Fanatical love for products like the Tesla Model Y often stems from simplicity rather than perfection. There is no such thing as a perfect product. Instead, there are simple products and complex ones. When a product becomes too complex, its core purpose gets lost. For example, a platform like Facebook can be confusing because it tries to offer news, reels, stories, and chats all at once. For a product to succeed, its purpose must be undeniable.

What I've learned over my career is that there is no such thing as a perfect product. But there is a simple product and a complex product. Purpose of the product has to be clear. When you look at a T-shirt, you know it is a T-shirt. When you look at a watch, you know it is a watch.

Mehul applied this philosophy when building Matic, which makes autonomous robot vacuum cleaners. He believes that a product should have an unmistakable identity. He wanted to design a robot that actually looked like a robot so its function was obvious. This type of intentionality and effort is why certain products develop a fanatical following.

Innovation is a function of the number of iterations and the speed at which you move between them. Mehul finds deep inspiration in the craftsmanship of Apple products from the early 2000s. He felt a sense of magic when using the first iPod and Macintosh. This attention to detail suggests that product design should be treated as an art form rather than just a utility. Great artists like Picasso were perpetually dissatisfied with their work. This same pride and drive for perfection should apply to building physical products.

Why do artists get to have a pride in their products? Some of these painters were never happy with their work. They were perpetually dissatisfied. Why are we not creating our own artwork through this product? Ultimately, make something wonderful. That wonderful itself is a storytelling.

While many startups follow the school of thought to ship fast, Mehul spent six years iterating before launching. This approach is necessary when building a minimum lovable product in an existing market. For a new product category, a simple prototype might work. However, when entering a market where customers have preconceived notions, the product must meet and then exceed those expectations. An electric car cannot just have four wheels. It needs a starter and a windshield to be taken seriously.

Current indoor robotics often lack true intelligence. Early robot vacuums functioned like a blindfolded person bumping into walls to map a room. Modern lidar robots are like people walking with one hand extended. They have many sensors but lack a brain to process the context of the environment. Mehul aims to give robots a visual cortex to provide a true understanding of the 3D world.

There are products that touch users and then there are products that users touch. We want to build the former, not the latter.

There is a common misconception that people want to buy robots or AI. In reality, these technologies are just tools. People want solutions to their specific problems, such as writing a better email or creating a video. Mehul notes that building a robot like R2D2 is often a mistake. Most people already have the functionality of such a robot in their iPhone. The focus should always be on the problem being solved rather than the technology itself.

No one actually wants robots. This idea that you want to buy a robot is a misnomer. It is false. People want solutions to their problems. And if you happen to solve that problem using robots because robots are the right way of doing it, great. AI is just a means to an end, in the same way robots are a means to an end.

Mehul recalls advice from Kevin Systrom of Instagram, who said that the most rejected piece of advice for entrepreneurs is to solve a problem. This happens because engineers and founders are often excited by elegant solutions and cool technology. Mehul experienced this with his first company, Flutter, which focused on gesture detection. Paul Graham from Y Combinator frequently reminded the team that they were a technology looking for a problem. It took over two years for the team to admit that gestures were not a product that people actively look to buy.

Paul Graham would always come to us and say, I know you guys. You guys are technology looking for a problem to solve. Have you found one yet? It took us two and a half years to grudgingly agree that what we were building was really cool technology but it is not a product because no one wakes up in the morning and says, today I am going to buy gestures.

A real product solves a tangible need, like a robot vacuum cleaner that keeps a floor clean. While having capital from a previous success makes it easier to think long term, the core challenge for any founder remains identifying a genuine problem to solve.

Starting a company is an iterative process that requires hands-on experience rather than just theoretical study. Learning to be a founder is much like learning to swim or ski. You can read every essay and watch every video, but you eventually have to jump into the water. Success comes from building muscle memory through the struggle of actually doing the work.

Starting a company is akin to trying to learn how to swim or how to ski. It doesn't matter how many videos you watch, it doesn't matter how much you learn. Theoretically, you gotta jump in that pool and try to swim and that's gonna be hard and you gotta struggle through it and you gotta iterate in the same exact way.

Mehul and his co-founder Navneet gained significant experience at Light.com, an early computer vision startup. While they were ahead of the curve, those years provided the foundation for their later work with Flutter. They eventually realized that Flutter was a technology looking for a platform rather than a standalone product. This insight led them to accept an acquisition by Google. This move allowed them to reflect on what they still needed to learn about the business world.

One major realization was the importance of building products that customers actually pay for. Mehul decided that his future projects must involve a direct payment model. Additionally, they realized the limitations of software-only solutions in computer vision. Because they did not control the hardware, like the frame rates or autofocus of webcams, their algorithms were like a brain trying to work with a failing eye. This led them to transition to Nest to learn the complexities of hardware and how to build physical products that people value enough to buy.

