Machine Minds

Hardware has long lagged behind software in speed, accessibility, and iteration. But that gap is starting to close.

Matthias Wagner, founder and CEO of Flux, joins Greg to unpack how AI is transforming electronics design from a slow, manual, and fragmented process into something far more collaborative, automated, and accessible. After years at Facebook and a deep frustration with legacy hardware tooling, Matthias set out to build what he calls the first AI hardware engineer. A system that can help anyone design, iterate, and manufacture electronics with the speed and flexibility of modern software.

From rethinking PCB design workflows to enabling entirely new classes of builders around the world, this conversation explores what happens when hardware finally gets its GitHub moment.

In this conversation, Greg and Matthias explore:

• Matthias’s journey from early software engineering to Facebook and ultimately founding Flux to tackle stagnant hardware design tooling
• Why hardware has lagged decades behind software in collaboration, automation, and developer experience
• How Flux acts as an AI hardware engineer, guiding users from concept to schematic to manufacturing-ready design
• The inefficiencies of traditional PCB design and how AI can consolidate complex systems into single, optimized boards
• Why building in the browser unlocks real-time collaboration, faster iteration cycles, and continuous product improvement
• How Flux integrates supply chain data directly into the design process to avoid costly delays and redesigns
• The shift from waterfall hardware development to more agile, software-like workflows
• Why democratizing hardware will unlock millions of new builders, not just make existing engineers more productive
• Real-world examples of non-traditional users building hardware, including farmers creating custom automation systems
• Where AI fits across the hardware stack, from component selection to simulation and layout optimization
• The reality of building a deep tech startup, including five years with no revenue and multiple near-death moments
• Lessons on fundraising for long-horizon products and why operator investors matter early on
• How AI is reshaping team structure, hiring, and what it means to be an effective engineer today
• Why tooling is the most underestimated lever in accelerating robotics and hardware innovation
• Matthias’s vision for the future where building hardware becomes so easy that “hardware is hard” disappears as a concept

If you are building in robotics, hardware, or just thinking about how AI will reshape the physical world, this episode offers a compelling look at the tools and mindset shifts required to unlock the next wave of innovation.

Website: https://www.flux.ai/

Job site: https://jobs.ashbyhq.com/flux

Connect with Flux on LinkedIn: https://www.linkedin.com/company/buildwithflux/posts/?feedView=all

Connect with Matthias Wagner on LinkedIn: https://www.linkedin.com/in/matthias-wagner-5220b047/

Flux X: https://x.com/BuildWithFlux

Matthias' X: https://x.com/MatthiasWagner

Connect with Greg on Linkedin: https://www.linkedin.com/in/gregtoroosian/

From climbing robots inspecting boiler tubes to AI-powered platforms optimizing the world’s most critical assets, Gecko Robotics is redefining how we understand and maintain the infrastructure that powers modern society.

Ed Bryner, Chief Technology Officer at Gecko Robotics, joins Greg to unpack how his journey from hands-on engineer to technical leader has been shaped by a deep focus on applied engineering, mission-driven teams, and building technology directly in the real world. With roots in robotics competitions, defense work, and industrial systems, Ed brings a uniquely grounded perspective on what it takes to move from prototype to production in some of the harshest environments on earth.

We explore how Gecko is building a vertically integrated stack from robot to cloud, why infrastructure health data is the missing layer in industrial decision-making, and how continuous inspection is unlocking entirely new ways to operate, maintain, and extend the life of critical assets.

In this conversation, Greg and Ed get into:

