquire any of the assets
pertaining to infrastructure you'd built?
No, assets.
We did grab quite a number of people out of Cruz
after they sort of shut down,
which by the way, I mean, GM still has like,
most of, you know, everyone that didn't sort of
explicitly leave on the engineering team
is still at GM working on ADAS over there.
So like, they still have a pretty sizable team,
but they shut down Cruz as a company
and they sort of managed to lose quite a bunch of talent
just through the way they kind of managed that transition.
But you know, that was a benefit for us.
We hired quite a few people out of Cruz.
This story seems familiar to me.
Yeah, what do you think is going to happen with,
you know, like you mentioned,
there is some talent still over at GM
and working for a company like Cruz is very different
than working for a legacy automaker like GM.
And, you know, what is your prediction, I guess,
of what GM is going to do with all that talent?
Yeah, yeah, they have, I mean,
that's the opportunity for them, right?
Like can they demonstrate that they're still hiring
and they've got good people there
and they're hiring good people?
That's a very tough market for, you know,
it's a, you know, a talent marketplace out there right now.
And a lot of the good AI talent
is getting scooped up by every single lab.
So I think it's definitely a challenge to keep hiring,
but they seem to keep hiring good people over there.
So...
Well, hiring isn't always the result in the thing.
It's just really interesting to me to see,
you see a company like Rivian, for instance, that's like...
Right.
Yeah, not only are we gonna do these joint ventures
with like on the software side of things,
we're gonna go all in on autonomy
and we're gonna deploy Robotoxys.
Like it's fascinating to me that automakers
are still kind of going after that.
And then you see companies like GM and Ford,
which seem to be more like, okay,
we're gonna ease into this, we're not gonna do the full.
Wondering what your view is, is the right approach
without like calling out a specific company?
Yeah, yeah, yeah, yeah, no, yeah, exactly.
I think like without sort of saying like,
is GM specifically going to win at that or not?
I would say across the industry,
we're still seeing a lot of OEMs
that are trying to go at them themselves,
trying to go it alone, right?
And some of them will be successful,
almost certainly Tesla will win at that
vertically integrated approach,
possibly others like Rivian will,
possibly GM will, that's the opportunity.
But I think that we're probably still
in a little bit of an unsustainable number of people,
like building an autonomy stack is incredibly difficult.
I don't think that there will be that many autonomy stacks
for driving 10 years from now than we have today.
I think that there's still consolidation to happen.
So there will be some very successful suppliers
who are working with OEMs.
There will be some OEMs
that have built a vertically integrated
and just paying the enormous amount
that it costs to keep building it.
But I think there's still gonna be
some more consolidation in that space.
And that's why, you know.
Yeah, I mean, you see companies like Wave
and even Nero now who are like,
that are able to, their pitch is,
hey, we'll license this
so you don't have to build the autonomy stack.
And then for a company like Rivian,
it'll be interesting because we've seen
that it was very lucrative for them on the software side.
I mean, they have,
are getting $5.8 billion from VW.
And that doesn't include AI or autonomy.
So-
Yeah, it doesn't include autonomy.
If they're successful in that,
you could see them also becoming a license
or if like big, big quotation marks right now.
It'll be interesting to see what happens.
I think what you'll see play out over the next few years
is as like, I'm incredibly bullish on Wave, right?
They're in a very strong position
going after the OEM licensing play.
They have a really strong AI team
and they have AI-first architecture.
So, as we see some of those platforms
reach a point where they can actually deliver,
like right now everyone sort of
is in a little bit of a blind spot
because everyone's building their own thing
but very few OEMs have actually shipped
like an eyes off like level three product basically, right?
Like no one has really,
even Tesla is still not eyes off.
So once we start seeing like over the next few years,
you'll see a couple of models
that's gonna start coming out that are eyes off
and maybe can do freeway driving point to point
with maybe 30 seconds or a minute warning
that you need to come back online.
Then I think it'll probably be a bit more race on, right?
That might be when you see the next wave
of people sort of reconsidering their strategy
and should they partner or should they not?