The early environment at Y Combinator was scrappy and authentic. When Mehul first applied with Flutter, the startup was rejected at the application stage. Following a personal rule of never taking the first no for an answer, he guessed various email addresses for Paul until he found the right one. Paul initially felt the technology was a solution looking for a problem, but a well-timed mention of support from John Collison at Stripe changed the trajectory.

I have this rule of thumb that I don't take no as an answer for the first time. I'll at least try again. So I guessed a bunch of Paul's email address possibilities and sent him an email and I got one of them right.

This persistence led to an interview and eventual acceptance into the program. Interestingly, Paul later forgot he had ever rejected the team. This realization led him to wonder how many promising startups are accidentally overlooked during the massive selection process.

He said, I wonder how many babies we throw out with the bathwater. It was just this idea and I remember thinking, wait a minute, he has no recollection.

At Nest, the environment was incredibly energetic. During the transition from Google to Alphabet, there was a palpable excitement about the work being done. However, this period also highlighted why a vast majority of acquisitions fail. Statistics suggest that nearly 98% of acquisitions do not succeed, and the primary reason is rarely the product or strategy. Instead, failures usually stem from cultural mismatches. Nest functioned as a mini Apple, characterized by a secretive, top-down culture with strong ownership. In contrast, Google operated with a more democratic, bottom-up approach. These two styles were like oil and water.

Based on how people behaved, I could tell whether they joined Nest pre acquisition or post acquisition, because if they were pre acquisition, they were just absolutely, absolutely devoted to Nest. There was just the sense of pride and ownership and thinking that it was just phenomenal.

The difference between employee groups was stark. Those who joined before the acquisition were devoted to the specific Nest philosophy. New hires who arrived after the purchase often identified more with Google culture and questioned whether they wanted to buy into Tony Fadell's leadership style. This cultural friction makes it difficult for a startup to maintain its soul after being bought. For those who built these products, it is painful to see them languish or get discontinued under new ownership. Even years later, the quality of that original work shines through. Mehul notes that people still search for old units of the Nest Protect smoke alarm on Reddit because they love the product so much, even as Google moves away from it.

In 2017, the world was filled with hundreds of self driving car startups. Mehul noticed that while many teams were trying to solve autonomous driving, nobody was building a home robot that could navigate without bumping into furniture. The existing robot vacuum market was worth billions, yet the products were underwhelming. Many had a negative net promoter score because customers were actually warning others not to buy them. Mehul saw an opportunity to apply the concepts of level five autonomous driving to the home.

If level five cars means that cars drive like humans, then level five robots must be that they behave like humans, they navigate like humans, they clean like humans, they manipulate objects like humans.

To achieve this, the team decided to move away from a traditional Christmas tree of sensors and focus on a vision based system. They wanted to give the robot actual eyes and a brain. Mehul and his team aimed to build the Apple of home robotics by creating a product that earned consumer trust through performance. However, the journey was much longer than anticipated. While they expected to ship by 2020, they did not ship until 2024. This delay was caused by many unknown factors and the need to build their own tools, as the infrastructure provided by companies like NVIDIA did not exist yet for home robotics.

Before building the technology, the team focused on a problem deck rather than a solution deck. Mehul identified that homeowners want a perpetually clean environment with the least amount of cognitive load. Most existing robots were dumb and required constant babysitting to ensure they did not get stuck on wires or damage expensive rugs. One early experience with a Dyson robot vacuum served as a turning point. The machine had high suction but lacked the intelligence to realize it was destroying a rug patch.

All these robots are just, for lack of a better way of saying it, are just dumb. And how do we actually build an intelligent robot, the one that just works, that you don't have to pre clean for that you can just trust and you don't have to babysit?

The goal became creating a robot that required no pre cleaning or supervision. This required a shift in perspective. Instead of adding more sensors, they simplified the hardware and focused on the AI and perception systems needed to understand the context of a home.

The common disc shape for robotic vacuums is not the best design for cleaning. While many robots are circular, this shape makes it hard to reach corners and sides. Most manual vacuums are not round for this reason. On a circular robot, the actual vacuum part is often two inches away from the wheels. This means the cleaning area is small and the robot misses many spots.

If circle or disc shape was the right shape, all the manual vacuums would have shipped that way. But disc shape, by definition, is very bad in terms of going to sides and corners.

Mehul believes that a robot must solve the problem of cleaning first. If a robot cannot reach the corners, it is not doing its job well. The focus should be on efficacy to ensure the product provides real value to the user.

Mehul explains that people do not want robots just for the sake of technology. They want solutions to specific problems. When designing a home robot, it is important to consider how children and pets react to it. Many existing disc robots can be scary for families. Homeowners often try to hide ugly appliances behind cabinets. A robot should look like it belongs in the home rather than being an eyesore in the living room.

The purpose of the product has to be clear when you use it. When you look at your T-shirt, you know it is a T-shirt. When you look at a watch, you know it is a watch. But when you look at a disc robot, you are like, wait a minute, is this a vacuum? Is this a speaker? What is it?

