A prototype can be brilliant and still be useless. Running well on a bench with the designer standing nearby is fine, but falling apart the first time it meets real-world variability can be disastrous. Production is where technology needs to graduate from an idea to a commitment: the same behavior, again and again, without an engineer in the room.
That shift is where manufacturing wins or loses. Parts can arrive late, and minor ergonomics annoyances turn into churning adoption problems. In U.S. factories, weeks matter, and poor handoffs can quickly become expensive, often with customers feeling the fallout.
Clayton Haight, Director of Robotics Hardware at Sunday Robotics, always keeps this in mind. He’s the co-inventor behind Sunday’s wearable training work, including a patented approach centered on time-of-flight sensing, another on wearable data collection. Now he’s focusing on the harder proof: building hardware and processes that hold up in production.
Skill Capture: Putting Ergonomics First
Haight sees a consistent friction point in data collection: many new modalities don’t feel natural at all to use. For example, data collectors using parallel jaw grippers need to invent new ways to manipulate and move through space due to the limitations of the data collection system. But a brand new operator doesn’t pause to explain how much pressure to apply, or the exact angle of a wrist. They make dozens of small corrections on instinct, and having a data collection form that is familiar makes this so much easier.
This is the case for exoskeleton wearables: they can capture motion while work is happening instead of relying on unnatural, out-of-context scenarios. But wearables only help if they feel natural. “If the device is uncomfortable, people will start making unconscious adjustments to compensate for it. At that point, you’re recording a workaround instead of the real motion. Comfort is so important in protecting the integrity of the training signal,” Haight explains.
At Sunday Robotics, Haight led the mechanical design, system integration, firmware, and assisted with manufacturing the patented Skill Capture Glove, a wearable that records how people do common household tasks. The goal was to build something that was comfortable to wear while collecting hours of usable training data, and then make it repeatable enough to manufacture without relying on an engineer to assemble or diagnose every unit. A product that can’t be built consistently, he argues, is more of a research artifact.
Using computer-aided design (CAD) software and 3D printing, Haight was able to run fast cycles—often two to three prototypes per day—then have real users and employees test fit and manipulation capability. If something felt wrong, the feedback went back into the CAD the same day. This fast, iterative loop can be credited with removing the kinds of friction that kill large scale projects once you leave a controlled environment.
Scaling Past the Demo
The same standard carries into robotics. Once you’ve proven something can work, you still have to prove it can live outside the lab. Industrial robotics is already operating at volumes that reward repeatability over cleverness, with annual installations reaching 541,000 units and a global operational stock of roughly 4.3 million. That comes with expectations that each unit will behave consistently and reliably.
Haight brought that mindset into the prototype development of Sunday Robotics Memo robot. He managed engineers, timelines, and suppliers to keep the work moving, using weekly coordination and direct check-ins as risks surfaced. The robot became the company’s main work horse, used internally for research, promotion, and marketing, and externally for product testing. It also marked the first time the company showed the world its product vision and industrial design, raising their bar from “can we demo it?” to “can we run it under pressure?”
And the pressure starts early in the manufacturing. Supply and build readiness are one and the same, Haight says. In 2024, almost 80% of organizations reported supply chain disruption over the prior twelve months, and nearly half experienced disruption due to third-party failures. One component slips, a test gets pushed, an integration window gets missed, and suddenly the schedule is bending around what didn’t show up. If you wait until it’s a big issue, it’s already in the critical path. His approach is to pull those realities forward: keep suppliers in the same operating cadence as engineering, surface what breaks the schedule early, and make changes while there’s still room to do so.
Haight believes demoing was part of what forced Sunday Robotics to think like a robotics manufacturer. “Live demos have a tendency to inspire others” he says. “Everything is out in the open. Any shortcuts become a liability, and a recorded demo tells you something. If you can’t run it in front of people, can you really say it works?”
Reliability Is What Lets More People Build
Reliability, explains Haight, decides whether a robot can become a shared platform or stays a fragile science project that only a few people trust. Reliability also has an economic shadow that’s hard to ignore. Unplanned downtime is estimated to cost the world’s largest companies nearly $1.4 trillion per year—about 11% of their revenues. In robotics, that can show up as stalled work and weaker experimentation and, worse yet, progress limited to whoever has the time to babysit the system.
With their newest platform, Sunday Robotics moved researchers onto a platform that was three times more reliable than the previous platform, with materially better performance. Just as important, the team built 10 times more robots than previously available, which changed how their work was happening day to day. Work stopped bottlenecking around a small number of good units, and more people could iterate and test and contribute without waiting for their turn.
Early prototypes also remained on the path to product certification, with work underway with regulatory agencies. Certification makes ‘good enough’ a moving target, says Haight. “We learned pretty quickly that the supply chain determines your bandwidth for testing and learning speed.”
More stable units made broader internal testing possible without turning every test cycle into a scheduling fight. More units also surfaced inconsistencies faster, which is exactly what Haight wanted: problems that show up early enough to fix, before it’s too late.
Manufacturing as the Robotics Competitive Edge
Robotics is heading into a phase where the market rewards operational execution as much as technical ambition. The global service robotics market is projected to reach about $107.75 billion by 2030 as the technology finds more of a foothold. As that grows, more teams will be able to build impressive prototypes. Fewer, though, will be able to turn those prototypes into something customers can depend on.
Haight keeps a keen eye on the company’s manufacturing pipeline because he recognizes where so many robotics systems stall. Even if all the right technology and product risks are addressed, a weak link in the supply chain could cause everything to fail. For Sunday Robotics, the near-term work is clear: carry what they have proved further, and lock down the processes that make scale possible. That means keeping supplier coordination tight, protecting reliability thresholds as volume grows, and turning prior wins into practices that hold up when the team expands and the schedule becomes less forgiving.
“Manufacturing discipline is really what makes robotics viable,” Haight says. “It’s the difference between something that works in the hands of the people who built it and something that works for everyone else.”