But right now, because no one has actually launched
a level three product,
that pressure hasn't really arrived yet.
Maybe that's my hot take on it for you.
So, I've got a little bit of a...
I'm sorry, do you wanna keep going?
No, I just wanna frame whatever you say next by,
ugh, I predicted this, go on.
I don't think you did, it's a little out there.
So you're in what you would call
sort of like a picks and shovels business.
Now it sounds like from what you're talking about,
the companies you're working with are pretty big.
So I think maybe where my question is going
may not be directly affecting you
but I'm very curious to get your thoughts about it.
And I wanna preface this by saying
I'm way over my skis on this,
more so even than usual,
but I just wanna tease out this,
what seems to be a little bit of a trend
and see it get your thoughts on it.
It seems like with vibe coding,
one of the things that is really being addressed
is actually small operators are able to kind of build,
like spin up their own,
like I've talked to friends who run small businesses
and they're all of a sudden they don't need to go
and lock in with big SAS platforms
because they can spin up stuff that works for them.
That the picks and shovels are really,
a lot of what's getting commodified by vibe coding.
I'm curious how you think about that
in the context of the work that you're doing.
Yeah, yeah, I actually, yeah, I mean 100% agree,
like my brother is in real estate
and he's built his own hall,
like never coded a thing in his life.
And then I'm on the phone to him and he's like,
oh yeah, I built my own CRM.
I'm using like Django and Next.js.
And I'm like, what the hell?
Like you don't even know these words.
He's like, oh yeah, you know,
like Claude told me how to do it.
I was like, what is going on?
But like there's a more specific nuance there, right?
Which is that like tools are getting commoditized,
especially tools with a small number of users.
If you're building them for yourself
or for a few people,
tools that are like absolutely getting commoditized.
You can build your own little personal workflow tools.
You're the only user of it.
You vibe code a tool and then you roll it out
to 300 person autonomy org or 1,000 person autonomy org.
Now you got 1,000 mouths to feed.
They're all coming to you.
And now you're still the guy, like it doesn't matter.
Sure, the AI is accelerating your development.
It's accelerating ours too,
but still you're sitting there with bug reports
and feature requests and like trying to deal
with all this stuff.
And you can't just have a free for all of like 1,000 people,
you know, contributing randomly to the tool.
But tools are definitely getting commoditized.
However, infrastructure is growing like crazy right now, right?
So like the thing that you do not want to do
is put Claude in charge of like your production database.
How many stories I've read of people that tried that
and asked them how it went.
So like a lot of these companies are, you know,
Vercel and SuperBase and a lot of these like
and Databricks is, you know, growing like crazy.
A lot of these, you know, data companies
and infrastructure companies are actually growing even faster.
Cause now all the AI is like, you know,
all the AI is recommending, oh yeah,
go use this system for your data.
Cause you need an adult in the room
that's going to like protect your data
and make sure you don't accidentally delete it all.
So that's kind of the, you know,
the nuance I draw there,
which is that we are absolutely both, you know,
both sort of even outside of the tech industry,
but also within customers
seeing a lot more people building little tools
to solve their workflows.
But, you know, infrastructure is actually
growing a lot faster in the past 12 months.
And so just from a strategy perspective, that keeps you,
and again, it sounds like you are already kind of focused
on bigger, more established companies
rather than trying to, you know,
bundle a whole lot of little operators together
or are you sort of agnostic about that?
Yeah, we actually go about both.
So like it's a really important, for me,
a really important part of our mission
that we support the little guys as well, right?
So like obviously we make most of our money
off the big companies,
but our biggest customer hasn't been founded yet.
Like I'm so long-term bullish on the robotics industry
and there are going to be so many thousand robotics startups
over the next, you know, five years
that are getting created and they're gonna get funded
and they're gonna deploy robots in production.
So it's really important to me that like,
A, we give all those companies a leg up
and like we actually have a free plan.
We have a lot of users on a free plan.
We have a very cheap credit card plans
for a small number of users, you know.