Mehul learned a valuable lesson from Tony Fadell and Matt Rogers at Nest. They taught him that a product must have a clear identity. If a robot were sent back to the 1960s, people should be able to guess its function. The team went through many iterations to find the right balance. They tested different wheel configurations and ensured the robot was safe if a child ran into it. The goal was to create a friendly machine that people could get used to easily.

Many people associate the loudness of a vacuum with its suction power. Because of this, companies sometimes artificially inflate the volume of their machines to make them sound more effective. However, noise has very little to do with actual cleaning performance. Real suction power comes from airflow rather than volume. Vacuums were originally designed as carpet sweepers for soft surfaces, but the cleaning needs of a home have changed.

Turns out actually noise has nothing to do with suction. Power has nothing to do with it. It actually is about airflow.

Mehul explains that cleaning hard surfaces requires agitation rather than just raw suction. He uses the analogy of fine dust on a car. No matter how fast you drive, the dust stays on the vehicle because of the way air moves over it. But if you gently nudge the dust with a finger, it comes off immediately. Hard surfaces need a brush roll to scrape and agitate the dirt. Once the dirt is loose, the machine only needs enough suction to transport it to the bin.

The amount of suction power that Dyson or some of these other vacuum companies have is great if you have sandy carpets. But most of us don't live near beach and we don't have sandy carpets. So just simple amount of brush roll is great.

Most current vacuums use preset settings that do not account for different floor types like shag rugs or delicate carpets. A more effective approach is to build robots that dynamically adjust their suction power and brush roll speed based on the specific floor surface. This allows the machine to clean efficiently by moving up and down to the right height instead of simply squishing the debris.

Mehul draws inspiration from established vacuum brands like Miele and Dyson. Miele is known for its quiet airflow and gentle treatment of floors, while Dyson excels in branding and suction. However, the goal is to create a product that people love with the same fanaticism they feel for a Tesla. Achieving this requires moving beyond just cleaning well and focusing on how the product integrates into the home environment.

Why do people love their Model Y or Model 3 or Model S in an absolutely fanatical way? And how do we do that and how do we create that bond with the product?

A central concept in this approach is continuous cleaning versus batch processing. Most appliances in the United States, like dishwashers and washing machines, are designed to process large batches. This requires users to wait until a machine is full before running it. During that waiting period, the home collects mess, which Mehul describes as entropy. This is especially challenging for busy parents because kids and pets act as tornadoes of entropy.

Every single appliance built in the United States and the First World are built as a batch processor. You have to collect your dirty dishes until the dishwasher can be full. You collect laundry until it is completely full. Because you are trying to make sure that you are using it efficiently, you are collecting entropy.

Mehul points to his experience growing up in India as a model for a different way of living. In India, domestic help often cleans dishes after every meal and mops floors twice a day. This keeps the home in a perpetual state of cleanliness. Robots are the perfect tool to replicate this because they do not get bored or tired. By cleaning continuously rather than once a day, a robot can maintain a sense of zen in the home. The vision for this was inspired by a scene in the movie Passengers where a small robot immediately cleans up spilled cereal.

The journey from a wooden student project to a functional robot involved significant iteration and a steep learning curve. Mehul and his team started with a background in computer vision rather than robotics, leading them to build early prototypes simply to see if they could make a robot work. These early versions allowed them to test features like quieter operation, cyclone vacuums, and navigation using only two cameras. They even explored interaction methods like voice and gestures, proving to themselves that these features were possible before focusing on scaling them for a final product.

We just sort of built it that way and see if we can make it even work. And then we were like, okay, we know how to do it. So the next three prototypes were some of the proof of concept.

User research revealed that the traditional goal of cleaning an entire home did not match how people actually live. Most homeowners ignore low-traffic areas like attics or bedrooms on a daily basis and focus instead on high-traffic zones like the kitchen and dining room. This insight shifted the focus toward a robot that could handle daily chores in these messy areas. Additionally, the team realized that modern homes have moved away from wall-to-wall carpets toward hard surfaces and thick rugs, requiring a complete redesign of the mobility systems and wheels used in older vacuum models.

Testing moved from reading history and reviews to conducting 50 different home demos. Mehul eventually placed early units with families to observe real-world usage. Even though these early versions were often unreliable in low light or frequently got stuck, a subset of users found them so valuable that they refused to return them. This desire to keep a buggy prototype proved the product was solving a genuine pain point for busy parents who wanted to reclaim time spent on daily chores.

Some people obviously stopped using it. But then there were a set of customers who just kept using it in spite of it being so bad. They just figured out when it works extremely well, and they kept using it for a year and in many cases even refused to return the robot.

A product does not need to be perfect to find success if the problem it solves is intense enough. Early versions of the robot were 3D printed and functional only about half the time. Despite this, users loved the solution because the underlying need was so strong. This validation allowed the team to focus on rapid iteration rather than aesthetic perfection in the early stages.