• Ed’s path into robotics, from a family of engineers to high school competitions that sparked a passion for building at the intersection of hardware and software
• The founding story of Gecko Robotics, starting with a wall-climbing robot designed to inspect boiler tubes and eliminate dangerous manual inspections
• How Gecko evolved from a robotics company into a data and AI platform creating “health records” for industrial infrastructure
• Why infrastructure inspection has historically been so challenging, from scale and complexity to reliance on manual, high-risk human labor
• The power of multimodal data collection, combining ultrasound, LiDAR, and visual data to create high-fidelity digital twins of critical assets
• What it means to build and deploy robots in extreme environments like power plants, submarines, and refineries and why lab-only development fails in the real world
• How continuous data collection, even while assets are operating, is transforming maintenance cycles, planning, and operational availability
• A real-world example of reducing unplanned outages from 12 to zero using Gecko’s inspection and analytics platform
• The shift from static reports to interactive, software-driven decision tools that connect operators, engineers, and executives around a shared source of truth
• The challenge of reliability in robotics and what it takes to build systems that survive dirty, high-risk industrial environments
• How Gecko structures its teams around vertical integration, bringing hardware, software, and domain experts together to accelerate innovation
• Why “orientation” and getting engineers into the field is critical to shortening development cycles and building products that actually work
• The balance between experimentation and scaling and Gecko’s philosophy of proving value with a few customers before expanding broadly
• How advances in AI and developer tools are accelerating experimentation and enabling engineers to work across the full stack
• Ed’s long-term vision of improving the health, lifespan, and sustainability of the world’s built infrastructure

For anyone building in robotics, industrial automation, or physical AI, this episode is a deep dive into what it really takes to deploy technology in the real world and create lasting impact on the systems society depends on every day.

Learn more about Gecko Robotics: https://www.geckorobotics.com/

Connect with Ed Bryner: https://www.linkedin.com/in/edwardbryner/

Connect with Greg Toroosian: https://www.linkedin.com/in/gregtoroosian/

Mobile robots are rapidly spreading across warehouses, hospitals, factories, and beyond. But as fleets grow and companies deploy robots from multiple vendors, a new challenge has emerged. The robots often cannot communicate with each other. Founder and CEO Aldus von der Burg joins Greg to discuss the “interoperability gap” in robotics and why solving it could unlock the next wave of automation.

Aldus shares the unconventional journey that led him into robotics. After studying automotive engineering and working at startups in Denmark, he explored drone delivery before regulatory hurdles forced a pivot. That experience led to the founding of Meili Robots in 2019, and eventually to a realization that the biggest barrier to scaling robotics was not hardware capability, but the software infrastructure needed to coordinate diverse robot fleets.

Today, Meili Robots is building a universal fleet management platform that allows robots from different manufacturers to operate together seamlessly. By taking a hardware-agnostic approach, the company aims to remove friction for operators, integrators, and manufacturers deploying robots across industries.

In this conversation, Greg and Aldus explore:

• Aldus’s path from automotive engineering and drone startups to founding Meili Robots
• The “interoperability gap” preventing robots from different manufacturers from collaborating effectively
• Real-world examples of robot gridlock and how poor coordination creates downtime, safety risks, and lost productivity
• Why many robotics companies build great hardware but ship weak or outdated software stacks
• How Meili’s platform enables vendor-agnostic fleet management across industries like warehousing, healthcare, agriculture, and mining
• The importance of operator independence through configurable tools and no code interfaces
• Why lab demonstrations of autonomy rarely survive real-world deployment environments
• Lessons learned selling into enterprise and industrial automation markets, including the slow pace of procurement and compliance
• Aldus’s hiring philosophy for early stage robotics teams, focusing on personality, curiosity, and strong engineering culture
• His candid take on the robotics hype cycle, including why humanoids may be overhyped compared to practical automation solutions

This episode is a deep dive into the invisible infrastructure layer that will determine whether robots remain isolated tools or become collaborative systems that scale across entire facilities and industries.

Learn more about Meili Robots: https://www.meilirobots.com

Connect with Meili Robots on Linkedin: https://www.linkedin.com/company/meilirobots/

Connect with Aldus on Linkedin: https://www.linkedin.com/in/aldusvdb/

Connect with Greg on Linkedin: https://www.linkedin.com/in/gregtoroosian/

Humanoid robots are dancing, backflipping, and going viral. But are they ready to do real work?

Theo Nash, founder and CEO of Mundane, joins Greg to challenge the hype cycle and refocus the conversation on what robots are actually for: eliminating dull, dirty, and dangerous work while amplifying human capability.

From growing up around London auto garages to studying at Stanford and building teams across Palo Alto, Shenzhen, and Vancouver, Theo shares how his path shaped a mission centered on safety, embodiment, and human–robot collaboration. Mundane is not chasing flashy demos. Instead, the team is rebuilding the robotics stack from first principles, with a bold bet on telepresence, tactile sensing, and intuitive control systems that feel less like a video game and more like becoming the robot.