So we help those people get a leg up.
And also, you know, obviously we wanna catch them early,
right, from a strategy perspective.
We wanna be part of their journey on
so that they trust us and can grow with us.
So who are these robot companies that are gonna get started?
This is, where do you see the new kinds
of robotic companies starting?
Yeah, every, like there are so many, you know,
there's a company in San Francisco
that has like a tire-changing robot
that they sell to like, you know, auto mechanics
and just like, hey, you know, it's just fully ordered.
Like every piece of machinery that humans operate today
is getting rethought for a world
of physical AI and is getting rethought.
And so there is maybe like one vision of the future
that you'll hear articulated that's like,
well, we're just gonna build billions of humanoids
and then the humanoids will be there
like pushing the buttons.
And I think, you know, maybe that is true for a long tail
but these humanoids for a long-term, you know,
that's bullish and long-term bullish on humanoids
but that's a very inefficient way of doing things
compared to just like rethinking the piece of machinery
so that it has actuators
and so that it can do these tasks itself, right?
So like, it's not that there's any one thing
that I think is like, there are thousands of things.
Every way you look that there's like a machine
that's getting, you know, kind of operated by humans today.
You're like, why have we got a human
that is pushing the buttons
instead of the robot doing it itself?
And do you think because, you know,
we saw this a bit in the AV space as well, right?
Where you had companies like Tesla saying, you know,
we're gonna create a general solution
like all of your, you know, the car that you own
is just gonna be able to do, you know,
and clearly we've had 10 years of that.
Now, maybe they don't, but like clearly
if you'd spent that 10 years
focused on something a little bit more attainable,
it seems like in the robotics space
we see that same dichotomy happening, right?
Where you have any investor, you're gonna have
a portfolio, you're gonna have your long shots
that, you know, are small chance of success
but big payoff if it happens
and you have your sort of safer bets
and you have a hedge portfolio.
Do you see the investment and the new companies
that are starting and the resources
that are getting allocated to them?
Do you think the balance there is about right?
Do you think there is too much going on these like
sort of maybe overly ambitious humanoid side?
Would you like to see more going into kind of
a little bit more like limited
but pragmatic kind of companies?
Like, how do you do that?
Yeah, I mean, you know, I don't think
these things are necessarily a zero sum game.
Like if you look at the amount of capital
that's going into robotics and physical AI
compared to AI more generally, I would say
that we should be allocating a lot more
like all of the, you know, all of the chatbots
and things that are only accelerating
the productivity of the digital world.
The physical world is like,
two orders of magnitude larger than the digital world, right?
So like every single physical industry
is larger like than the entire global SaaS industry.
Even individual like industries like construction
is sort of like 10 times bigger industry
than the entire global SaaS industry.
So I would say like first and foremost,
we're nowhere close to a bubble in physical AI
we should have a lot more capital going into that
versus the digital world.
But within physical AI and sort of the robotics
more broadly, there is a lot of focus right now,
you know, sort of early 2026 on, you know,
large foundation models on humanoids
and sort of consumer robotics
are starting to get a real wave.
So like, these are real categories.
It's good that money are going into them
but there are a lot that like I say,
tend to get lost in the sidelines
where, you know, there are huge opportunities
where a company could scale to a billion dollars in revenue
building some of these, you know,
just a lot less sci-fi sounding robots, right?
Where you don't hear as much about the funding rounds
but that will pick up I think,
especially as we start to see some breakaway successes
and we start to see, you know, some,
you know, I'm gonna say like quote unquote
more boring AI, physical AI companies
that are just like solving real problems in the world.
When we started to see some breakaway successes
and people were like,
whoa, how are they making a billion in revenue
doing, you know, eggs?
You're like, okay, great.
Let's pull more money into those types of companies.
All right, now my final question.
Knowing what you know from your time at cruise
and we didn't even get into your time at Coinbase
but that's another show.
Yeah, yeah, so that's another whole story.
What do you drive?
What do I drive?
I drive a Rivian, R1S.