Speed and cost were the primary drivers for choosing 3D printing over traditional manufacturing. Injection molding parts or sourcing them from overseas is expensive and slow. By keeping production local and using modern 3D printers, the team could iterate through 250 different versions quickly. Mehul notes that this approach saved both time and money by avoiding the delays inherent in global supply chains.

The reason we did it was for two reasons. If you try to injection mold any of the parts, it is expensive. Also, because we do not have a good manufacturing industry left in the United States, a lot of times you have to go outside of the country. All of that takes time.

This lean approach allowed the company to reach a mature stage with a much smaller team and lower budget than typical robotics startups. While similar companies often hire 300 people and raise massive amounts of capital, this team remained vertically integrated with only 70 people. They spent roughly 15 million dollars to reach their current milestone, which is a fraction of the industry standard.

Establishing a global supply chain involves more than just finding a factory. Mehul visited China in 2023 to begin spinning up production, and the company now maintains dedicated teams in Hong Kong and Taiwan. The shift toward injection molding made looking at international suppliers a necessity. Even when trying to work with domestic vendors in the United States, those vendors often source their own components from China or other parts of Asia.

Once we got to injection molding, it just made sense. Even if you try to do it from the US, sometimes you get vendors who themselves get parts from China or different parts of Asia.

While the initial focus was heavily on China, the strategy has evolved over the last year. Production is now being diversified across several different countries, including Vietnam, Taiwan, and Malaysia. There is also an ongoing effort to spin up manufacturing capabilities in Mexico to further distribute the supply chain footprint.

Building a new type of robot requires a balance between custom engineering and standard parts. Mehul explains that they prioritize off-the-shelf components like motors, batteries, cameras, and Nvidia GPUs whenever possible. This strategy allows the team to focus their energy on the unique physical design and plastics of the robot rather than reinventing basic sensors. Even with these standard parts, the journey to a functional product is long.

Success in a lab environment is very different from success in a real home. Mehul mentions the importance of testing with family members. If people who are biased to love you are not happy with the cleaning performance, then the product has failed. One major technical milestone was watching the robot navigate through a complex dining area. Weaving around chairs without bumping into them felt like a human-like achievement.

The moment I remember the most is the first time I got a bag full. It was a really exciting thing that not only had I used this robot, but it cleaned enough that our bag was full. I showed it to the entire team and took a picture. It was proof that the robot works.

It took six years of development just to get that first full bag of dirt. Since that breakthrough, the scale of operation has grown immensely. The robots have now cleaned over 100 million square feet in total. On average, users run their robots seven times a week. This means most people are cleaning their entire homes every single day. In total, the robots have driven about 75,000 miles inside customer homes.

The concept of an eleven star experience involves pushing past perfection to design something almost impossible that truly delights customers. For home robotics, this means a robot that completely removes the burden of floor cleaning. The robot should tell you when it has everything under control or precisely what it needs to finish the job. The goal is to eliminate the cognitive load entirely so you never have to think about cleaning again.

The eleven star experience that we are trying to drive towards is the robot comes into your home and says, hey, I got it. You never have to worry about it from this moment onwards. If it turns out that there is a stain that I cannot clean, I will tell you about it.

Mehul notes that humans are naturally a creative species rather than a repetitive one. Giving people their time and energy back results in much higher productivity. History shows this pattern clearly. Before the industrial revolution, productivity was flat. Technology like planes and cars shrunk the distance and gave people time back. In 1920, the average person traveled only 30 miles from home. Today, technology enables us to traverse the globe with ease.

By shrinking time or giving it back, technology allows people to pursue their dreams. Whether someone wants to be a better parent, an artist, or a piano player, removing mundane tasks enables that growth. The ultimate hope is to empower customers to become the people they want to be.

The six-year period spent tinkering on a product before it shipped was marked by moments of intense uncertainty. One of the most painful periods began in March 2020 with the onset of the pandemic. At that time, the company had not raised significant funding and relied on bootstrapping and angel investors. The sudden shutdown raised existential questions about whether the business could survive a year of inactivity.

The question was either you go back to the company and wear mask and work that way, and we did that for about a year and make progress or you die.

While some engineers tried to maintain momentum by taking 3D printers home to iterate on designs, the team found that they were not making enough progress remotely. By June 2020, based on early literature about safety protocols, Mehul and his team made the decision to return to the office. They spent a year working in masks because physical proximity was the only way to ensure the survival of the company.

Mehul initially assumed that simultaneous localization and mapping, or SLAM, was a solved problem in robotics. He expected to find open source libraries that would work perfectly. Instead, his team spent over a year discovering that these tools were only eighty percent accurate at best. This realization forced them to spend three years building their own system from scratch using neural networks and classical ideas.

I believe our slam system at the moment is order of magnitude better than anything out there. And that was just pure grind of three years.

