Humanoid robots are everywhere on social media right now. They dance, they backflip, they carry boxes across a stage. It makes for great video. It also raises a fair question: what is really happening behind the curtain?

That question sits at the center of the 124th episode of Machine Minds, a conversation between host and CEO of Samson Rose, Greg Toroosian, and Nikita Rudin, co-founder and CEO of Flexion Robotics. Nikita has spent most of his career on one stubborn problem: how do you teach a robot to move, adapt, and recover in the messy real world without a person steering every step?

Over roughly 46 minutes, the two cover everything from a childhood in Switzerland to the economics of industrial deployment. Here are the biggest ideas from the episode, and why they matter if you build, fund, or hire for robotics.

From Dreaming About Space to Building Robot Brains

Nikite grew up in Switzerland with his eyes on space exploration. That curiosity pulled him through physics, then engineering, and eventually into robotics. A thread he returns to again and again is a pull toward faster, more tangible impact. Pure theory was never quite enough. He wanted to watch things work.

His PhD at ETH Zurich put him at the front of a major shift in the field: using reinforcement learning to control legged robots. Early work on neural network-controlled quadrupeds showed what was possible, and Nikita makes the case that reinforcement learning became the only path forward that truly scales. From there, he joined NVIDIA, where he helped shape Isaac Gym and Isaac Lab, simulation tools that are now widely used across the industry. Those years set the stage for Flexion.

The Hidden Human Effort Behind Robot Demos

One of the most refreshing parts of the conversation is how honest Nikita is about demos. A polished video can hide a lot of human work, including hand-engineered behaviors and teleoperation. That approach is fine for a stage. It is a problem for a product.

Here is the issue in plain terms:

Hand-engineered behaviors need specialist time for every new task and every new robot.
Teleoperation keeps a person in the loop, so labor costs never really go away.
Neither approach prepares a robot for the surprises of a real workplace.

If the goal is a general-purpose robot, someone has to build a system that learns instead of one that is scripted. That is the bet Nikita and his team are making.

Why Reinforcement Learning and Simulation Matter

Reinforcement learning sits at the core of Flexion’s approach, and Nikita explains why in practical terms. Robots trained this way tend to be more robust. Instead of collapsing when something goes wrong, they can recover. Anyone who has watched a robot topple over after a small bump will understand why that matters.

Simulation plays a big part here, because it lets a team train robots safely and at scale. A robot can fall a thousand times in a virtual world without breaking a single actuator. The catch is the sim-to-real gap, the difference between what works in a simulator and what works on a factory floor.

Nikita is clear that simulation has limits. Real-world data still matters, and the strongest results come from mixing both. It is a balanced view, and a useful one for anyone tempted to treat simulation as a magic wand. Teams that lean only on virtual training risk surprises on the floor, while teams that skip simulation burn time and hardware learning the hard way.

One Brain, Many Bodies: The Horizontal Platform Idea

So what is Flexion actually building? Think of it as a horizontal AI software platform that can act as the brain across many different robot types. Rather than designing one specific machine, the company wants to supply the intelligence that many machines can use.

Nikita argues that a successful horizontal platform could lift the whole ecosystem. Hardware makers could spend less time reinventing the software stack and more time perfecting their robots. Customers could get capable machines sooner. It is an ambitious idea, and the conversation does a nice job of showing both the promise and the work involved. Flexion’s robots in simulation research give it a head start, since its founders helped build tools the wider industry already relies on. Still, a platform only earns its place when it works across different bodies, tasks, and environments, and Nikita does not pretend that is easy.

Where Demand Is Already Real

There is plenty of talk about humanoids and general-purpose robots someday arriving in homes. Nikita points to a more immediate story. Demand is already strong in:

Logistics
Manufacturing
Material Handling

These are industries with repetitive, physical, and often hard-to-staff work, which makes them a natural fit for capable robots.

He also shares a tension that many frontier startups know well. Working closely with customers keeps a company grounded, but chasing every request can pull the team away from building the long-term platform. Finding the balance is part of the job. Looking ahead, Flexion’s near-term milestones include moving from demos to real industrial deployments and making the economics work at scale, which is where many robotics companies either prove themselves or stall.

What Makes a Great Roboticist

Greg asks Nikita what he looks for when hiring, and the answer is not a list of buzzwords. Nikita highlights three qualities:

Action-oriented thinking: A bias toward trying things and learning fast.
Healthy stubbornness: The persistence to keep going when a problem refuses to budge.
Strong collaboration: Because no one builds a robot brain alone.

The conversation also shines a light on roles that rarely get attention. Infrastructure engineers, along with the people who build data and deployment pipelines, quietly make everything else possible. Nikita’s own leadership style has changed as the team has grown past 40 engineers, and he speaks openly about what shifts when a small group of researchers becomes a real company.

Hype, Honesty, and Advice for Founders

Robotics gets a lot of hype, and Nikita does not wave it away. He sees a healthy side to it, as long as expectations and timelines stay honest. Where he sees risk is in over-promising, especially around consumer and home robots, where the gap between a demo and a dependable product can be wide.

His advice for founders is practical. Most people underestimate three things: how big the team needs to be, how long the timeline will run, and how many operational realities come with building a company. If you are planning a robotics startup, those three lines alone are worth listening to.

Key Takeaways for Robotics Teams

If you only have a minute, here is what stands out from the conversation:

Demos are not deployments, and the hidden human effort behind a demo is the real cost to watch.
Reinforcement learning helps robots recover from failure instead of falling apart.
Simulation and real-world data work best together, so plan for both.
A horizontal platform could speed up the whole robotics ecosystem.
Honest timelines build more trust than bold promises.

Why This Episode Is Worth Your Time

If you are building robots, investing in them, or trying to understand how research breakthroughs become real-world impact, this episode delivers a grounded and forward-looking view. Nikita blends deep technical knowledge with a refreshingly straight take on timelines, teams, and trade-offs.

It also shows what makes Machine Minds stand out. Greg Toroosian asks questions that reach past the headlines and into the decisions founders actually face. The result feels less like a product pitch and more like a candid chat between two people who care about getting robotics right.

About Your Host: Greg Toroosian and Samson Rose

Machine Minds is hosted by Greg Toroosian, CEO of Samson Rose. Greg spends his days inside the robotics and hard tech world, helping companies find the engineers and leaders who turn ambitious ideas into working machines. That vantage point shows in every episode. He understands the technology, the people behind it, and the hiring challenges that come with growth.

If your team is scaling and you need robotics talent, Samson Rose is worth a conversation. And if you want more founder stories like this one, follow Machine Minds on your favorite podcast app.

Deepen Your Understanding of Robot Learning and Intelligence

Explore More from Machine Minds and Flexion Robotics:

Listen to the Full Discussion: Head over to Spotify or Apple Podcasts for Building the Brain Behind the Next Generation of Robots with Nikita Rudin.
Analyze the Technical Architecture: Visit the official website of Flexion Robotics to explore their robot learning approach, simulation work, and company updates.
Connect with the Speaker: To follow insights on reinforcement learning, humanoid robotics, and building general-purpose robot intelligence, look up Nikita Rudin on LinkedIn.

Building the Brain Behind the Next Generation of Robots: Key Lessons from Nikita Rudin