It's a very nice vehicle.
I actually, so we, I had RJ at our conference.
We run the largest robotics developer conference
actually, it's happening this year in August.
And so we did it last year.
I had RJ on stage and did a fireside chat with him
and he's just like super cool guy down to earth,
very smart, very deep understanding
of everything in his company.
So I just, you know, I was real excited about it
and I needed to upgrade.
I had a, my wife as a model Y
I knew it.
I was waiting for it.
Yeah, the model Y.
And then which, you know, and you know, the autonomy,
I mean Tesla very far along in autonomy.
And then, yeah, but I upgraded.
I had the Subaru, Subaru Cross Trek.
It's a good ski car.
That's a very nice cross section of vehicles.
I commend you, Edward.
Yeah, but I think the Subaru,
the adventuring Subaru to Rivian,
this is has probably an underrated upgrade path.
This is, yeah, no, it's a very,
this is like, I've seen this happen actually in Portland,
across Portland with that Subaru and Rivian.
Yeah, evolution is definitely a real thing.
If we take your wife's model Y drive ownership
and combine that with your remark about Tesla
being far along, the audience,
if you could have seen the symphony of emotions
crossing Edward's face.
But we'll leave it at that because we come
to the end of our show.
Alex, I read a whole book.
Like the emotion, I purged it.
Yeah, yeah.
If anyone knows you wrote a book,
it's the audience of this show.
Kirsten.
I did have one quick follow-up.
And because you said you had RJ on your show
and this was for your developer conference.
At our conference, yeah.
So what do you think about his pursuit
or take on what he's doing with Mind Robotics?
Because that's kind of an interesting.
Oh yeah, very excited.
Yeah, we know that Mind Team pretty well
and they're early in that journey.
But this is a huge, I mean, Robotics and manufacturing
is a thing that you're only just starting
to hear a bit about.
There's a couple of companies going after this.
I think that was a very smart move.
Mind is being spun up as a separate company
but they're basically, you know,
Rivian is their first customer
and probably their only customer
for the near-to-be future.
They prove that they can do it with Rivian.
You know, coming back to the GM point,
GM has a robotics team working on
bringing autonomy into manufacturing as well, right?
Besides Sepper from the ADAS play.
So there is incredible potential
for autonomy and manufacturing.
I mean, you know, there are entire factories
running lights out in China
that don't have any humans in them.
So there is just like, you know, massive opportunity there.
The smart thing about spinning it out
I think is that that allows, you know,
with Mind obviously they're able to like
capitalize that separately and make a play
for like this is not just gonna be a,
not just be a, you know, a supply at a Rivian
but it can actually sell to the broader industry
which I think I think is like great.
So is Fox Glove then working with Mind?
No comment.
Loves that Rivian though.
We'll leave it there.
That's a perfect place to end.
The unanswerable question.
You know I was working towards that eventually.
Yeah, yeah, that's a good team over there.
Kirsten?
Yes.
Well, thank you, Adrian, for joining us
and thanks to our listeners
for tuning into another episode of the Atonic Cast.
Thanks for having me.
About this episode
Adrian Macneil, founder/CEO of Foxglove, traces his path from Cruise—where he helped build autonomy infrastructure—to launching a platform for the “data flywheel” that makes robots improve over time. He explains how Foxglove helps teams capture, mine, and debug long-tail events from real-world fleets, including incident-flagging and search workflows. The conversation covers why MCAP became “the PDF of robots” (standardizing logging formats), why robotics is easier now (compute, AI, and talent), and why edge-case problems like Waymo’s school-bus issue persist operationally.
In this episode of Autonocast, Alex Roy, Kirsten Korosec, and Ed Niedermeyer talk with Foxglove founder Adrian Macneil about how lessons from Cruise and the autonomous vehicle boom are now fueling a broader robotics revolution. Macneil explains why Foxglove is building the infrastructure layer for “physical AI,” helping robotics companies capture and learn from real-world machine data, and why today may be the best moment yet to launch a robotics startup.