This breakthrough is similar to the transition from early touchscreens to the iPhone. Before the iPhone, touch interfaces required heavy pressure or styluses. The iPhone made the experience silky and natural. Most robot vacuums today are like those early touchscreens. They are frustrating and lack the seamless precision required to be truly useful. This is largely because most indoor robots lack contextual awareness.

Standard robot vacuums use relative maps based on their docking station. If the robot does not start at its dock, it gets lost. This is known as the kidnapped robot problem. Mehul wanted to build absolute maps. A robot should know its location immediately, even if it starts from a random spot. It should work like a human who recognizes a room even if they were brought there blindfolded.

That was always a wrong way of doing it because that's like saying that I can only navigate my home if I enter through front door. Like I came to Alaska and I can't navigate if I don't understand where my house is in Alaska.

The hardware for these systems is deterministic and follows known patterns. However, the software for a vision-only system is indeterministic and much harder to predict. While adding more sensors like LiDAR might make the software development faster, it can make the final product too expensive. Mehul focused on making the business work by perfecting vision-only technology, similar to the path taken by Tesla.

The Roomba was not the first robot vacuum on the market. An earlier product from Electrolux failed because it cost $1,400. To succeed, Roomba aimed for a price below $200. This specific threshold is a psychological barrier. When a gadget costs less than $200, it often does not require a long discussion or permission from a partner to purchase. It is considered toy money. This accessibility is vital because almost no common consumer electronics cost more than $2,000. Above that price, the market shrinks significantly as products become tools for professionals rather than general consumers.

If you're building a robot and if it's not priced cheaply, it's game over. There is literally zero ubiquitous consumer electronics device that is priced higher than $2,000. Beyond $2,000, your market shrinks by over 10x.

Beyond pricing, hardware complexity is a major hurdle. Mehul mentions a rule of thumb learned from the founders of Nest: every single sensor added to a device requires three permanent software engineers on the team. Adding sensors increases calibration needs, supply chain complexity, and failure points. To keep products economically viable, companies like Tesla and Mehul chose vision only robotics. They used RGB cameras and moved the complexity into software instead of adding more hardware hardware.

When building a new product, it is better to solve small, constrained problems first. Instead of trying to build a perfect robot immediately, the focus was on simple milestones. The team asked if the robot could clean a single rug without obstacles or map a home manually. When the robot first shipped in November of 2024, it could not even clean the edges of a room or reach under kitchen cabinets. These features were added later through software updates. This approach follows the logic of making every detail perfect while minimizing the total number of details.

It is the idea that you are trying to climb Mount Everest. What is Basecamp one? What is Basecamp two? Can you define those milestones and can you hit those along the way so you know that you are making progress?

Shipping a product often involves a tension between moving too early and waiting too long. After six years of development, the pressure to release the robot became immense. Technical shifts, such as moving from an Umbrella system on a chip to Nvidia, added six months of delays. Customers who had already paid were questioning if the product would ever arrive. To manage this, Mehul and his team took a direct approach. They contacted their existing customer base and were honest about what the robot could and could not do yet.

We went into customers and we were very direct and very honest that this works and this doesn't work. And if you really want care about what doesn't work, please don't be an early adopter. But if you want to be early adopter and you just want to try it out and you want to take it as it is, we'd be there to support you.

Many customers chose to receive the robot despite its current limitations. These early adopters often understand the complexity of hardware challenges. They provide valuable feedback and are willing to wait for software updates to fix known issues. Finding customers who want to be part of the community and help innovate allows a company to iterate in the real world. Setting clear expectations transforms customers into partners in the development process.

In N Out serves as a powerful example of product consistency. For eighty years, the company has avoided changing its menu or process and rarely uses advertising. Despite this, its stores generate ten times the revenue of a typical McDonald's. This success comes from a simple principle. They make a promise and meet it consistently. Many products fail because they over promise and under deliver. Maintaining a reliable experience over decades is significantly harder than it looks.

Sometimes it is just making a promise and meeting it because most products don't. Most products over promise and under deliver. So if you can just make a promise and deliver it in a meaningful way consistently, that's good enough.

Trust can be destroyed through small, incremental changes. Chipotle serves as a cautionary tale where customers noticed portion sizes getting smaller over time. Mehul points out that customers are smart and observe these changes. When a brand erodes trust in even a small way, it creates a negative backlash. This is often described as death by a thousand cuts. Consistency in the user experience is vital because it preserves the value already built with the customer.

Simplicity is the goal, not perfection. And you shouldn't keep designing things or reinventing the wheel to get to that sort of newness.

Simplicity is much harder to maintain than complexity. Many tech companies suffer from an arrogance of design where they change interfaces just for the sake of novelty. These updates often confuse users who have already learned how to use the device. Mehul notes that apps like WhatsApp succeed because their core purpose remains clear. In contrast, platforms like Facebook and Instagram have become overwhelming because they include too many competing features like stories, Reels, and news feeds. A simple product stays focused on its original purpose.

