Founders in Arms

Mehul Nariyawala is the co-founder and President of Matic Robotics, a home robotics company building what he calls “robotics 2.0” — intelligent, vision-first robots designed to actually work in real homes. After early careers at Nest and a prior acquisition by Google, Mehul and his team spent seven years building Matic, challenging the assumptions behind robot vacuums, consumer hardware, and how robotics companies should scale.
In this conversation, Mehul breaks down why robotics is far harder than software, why most home robots quietly fail, and how Matic approached everything differently — from vision-only robotics and in-house manufacturing to avoiding subscriptions, ads, and premature market creation.
What you’ll learn:
Why robotics is “100× harder than software” — and where most teams underestimate the work
The difference between automation and true intelligence in home robots
Why negative-NPS categories can hide massive opportunities
How Matic beat entrenched incumbents like Roomba by fixing fundamentals, not adding features
Why vision-only robotics was a risky but necessary bet
The real reason humanoid robots are still far from consumer-ready
Lessons from Nest on why some hardware categories stay defensible for decades
Why creating a new market can be fatal for hardware startups
How Matic built robots in-house in California instead of outsourcing manufacturing
The tradeoffs between subscriptions, ownership, and consumer trust
Why great hardware products must earn word-of-mouth before growth

In this episode, we cover:
(00:00) Introduction to Mehul Naryawala and Matic Robotics
(01:10) Why robotics is dramatically harder than software
(03:00) The failure modes of early robot vacuums
(05:10) Identifying opportunity in negative-NPS markets
(07:45) Automation vs. intelligence in consumer robotics
(10:15) Why vision-only robotics was a foundational bet
(14:00) Lessons from Nest on defensible hardware categories
(17:30) Why Matic avoided creating a new market
(20:45) In-house manufacturing and vertical integration
(24:30) Scaling hardware without inventory risk
(28:10) The long road from demo to product
(32:00) Why humanoid robots are still overhyped
(36:20) Word-of-mouth, product-led growth, and brand trust
(40:15) Subscription fatigue and consumer psychology
(44:30) The future of home robotics and where Matic goes next

Jack Zhang is the co-founder and CEO of Airwallex, a global payments and financial platform valued at $5.5 billion. Founded in Melbourne, Airwallex processes billions in cross-border transactions and serves businesses expanding internationally. Jack shares his journey from starting the company to competing with giants like Stripe, navigating the complexities of global payments infrastructure, and building across multiple regulated markets.
What you'll learn:
Why cross-border payments remain broken despite decades of fintech innovation
How Airwallex competes against Stripe and other established payment platforms
The challenge of building financial infrastructure across multiple countries and regulations
Jack's perspective on fair competition versus FUD (fear, uncertainty, doubt) tactics in business
Why Airwallex is deploying $1 billion in the US market over the next three years
The reality of being a foreign founder building in America during geopolitical tensions
How payment infrastructure for global businesses differs from consumer fintech
The trade-offs between growth velocity and sustainable business building
Jack's philosophy on money, success, and what matters after achieving wealth at 30
Why he chose to stay in Melbourne instead of relocating to San Francisco

In this episode, we cover:
(00:00) Introduction to Jack Zhang and Airwallex
(02:34) Early days of Airwallex and the founding story
(05:12) The problem with cross-border payments
(08:45) Competing with Stripe and other payment platforms
(12:18) Building in regulated markets and compliance challenges
(16:23) FUD (fear, uncertainty, doubt) tactics in business competition
(19:13) Raj's experience with FUD at Lyft vs Uber
(22:47) Navigating geopolitical tensions as a Chinese-Australian founder
(25:36) The $1 billion US market investment commitment
(27:41) Product philosophy and fair competition
(31:15) Going upmarket vs staying with SMBs
(35:22) Life choices: Melbourne vs San Francisco
(37:49) Perspective on wealth - "not about the money"
(42:18) The future of payments infrastructure
(45:30) Advice for founders building in competitive markets