This is a deep dive into embodiment, reliability, vertical integration, and why scaling too fast could be the biggest mistake in robotics today.

In this conversation, Greg and Theo explore:

• Why most humanoid demos miss the real question: can the robot do economically valuable work reliably?
• Mundane’s founding principle that robots should eliminate drudgery, not replace human fulfillment
• The embodiment gap and why teleoperation must feel like true presence, not remote control
• Proprioception in robotics and what it means to “feel” where a robot’s limbs are in space
• Why wiping a table is harder for a robot than playing chess
• The hidden challenge of reliability, from operator accuracy to hardware consistency between units
• Lessons from China’s advanced manufacturing ecosystem and how speed and supply chains shape hardware startups
• Why scaling slowly, even in a hype cycle, may be the smartest long term strategy
• Telepresence as infrastructure and the vision for a VP of Teleportation
• How robots can reduce workplace injuries and remove humans from hazardous environments like oil rigs and disaster zones
• Why Theo believes AI and robotics are collaborators, not replacements for human workers
• Building a team culture that works hard, plays hard, and debates openly without hierarchy

If you are building physical AI, deploying robots in the field, or thinking about the future of human labor in an automated world, this episode is a thoughtful and ambitious look at what it will actually take to make robots useful at scale.

Connect with Theo Nash: https://www.linkedin.com/in/theo-nash/

Connect with Greg Toroosian: https://www.linkedin.com/in/gregtoroosian/

What happens when a neuroscientist stops asking why the brain works the way it does and starts building his own version in silicon?

Sam Fok, co-founder and CEO of femtoAI, joins Greg to explore the journey from studying neurons to designing ultra efficient AI chips for real world devices. Growing up in a household of scientists, Sam was drawn early to questions about intelligence and consciousness. That curiosity eventually led him from neuroscience labs to Stanford’s Brains in Silicon Lab and into the world of neuromorphic engineering.

Today, femtoAI is building brain inspired hardware and software that brings AI out of massive data centers and into everyday products such as wearables, appliances, robotics, and consumer electronics. The goal is simple but ambitious: make AI run efficiently in the real world without straining power grids, supply chains, and balance sheets.

Greg and Sam dive into the technical, commercial, and personal pivots required to turn deep research into a scalable company, including the hard decision to rebuild their core platform from the ground up.

Highlights:

• Sam’s path from neuroscience and recording brain signals to building neuromorphic chips at Stanford
• The founding story of femtoAI and why timing mattered as AI crossed from research curiosity to commercial force
• How femtoAI enables AI to run directly on devices like glasses, earbuds, appliances, and robots instead of relying on the cloud
• Why sparse computation and localized processing, inspired by biology, are key to unlocking efficient edge AI
• The massive pivot from PhD era technology to a commercially viable platform built for manufacturability and scale
• The tension between academic incentives and commercial realities, from benchmarks and novelty to reliability and reproducibility
• Two distinct customer profiles: companies that want turnkey AI features and those that want full stack tools to build their own models
• Why ease of use and integration often matter more than technical elegance
• Lessons on earning credibility in deep tech by delivering real, testable results rather than selling grand visions
• Building highly technical teams with intellectual curiosity and cross stack awareness
• The importance of trust and empowerment when leading engineers across silicon, firmware, compilers, and AI tooling
• Why Sam believes today’s dominant AI approaches alone will not be enough to reach truly general robotics
• What the next three to five years could look like as intelligence spreads across devices and environments

If you are building AI powered hardware, deploying robots, or thinking about how to bring intelligence out of the cloud and into physical products, this conversation offers a grounded look at what it really takes to make AI usable, efficient, and scalable in the real world.

Learn more about femtoAI: https://femto.ai/

Developer portal: https://developer.femto.ai/

Connect with Sam Fok: https://www.linkedin.com/in/sam-fok/

Connect with Greg Toroosian: https://www.linkedin.com/in/gregtoroosian/

From aerospace engineering to tactile sensing and physical AI, Jennifer Kwiatkowski’s path into robotics was driven by a simple but powerful motivation: taking math and physics off the whiteboard and making them work in the real world.