Pavel Durov manages Telegram with a team of only thirty people. Even with a billion users and a billion dollars in revenue, the company stays lean and remote. They find talent through difficult coding challenges on a dedicated website. This small team size acts as a powerful constraint. While companies like Facebook have over 100,000 employees adding endless features, Telegram relies on a single designer. WhatsApp and Instagram also started with very small teams. WhatsApp had only 48 people when it was acquired for $19 billion.

Constraints are great and constraints forces you to think. So when you only have 30 people like Telegram, you would say, is this feature going to help 90% of my users? If the answer is no, do not build it. But when you have 3,000 people, you say 300 million people will use it. That is still just 30% of your billion users. You can frame the number to make it feel important when it really is not.

Maintaining this level of discipline is difficult. Founders often serve as the soul of a company. When a founder leaves or loses control, the product experience usually changes. Mehul explains that to keep things simple, you must treat the company itself as a product. This involves constant iteration and repeating the core values of simplicity to the team. Mehul points to Netflix as a model company that has successfully maintained this discipline over time. He suggests using metrics like profitability per employee to measure this type of efficiency.

In 2018, an analysis of large public companies revealed that Netflix had the highest revenue per employee. While tech giants like Apple, Google, and Microsoft employ over 200,000 people, Netflix remains small by comparison with only about 12,000 staff members. This lean structure is highly intentional. Mehul notes that the entire company functions with only about 50 product managers. This focus on discipline allows the organization to maintain high efficiency even as it scales.

There are only about 50 product managers in the entire Netflix. So that discipline is really, really clean.

Simplicity requires active maintenance and a willingness to remove old features entirely. When Netflix decided to end its free trial program, they did not just hide the feature or comment out the code. Instead, they assigned a team to delete every trace of that feature from their system. This prevents the kind of code bloat seen in products like Windows, which contains billions of lines of code. Actively deleting rather than archiving is a difficult but necessary step for any company that wants to control complexity and remain efficient over time.

By deleting, you're simplifying. When you just archive or put it in a code, you're not simplifying it.

Mehul found his sense of intentionality through movies. As a young movie buff, he would watch films multiple times to analyze their craft. He would try to identify which scenes could be cut from a story. In the best films, he found that every single scene was essential. This focus on detail eventually led him to notice how editing can disrupt a narrative.

I would sit there and watch movies and say, which scene would I cut? I had watched that movie five times already. I would sit there thinking about which one to cut and I wouldn't cut any of them.

This perspective led to a critique of the Star Wars films. While those movies were groundbreaking for the 1970s, Mehul believes they are poorly edited. The constant use of transitional shots, like rockets flying by, breaks the flow of the story. It makes the narrative feel like a series of short episodes rather than a continuous link. This early interest in cinematic intentionality transitioned into product design when he moved to Silicon Valley.

Apple and Pixar served as major influences. Mehul was specifically fascinated by Pixar's ability to produce a long string of high quality movies. Maintaining that level of excellence across ten or fifteen films is incredibly difficult. He realized that achieving consistent hits requires immense intentionality and a commitment to the craft of the experience.

How do you come up with hits after hits after hits? It is really, really hard to do that. There was so much intentionality in every single movie.

A great story or product often builds depth over a long period. J.K. Rowling spent 15 years on the Harry Potter series, weaving clues into early books that only made sense years later. This kind of intentionality is fascinating when applied to building a company. Instead of looking for a quick exit, the goal is to create a multi-decade journey where every new release adds depth to the overall mission. This approach relies on the power of compounding, which is often called the eighth wonder of the world. While many startups look for shortcuts, the most successful companies, like Stripe or Gusto, succeed because they stay the course and compound their progress over time.

Anything worth building over time requires patience and building compounding. We realized in 2017 that we were never going to sell this company. The goal was to build products that give people their time and energy back.

The mission focuses on solving a massive problem for families. Research shows that families in the Western world spend 45 to 60 hours every week on home chores. This is essentially a full-time job that people do in small increments throughout the day. By building products that automate these tasks, a company can give people their lives back. The strategy for doing this follows the Tesla model rather than the Waymo model. Instead of spending decades on research without a product, it is better to ship useful products to customers immediately. This generates revenue and data while building the trust needed to release even more advanced products in the future. The ultimate test of technology is whether it works for a customer, not whether it looks good in a research paper.

The decision to ship a product that is not fully ready often comes from a simple necessity. For a startup, the choice is binary: you ship or you die. If a company stops moving, it faces a natural decline where meaning and brand value disappear. Innovation is fundamentally based on survival. Mehul explains that by late 2024, the team realized they had no future if they did not get the product into customers' hands immediately.

In November of 2024, we had no choice but to ship. I had come to the conclusion that if we do not ship this year, we do not have a future. We may not survive as a company. Either you ship or you die. This is the beauty of a startup. It is very binary. You do things with less resources or you die. You innovate or you die.

