Issue #043 — Weekend Notes

WEEKEND NOTES Bedrock Robotics’ first operator-free excavator deployments are the perfect cure for my humanoid fatigue.

I’ve been staring at my screen for three days, watching yet another generically named bipedal prototype fold a t-shirt or pour a cup of coffee with the enthusiasm of a hungover intern, and I just had to step away. Don’t get me wrong, I love the hype as much as the next VC looking for the next Nvidia, but there is something profoundly refreshing about a machine that doesn't pretend to be human—it just does a job that humans absolutely hate doing in the mud.

Last week, I ran into the news that Bedrock Robotics has officially kicked off its first operator-free excavator deployments (per The Robot Report). While the rest of the twitterati was debating the knee joint actuation on the latest Unitree or Figure bot, Bedrock was quietly dropping yellow iron into actual dirt.

This feels like the moment we stop talking about "robots" as sci-fi companions and start talking about them as... well, capital equipment.

I’ve been thinking a lot about the split in the industry right now. We have this massive divergence between the "Sexy Tech" crowd—humanoids, dexterous hands, generative world models—and the "Dirty Fingernails" crowd—construction, ag, mining. If you look at the sheer math, the opportunity in the dirt is staggering. The construction industry in the U.S. alone faces a shortage of roughly half a million workers (various industry estimates put this number higher, but let’s just say it’s massive), and it’s not getting better.

The prevailing narrative for the last two years has been that the humanoid form factor is the solution because our world is built for humans. That’s a fair point. Stairs, door handles, and steering wheels are ergonomic for us, bipedal primates. But have you seen a construction site lately? It’s barely ergonomic for humans. It’s a chaotic, dangerous, unstructured mess where the optimal form factor isn't a five-foot-nine android trying to balance on a gravel hill. It’s a tank treads with a giant bucket on the front.

That’s why Bedrock’s deployment hit me differently. They aren't trying to replace the human form; they are trying to replace the human operator's anxiety and back pain. The article notes they are tackling labor shortages with AI-powered heavy machinery. This is the "brains and brawn" argument that Agtonomy’s founder has been shouting from the rooftops (also reported by The Robot Report this week). The thesis is simple: for agricultural and construction equipment makers to stay relevant as workforces dwindle, they need to add AI.

It sounds obvious, but it’s a departure from the "let's build a new species" approach.

We are tracking roughly 600+ robot models across ~350 companies globally in the Maze/FinBrain catalog right now. It feels like 100 of them are humanoids. But when I filter for "construction autonomy," the list thins out considerably. You’ve got the big legacy players trying to bolt tech onto their existing fleets (often with the grace of a dinosaur learning to code), and then you have these startups like Built Robotics, SafeAI, and now Bedrock, treating the machine as a software platform.

What I find most interesting about Bedrock specifically—and this is where I might get a little speculative—is the "operator-free" phrasing. A lot of autonomy in this space is actually "tele-operation 1.0," where a human sits in a trailer somewhere with a joystick, ready to take over if the AI gets confused. That’s a valid bridge technology, but it’s not the promised land. True operator-free means the machine deals with the edge cases. It sees the rock that's too big, it adjusts the bucket angle, it digs the trench, and it moves on. If Bedrock is actually doing this without a human holding its hand in a loop, they’ve cleared a massive technical and psychological hurdle.

Of course, the heavy iron world has its own unique flavor of pain. You can't just iterate hardware at the speed of software. If a humanoid falls over, you pick it up and maybe fix a servo. If an autonomous excavator gets stuck in a mud slide three counties over, your burn rate just got a lot more expensive.

And let’s not forget the geopolitical elephant in the room. The regulatory environment is shifting fast. Just recently, we saw news that the U.S. government is moving to ban new foreign-made humanoids and robot dogs, citing national security risks (per TechCrunch). While that ban is largely targeting Chinese imports—and let’s be honest, China dominates a lot of that manufacturing right now—it highlights a sensitivity around autonomous infrastructure that you don't see with a robot folding laundry.

If a Chinese-made robot dog malfunctions, it’s a surveillance concern. If a foreign-made autonomous excavator with a 3-ton payload malfunctions on critical infrastructure... that’s a different risk profile entirely. It makes me wonder if the "Dirty Fingernails" robotics companies are going to end up being the most protected and subsidized segment of the U.S. market, simply because they are physically tied to the ground we stand on.

There is also a subtle thread here about sensors that I think is worth mentioning, even if it seems tangential. To make these machines work in the dark, in the dust, and in the rain, you need eyes that can handle the real world. IDS Imaging just added a Nion ToF sensor to their portfolio (per The Robot Report), specifically targeting applications where precise 3D data is required at real-world process speeds. Time-of-Flight technology isn't exactly the new shiny toy compared to LIDAR or some of the new visual SLAM approaches, but for heavy machinery, robustness beats fancy every time. You don't need to know the color of the rock; you need to know exactly how far away it is so you don't smash the hydraulic piston.

I suspect the winners in the construction robotics space won't be the ones with the most elegant code or the best PR team. They will be the ones who can survive the vibration, the mud, and the angry site foreman who just wants the ditch dug by 5 PM.

It brings me back to my original thought: fatigue.

I’m tired of the "Uncanny Valley." I’m tired of wondering if a robot has "feelings." I’m ready for the era of "Uncanny Utility." I want to look at a machine and say, "I have no idea how that works, but thank god I don't have to do that job myself."

Bedrock’s excavators might not win a beauty contest. They won't be interviewed on podcasts about the future of consciousness. But if they can actually lay foundation without a human in the cab, they are worth more to the economy right now than a thousand dancing androids.

So, as we head into the weekend, I’m going to keep watching the humanoid space (I have to, it’s my job), but my heart is with the muddy robots. There is something deeply honest about a machine that digs.

I’ll be digging deeper into the unit economics of construction autonomy next week—specifically, whether the hardware margins can survive the brutal construction procurement cycles. If you have any hard data on rental rates for autonomous bulldozers vs. standard ones, you know where to find me.

Stay dirty.

Ken