Ryan Gariepy is the co-founder and former CTO of Clearpath Robotics and Otto Motors, acquired by Rockwell Automation for $600M+ in 2023. He bootstrapped the company for five years with only $300K in funding, reached profitability in 18 months, and spent 14 years building mobile robotics platforms that became the industry standard for research and industrial automation.
(If you’re looking for inspiration and lessons from other founders, Founders in Arms is hosting a founders roundtable with Rajat Suri, Immad Akhund, and Max Mullen next Wed Jan 14th at Mercury HQ. Discussing war stories and sharing lessons with a group of founders, as part of Founders-in-Arms podcast. Will be food and drinks. Capacity strictly limited at 50 so apply early if you’re interested: https://luma.com/dk97inyk )
What you'll learn:
Why robotics is a systems discipline where progress stacks rather than explodes
How to bootstrap a hardware company to $10M revenue before raising venture capital
Why robotics follows 20-50% sustained growth for decades vs. software's boom-bust cycles
The "promise problem" with humanoid robots and why form factor shapes user expectations
How manufacturing in Canada (not China) became a strategic advantage for Clearpath
Why founders overestimate 2-year progress but underestimate 10-year impact in robotics
The real economics of humanoid robots: $20K cost becomes $80K landed price
How robotics investment differs from software: less competitive, more defensible
Why experience compounds in hardware but expires in software careers
Investment criteria for robotics: engineering risk vs. technical risk and go-to-market strategy

In this episode, we cover:
(00:00) Introduction and live event announcement
(03:29) Ryan's background: Clearpath Robotics and Otto Motors
(04:06) Building two brands under one company
(06:29) The 14-year journey: challenges and non-linear growth
(07:11) Bootstrapping robotics when "nobody thought you could make money"
(08:17) Reaching profitability in 18 months with research customers
(10:28) Building robotics platforms for MIT, universities, and research labs
(11:03) Manufacturing in Canada vs. outsourcing to Asia
(15:05) Reconnecting after 20 years: the Waterloo entrepreneurship connection
(16:17) Working at Kiva Systems (now Amazon Robotics)
(18:10) Why robotics is more exciting now than ever in history
(19:21) Robotics as systems discipline: no single breakthrough technology
(21:22) The overhype cycle and realistic expectations
(22:14) Software explodes then crashes; robotics compounds for decades
(23:36) Why hardware is harder but more mission-driven
(25:27) The talent pool advantage: people irrationally love hardware
(27:30) Physical AI and real-world impact beyond software optimization
(28:07) Humanoid robots: incredible tech, miscalibrated expectations
(32:41) The "promise problem": form factors make promises to users
(34:35) Consumer robotics examples: Matic cleaning robot
(35:59) Asia...

Emmett Shear is the founder and CEO of Softmax, an alignment research company, and previously co-founded and led Twitch as CEO. He was also a Y Combinator partner and briefly served as interim CEO of OpenAI.
What you'll learn:
Why AI alignment and AGI are fundamentally the same problem
How theory of mind is the critical missing piece in current AI systems
Why continuous learning requires self-modeling capabilities
The dangerous truth: alignment is a capacity for both great good and great evil
Why "aligned AI" really means "aligned to me"—and why that's concerning
How societies of smaller AIs will outcompete singleton superintelligences
Why AI needs to be integrated with humans, not segregated into AI-only societies
The Twitch lesson: people don't want easy, they want good
Why 99% of AI startups are building labor-saving tools instead of value-creating products
How parenting and AI development mirror each other in surprising ways
Why current AI labs are confused about continuous learning
Conway's Law applied to AI: you ship your org chart
The problem with mode collapse in self-learning systems
Why emotions are training signals, not irrational noise
Emmett's biggest mistake at Twitch: chasing new products instead of perfecting the core