Internally, teams are often the harshest critics. They spend all day looking at problems and trying to improve them. This focus can make the product seem worse to the creators than it does to the users. To overcome initial apprehension from customers, the team focused on making the robot feel like a friend within thirty seconds. They designed it to roll itself out of the box and greet the family by name. They even included stickers. When a child personalizes the robot with stickers, they build an instant bond, much like naming a new pet.

Gathering honest feedback is a challenge because people tend to be polite in person. Mehul argues that a single piece of negative feedback is worth a hundred positive ones. They focus specifically on feedback from paying customers because those users have high expectations. If a customer pays for a product and decides to keep it rather than return it, that is the most honest indicator of success. While selling a company brings a sense of relief, the real joy comes from seeing users love a futuristic product that the team poured their hearts into.

Mehul highlights a significant milestone of saving 12,000 hours of labor for customers. This success followed a difficult period where early product launches failed to gain attention. This persistence mirrors the history of Airbnb, which launched many times before finding its footing. The goal is to build a vision only robot that functions like full self driving for the home.

We tried to launch in 2023 and no one gave a shit. We launched again in November of 2023 and no one cared. If no one pays attention, you can just keep on launching. It was good to finally see that we are doing something truly innovative.

The strategy for the next two decades focuses on solving intense customer problems rather than just building robots. Mehul believes in working backward from a pain point, like toy cleaning or shoe organization. This differs from a robot first approach. While a humanoid robot like Tesla Optimus is impressive, its specific daily purpose for a homeowner is not yet clear.

What we build as a product is just a means to an end. We start with the problem first and work backwards versus starting with robots and working backwards.

A surprising lesson from development is that trivial tasks require much higher precision. People are happy to collaborate with AI for complex tasks like coding or video production because those skills are hard to learn. However, because everyone knows how to vacuum perfectly, the bar for accuracy is higher. Customers want to delegate simple tasks entirely. If a robot leaves a single piece of popcorn behind, the user loses trust immediately.

If the tasks are simpler, people just want to delegate. They don't want to collaborate. And if you're delegating, the bar for accuracy is much higher. We get email if a single popcorn is left behind.

The future likely includes humanoid robots, but the timeline is uncertain. Startups must find ways to provide value and survive until the technology fully matures.

Timing is everything for a startup. Mehul Nariyawala explains that his team could not have built their current product in 2012 because the necessary technology simply did not exist. The compute power, 3D printers, and even the Rust programming language were not ready for high-level use until much later. This highlights a common pattern where infrastructure must be in place before a product experience can truly succeed. For example, Netflix always wanted to be a streaming service, but the internet of 1997 forced them to start as a DVD company until broadband became ubiquitous.

Infrastructure has to be around for you to be able to deliver that product experience that we want. Netflix and Reed Hastings never wanted to build a DVD company. They wanted to build a streaming company. But they could not have done it in 1997.

Every startup is essentially a ticking time bomb. The day the bank account hits zero, the company dies. While raising venture capital adds more time to the clock, it does not actually diffuse the threat. The only way to stop the clock is to become cash flow positive. This is why companies like Tesla and Waymo have been able to lead in self-driving technology. They have massive cash cows, like car sales and search ads, that allow them to fund long-term development without the immediate fear of running out of money.

Mike points out that successful companies also possess a specific muscle memory for shipping. They do not work in total darkness for decades. Instead, they learn by shipping products to small groups of customers and scaling up over time. This shipping DNA is vital for survival. If a startup raises billions of dollars without ever releasing anything, they often fail to develop the internal culture needed to actually deliver a finished product to the world. Transitioning from a research-heavy environment to a shipping culture is incredibly difficult for most organizations.

Product development should always begin with a customer problem. It is easy for engineers to become mesmerized by the possibilities of technology, but the goal is to build something useful rather than just something cool. This requires a constant focus on whether a solution actually solves a problem for the user. One of the most important habits is putting yourself in the shoes of the customer to determine if a feature is simple or unnecessary.

We think of a solution, we put it there, and then we keep removing things to get to the simplicity. And it's like, why do we need three buttons? Can we get away with one? That requires iteration and that requires time and you have to walk away from the product.

Mehul follows a strategy inspired by Steve Jobs where a solution is built and then set aside for a few days. Upon returning, the team asks if every button or feature is absolutely necessary. They peel back the layers of the product until only the essence is left. While it is easy to add an extra option or button to solve a design conflict, it is much harder to do the work required to make a product simply work without them.

Scaling a physical product is often harder than designing the initial version. Mehul mentions the idea that the factory itself is the product. Building one unit is easy and building ten is manageable. However, jumping from one unit to ten thousand is a massive challenge. It requires entirely different systems and tools than a small operation. Matic moved from shipping 300 units in late 2024 to 3,000 units shortly after. The next goal is to reach 60,000 units in a single year.

Navneet always said that if you build one product, it is easy to imagine how we are going to build 10. If you build 10, it is easy to imagine how you will build 100. But going from 1 to 10,000 is hard. That is very hard to imagine.