Now an AI Specialist at Robotiq, Jennifer works at the intersection of hardware, software, and customer reality. In this conversation, she joins Greg to unpack what it actually takes to deploy robots that deliver ROI, why integration is often harder than the robot itself, and how the industry can avoid getting lost in hype cycles.

Jennifer brings a research background in robotic manipulation and tactile sensing, along with firsthand experience navigating the fragile complexity of physical systems. Her perspective cuts through buzzwords and focuses on what matters: reliability, safety, usability, and long term value for customers.

In this conversation, Greg and Jennifer explore:

• Jennifer’s journey from mechanical engineering at McGill to a decade deep in robotics and AI
• Why engineers, not just scientists, are the ones who bring theory into the physical world
• The hidden challenge in robotics: integration can account for 50 to 80 percent of total automation cost
• Lessons from building automated pick and place systems for data collection and the real data barriers in physical AI
• What separates shiny robotics demos from systems that actually deploy and scale
• Why vertical focus and application specific reliability win over general purpose promises
• The hard truth about automation: if a human still has to supervise it, you have not automated anything
• How AI meaningfully adds value in unstructured environments, and when it simply adds unnecessary complexity
• The importance of setting the right expectations with customers and practicing real customer empathy
• Why supply chain and support operations become major bottlenecks when robotics companies scale
• Using OKRs to align teams without creating decision making bottlenecks
• Safety and privacy concerns as robots move into unconstrained human environments
• The current hype cycle around humanoids and where Jennifer sees sustainable growth in robotics adoption
• Why she is excited about “weird robots” and unconventional form factors that move beyond human imitation

For founders, engineers, and operators building physical AI systems, this episode offers a grounded look at what it takes to move from prototype to production, and from hype to durable impact.

Learn more about Robotiq: https://robotiq.com/

Connect with Jennifer Kwiatkowski on LinkedIn: https://www.linkedin.com/in/jenniferkwiatkowski/

Connect with Greg Toroosian on LinkedIn: https://www.linkedin.com/in/gregtoroosian/

As robots move out of cages and into shared human environments, safety is no longer just about stopping motion. It is about enabling systems to keep moving productively without increasing risk. 3Laws Robotics is tackling one of the hardest problems in autonomy: how to guarantee safety while machines remain in motion.

Andrew Singletary, co-founder and CEO, and Amir Sharif, COO of 3Laws Robotics, join Greg to unpack how a decade of academic research in safety critical control evolved into a commercial platform for dynamic safety across robotics, vehicles, and autonomous systems. From warehouse robots and mobile manipulators to aircraft and automobiles, their work aims to make autonomy scalable, certifiable, and trusted in the real world.

Andrew traces his path from high school robotics competitions to nuclear engineering at Georgia Tech, a PhD at Caltech, and ultimately founding 3Laws to productize safety technology that companies repeatedly asked him to integrate. Amir brings the operator’s lens shaped by multiple startups and exits, focusing on product discipline, execution, and building an organization that values getting it right over being right.

Together, they explore what it actually takes to move beyond academic proofs into functional safety that survives broken sensors, real world uncertainty, and regulatory scrutiny.

Highlights:

• The origin of 3Laws Robotics and the influence of Isaac Asimov’s Three Laws on the company’s philosophy
• Why traditional industrial safety equates safety with stopping and why that breaks down for mobile robots, vehicles, and aircraft
• The concept of dynamic safety and how robots can remain in motion while maintaining established safety guarantees
• How safety events, gridlock, and human resets quietly destroy productivity across robot fleets
• The gap between academic definitions of safety and industry expectations for functional safety
• Why safety must be designed in early rather than bolted on at the end of development
• How 3Laws works with both established robot form factors and emerging systems like mobile manipulators and humanoids
• The role of exposure and scale in redefining safety risk as robot fleets grow
• Lessons from hiring early team members and why skepticism is often a signal of deep understanding
• How Andrew and Amir divide responsibilities between vision, product, and operations without rigid boundaries
• Why culture, values, and hiring standards act as long term accelerants rather than speed bumps
• The challenge of focus when a single safety platform can apply to logistics, automotive, aerospace, and beyond
• What founders often get wrong about changing industries versus working within existing frameworks
• Why certification and third party validation are critical to trust in autonomous systems

For founders, operators, and engineers building robots that must operate alongside people, this conversation offers a grounded look at how safety, productivity, and autonomy intersect and why the future of robotics depends on systems that can move intelligently, not just stop.