In this episode, we cover:
(00:00) The dangerous truth about AI alignment
(01:13) Introduction to Softmax and organic alignment
(02:05) What alignment actually means (and why most people are confused)
(03:33) The output: training environments for theory of mind
(05:01) Continuous learning and why it's so hard
(06:25) Multiplayer reasoning training in open-ended environments
(07:14) Aligned to what? The critical question everyone ignores
(08:40) Why alignment is always relative to the aligning being
(11:07) Cooperation vs. competition: training for the real world
(12:56) Is AGI an urgent problem or do we have time?
(13:15) AGI and alignment are the same problem
(15:25) Alignment capacity enables both good and evil
(17:13) The singleton problem and why societies of AIs make sense
(20:41) Building alignment between AIs and humans
(22:09) Why Elon's "biggest cluster" strategy might be wrong
(23:06) AI must be aligned to individual humans, not humanity
(25:03) What does the atomic unit of AI look like?
(28:02) Adding a new kind of person to society
(29:06) Everything will be alive: from spreadsheets to cars
(30:00) From Twitch retirement to Softmax founding
(31:26) Research vs. product engineering at early-stage startups
(32:41) Raising money for AI research in the current era
(34:30) Why Softmax will ship products
(34:50) Ilya's closed-loop research vs. open-loop learning
(36:36) How you do anything is how you do everything
(37:28) The continuous learning problem explained simply
(38:29) Mode collapse: why AIs become stereotypes of themselves
(39:33) The reward problem and why humans need emotions
(40:48) Why LLMs are trained to avoid emotions
(41:52) Watching children learn while building learning AI
(43:04) Advice for first-time AI founders
(45:08) Treat AI as clay to be molded, not a genie granting wishes
(45:50) The Twitch lesson: people want good things, not easy things
(47:22) Why 99% of AI companies are building the wrong thing
(48:16) Rapid fire: biggest career mistake at Twitch
(50:15) Which founders inspire Emmett most
(50:56) The passing fad: AI slop generators

Immad Akhund and Raj Suri reunite for a one-on-one conversation covering the biggest tech shifts of 2025, from Mercury's public launch of Personal Banking to the quieting of AGI doom discussions. This wide-ranging episode explores why self-driving cars may matter more than AGI, how vibe coding is changing software development, and the strategic decisions founders make when everyone else disagrees.
What you'll learn:
Why Immad launched Mercury Personal despite investor and team skepticism—and the founder lesson about following conviction
How Mercury Personal brings business-grade financial controls to personal banking (collaboration features, automatic categorization, 3.5% savings rates)
The existential threat facing OpenAI and Anthropic as AI models commoditize and Google leverages distribution advantages
Raj's vibe coding experiment: Building a full-stack app with Postgres backend using just prompts (and why Replit won)
Why Tribe is rejecting the $30/user ad model to build a premium, ad-free group chat platform
The retention metrics showing Tribe's product-market fit (20-40% six-month retention with minimal marketing)
How AI hype shifted from AGI doom conversations to practical commercial applications in 2025
Why self-driving technology (Waymo, Tesla FSD) represents a more immediate transformation than AGI
The best and worst of 2025: renewed tech energy vs. immigration scapegoating and Doge's failure to deliver government efficiency
Why supply constraints (chips, power) signal AI demand is real, not a bubble
In this episode, we cover:

(00:00) AGI conversations cooling down in 2025
(01:50) Mercury Personal launch after year-long waitlist
(02:42) Business-grade controls for personal banking
(04:30) 3.5% savings rates and Treasury/Invest products
(06:15) Following founder conviction despite opposition
(07:33) Balancing product shipping with polish
(08:26) OpenAI's Code Red and focus strategy
(09:23) Google's distribution advantage vs. OpenAI
(10:33) The API commoditization threat to Anthropic
(12:34) Why ad economics dominate the internet
(14:58) Facebook's $30/user vs. subscription models
(17:22) Tribe's progress: retention, AI features, monetization plans
(21:42) Vibe coding experiment: Replit vs. Lovable vs. Wix
(26:31) Why Replit might own the vibe coding market
(28:05) Enterprise use cases for AI-generated apps
(33:26) 2025's best: renewed tech energy and deregulation
(34:51) 2025's worst: immigration scapegoating and Doge's failure
(40:48) Self-driving breakthrough: Waymo and Tesla FSD
(42:31) Why AGI talk has quieted down
(43:43) Supply constraints proving AI demand is real