Manufacturing at scale reveals issues that do not show up during small production runs. One specific crisis involved a Japanese motor supplier that had been reliable for years. Suddenly, 80 percent of the robots started failing noise tests. After a deep investigation, Matic discovered that a sub-supplier changed the glue on a small part of the motor. This tiny change caused friction and noise that ruined the robots. The team had to fly in replacement parts and solve the issue while orders were flooding in from customers.

The motors that have been amazingly reliable for two and a half years of us iterating all of a sudden are just noisy. It turns out, unbeknownst to them, their own supplier had changed the glue that goes on the impeller on the motor, which results in higher friction, which results in higher noise.

Customer service is another critical part of the scaling process. Mehul follows the philosophy that the product is the entire journey. This starts from the first time someone hears about the company until they stop using the device. When a product fails at scale, the response matters most. Matic avoids scripts and templates for customer support. They treat every customer issue individually to avoid the frustration of robotic answers. Mehul still receives email alerts for every customer ticket to stay connected to the user experience.

Mehul stays connected to the customer experience by personally reaching out to everyone who wants to return a product. This direct feedback loop provides a learning opportunity to understand why expectations were not met. In the early stages of a hardware company, reliability testing is an iterative process. By identifying failure points through longevity testing, the team can establish critical checks to ensure the system works as intended.

We're much rather off taken pain and double and triple check everything and handle them and understand what's needed than we not. And then as we understand that there is a reliability, we can relax certain constraints.

In hardware, you must assume things will not work as well as expected. Focusing on the first thousand customers is vital because these individuals become evangelists for life if they have a good experience. If they receive a bad product and the company does not care, the reputation is lost. Unlike software companies that can scale to millions of users in months, hardware is physically limited by supply chains. Parts often must be ordered nearly a year in advance. This creates a forced discipline of compounding growth.

The advantage of hardware is that your supply, if you find a product market fit, you will be supply constrained, which means you are forced to grow deliberately. And that deliberate growth is necessary because every single level of growth uncovers bugs, uncovers issues, and it gives you time to push it.

When demand exceeds supply, the focus shifts. The actual product becomes the factory and the assembly line itself. This slower, deliberate pace allows the team to maintain quality and patience as they scale.

A look at the top companies by market cap reveals that eight or nine of the world's most valuable businesses are hardware companies. From Apple and SpaceX to Nvidia and TSMC, these organizations demonstrate that while compounding takes longer in hardware than in software, the results are often more transcendent. Software companies experience much faster cycles of growth and decline, but hardware moats tend to be significantly deeper.

I think one of the best moats is just the required pain that must be experienced before the thing can work. And if you have seven years of pain as a moat, that is a pretty good moat.

The history of Nest products provides a clear example of how these moats function in the real world. Mehul notes that when the Nest camera launched, it became hyper competitive almost immediately. By late 2015, dozens of new security cameras were entering the market every week. In contrast, the Nest thermostat has faced almost no serious competition even fifteen years after its debut. This is a surprising phenomenon because every household in the country requires a thermostat, and the profit potential is clear. While microeconomic principles suggest that profit attracts competition, the sheer difficulty of executing on hardware can keep a market leader like Nest in a dominant position for over a decade.

Choosing which product to build often comes down to the level of competition and the barriers to entry. While many people focus on security cameras, the market is flooded because they are easy to make. You can buy a camera in China, add some software, and have a product ready quickly. In contrast, products like thermostats and smoke alarms are incredibly difficult because they are tied to old systems and heavy regulations. A thermostat must be compatible with decades of different heating systems, and a smoke alarm has to follow hundreds of pages of rules that change by state and county.

Mehul explains that these unsexy and tedious requirements actually act as a shield. Big companies like Google or Apple often avoid categories that are too messy or unglamorous. Even top tier engineering talent prefers working on prestigious projects like self-driving cars rather than household appliances.

If you graduate with a computer vision PhD from Stanford and tell your mom I am working on self driving cars, she will probably say my son is amazing. But if you tell her you are working on robot vacuums, she will probably ask what is wrong with you.

This lack of social prestige creates a huge opportunity for a startup. Mehul points to iRobot, which generated hundreds of millions in revenue despite a lack of significant innovation for over twenty years. By entering a category that others find boring or too difficult to navigate, a company can build a sustainable business with less competition from big tech or other startups.

Patience can be the most difficult internal struggle for a leader. Mehul finds that his current role at Matic is the longest he has ever held in his career. His previous roles usually lasted about three years due to acquisitions and other changes. Being naturally impatient makes the long-term commitment of building a company particularly challenging. It requires a constant effort to maintain faith and remind the team that progress is happening.

It is really, really hard to learn to be patient. I am by definition very impatient. Just being patient and continuously preaching that we are making progress and keeping that faith was absolutely hard. It is not easy.

This journey also involves the emotional weight of seeing people leave. It is difficult to watch team members walk away when you expected them to stay for the long haul. When individuals lose faith in the mission, it tests the resolve of those leading the organization.