Learn more about 3Laws Robotics: https://3laws.io

Connect with Andrew Singletary on Linkedin: https://www.linkedin.com/in/andrewsingletary/

Connect with Amir Sharif on Linkedin: https://www.linkedin.com/in/asharif/

Connect with Greg Toroosian on Linkedin: https://www.linkedin.com/in/gregtoroosian

From academic breakthroughs in legged locomotion to building a horizontal AI platform for millions of robots, Flexion Robotics is taking on one of the hardest problems in the field: how to make robots learn, adapt, and scale in the real world.

Nikita Rudin, co-founder and CEO of Flexion Robotics, joins Greg to share his journey from growing up in Switzerland with dreams of space exploration, to pioneering reinforcement learning for robots during his PhD, to helping shape NVIDIA’s simulation tools, and ultimately to founding a company aimed at becoming the “brain” behind general purpose robots.

Nikita brings a rare perspective that spans deep research, real-world demos, and early-stage company building. He explains why hand-engineered behaviors and teleoperation do not scale, how reinforcement learning and simulation unlock robustness, and what it really takes to move robotics from impressive demos to economically viable deployments.

In this conversation, Greg and Nikita explore:

• Nikita’s path from physics to engineering to robotics, and why he has always been drawn toward faster real-world impact
• Early breakthroughs in neural network controlled quadrupeds and why reinforcement learning became the only scalable path forward
• The hidden human effort behind robotics demos, and why that model cannot support general purpose robots
• How Flexion Robotics is building a horizontal AI software platform to serve as the brain across many robot types
• The role of simulation, reinforcement learning, and the sim-to-real gap in training robots safely and at scale
• Why reinforcement learning leads to more robust systems that can recover from failure rather than collapse
• The limits of simulation, where real-world data still matters, and why mixing both is essential
• Why demand for humanoid and general purpose robots is already strong across logistics, manufacturing, and material handling
• The tension between customer engagement and long-term platform development in a frontier technology startup
• What Nikita looks for in exceptional roboticists, including action-oriented thinking, healthy stubbornness, and strong collaboration
• Underappreciated roles in robotics startups, from infrastructure engineers to data and deployment pipelines
• How Nikita’s leadership style has evolved as the team scaled to more than 40 engineers
• Where the robotics industry risks over-promising, especially around consumer and home robots
• Why hype can be healthy if expectations and timelines remain honest
• Advice for founders on underestimating team size, timelines, and the operational realities of company building
• Flexion’s near-term milestones, including moving from demos to real industrial deployments and making the economics work at scale
• How a successful horizontal robotics platform could accelerate the entire ecosystem

If you are building robots, investing in them, or trying to understand what it will take to move from research breakthroughs to real-world impact, this episode offers a grounded, deeply technical, and forward-looking view of where robotics is headed.

Learn more about Flexion Robotics: https://flexion.ai

Follow Flexion Robotics on LinkedIn and X for updates and technical insights

Connect with Nikita Rudin on LinkedIn: https://www.linkedin.com/in/nikita-rudin

Connect with Greg Toroosian on LinkedIn: https://www.linkedin.com/in/gregtoroosian

From gigabytes of robot telemetry per minute to natural language search across multimodal data, Alloy is tackling one of the most underappreciated bottlenecks in robotics: making sense of what robots are actually doing in the real world.

Joe Harris, founder of Alloy, joins Greg to unpack how his background in electrical engineering, machine learning, and growth teams shaped a product that helps robotics companies move faster, ship more reliably, and avoid rebuilding the same internal tooling over and over again. What started as frustration with inaccessible data and slow feedback loops has become a platform designed to turn robot data into a shared, searchable source of truth across engineering, validation, and commercial teams.

The conversation dives deep into why replay tools break down at scale, how modern LLMs are changing what’s possible with robotics telemetry, and why deciding what not to build is one of the most important skills for early-stage founders.

Highlights:

• Joe’s path from electrical engineering and machine learning research into growth teams at scale, and how feedback loops became a unifying theme across software and robotics
• Why robotics companies are drowning in data but starving for insight, with robots generating gigabytes per minute across video, sensor data, and logs
• How Alloy helps teams move beyond one-off replay by enabling cross-sectional analysis, natural language search, and summarized field test reports
• The validation and verification teams who feel the value first, and how faster analysis turns into faster deployments for customers
• Why most robotics startups should not build their own telemetry and analysis stack, and how the industry is entering a tooling renaissance similar to early cloud software
• The importance of pain times frequency when deciding what features to build and what to cut
• Lessons from early mistakes, including why free pilots often fail and how paid pilots create real commitment on both sides
• Joe’s philosophy on early hiring, small teams, mission alignment, and building culture without unnecessary process
• What the next 12 to 18 months look like for Alloy as robotics fleets scale and foundation models reshape the landscape
• A long-term vision for a world of abundant automation, where robots learn continuously from experience and data interpretation becomes critical infrastructure

If you are building robots, deploying them at scale, or thinking about the unseen infrastructure required to make robotics reliable in the real world, this episode offers a candid and deeply technical look at what it takes to turn raw robot data into real-world progress.

Learn more about Alloy: www.usealloy.ai

Connect with Joe Harris: https://x.com/_joe_harris_

Connect with Greg Toroosian: https://www.linkedin.com/in/gregtoroosian

Manufacturing has long been the bottleneck between imagination and reality. From aerospace to automotive, complex physical products still take years to tool, validate, and produce. Machina Labs is working to change that equation by turning factories into flexible, software-driven systems that can build almost anything, anywhere.

Edward Mehr, co-founder and CEO of Machina Labs, joins Greg to unpack his journey from early software obsessions to SpaceX, and ultimately to founding a company focused on rethinking how the world makes metal parts. Drawing from hands-on experience across software, robotics, and aerospace manufacturing, Edward shares why factories themselves are the real product, and how breaking the link between tooling and design unlocks speed, resilience, and creativity.

The conversation explores how Machina Labs’ robotic “Robocraftsman” systems combine dexterity, AI-driven learning, and modular deployment to form metal without molds or dies. The result is manufacturing that adapts as fast as software, enabling rapid iteration, distributed production, and entirely new business models.

Highlights from the conversation include:

• Edward’s early fascination with computers and making things, and how hands-on craft and coding shaped his view of the physical and digital worlds
• Lessons from SpaceX on why manufacturing speed, not engineering ambition, is often the true constraint in hardware innovation
• Why traditional factories are locked to specific designs and materials, and how Machina Labs is building product-agnostic, software-defined factories
• The concept of the Robocraftsman, a robotic system that learns like a human craftsperson and adapts processes in real time using data and AI
• How Machina Labs captures data from both physical forming and simulation to train models that optimize force, tooling, and process parameters
• Early traction in aerospace and defense, including dramatically reducing lead times for legacy aircraft parts that once took years to replace
• Expanding into automotive manufacturing and enabling mass customization directly from OEMs without expensive tooling
• A major partnership in the UAE focused on rapidly deployable, distributed manufacturing for defense and commercial resilience
• The strategic importance of factories as national security assets in an era of fragile global supply chains
• How portable, containerized manufacturing systems open the door to off-world production on the Moon, Mars, and beyond
• The challenges of building multidisciplinary teams across robotics, AI, and materials science, and how leadership evolves as companies scale
• Edward’s vision for the future of manufacturing, where physical expression becomes as fast, personal, and iterative as software development

If you are building hardware, scaling robotics, or rethinking how physical products get made, this episode offers a deep look at what it takes to bring software-speed thinking into the world of atoms.

Learn more about Machina Labs: https://machinalabs.ai/

Machina Labs Advances Custom Automotive Manufacturing with AI and Robotics: https://machinalabs.ai/resources/machina-labs-advances-custom-automotive-manufacturing-with-ai-and-robotics

Strategic Development Fund Announces Investment and Initial Agreement with Machina Labs: https://machinalabs.ai/resources/uae-strategic-development-fund-announces-investment-and-initial-partnership-with-machina-labs

Connect with Edward Mehr on LinkedIn: https://www.linkedin.com/in/edward-mehr/