The robots I'm most impressed by right now aren't the humanoid ones. They're the surgical arms that help eye surgeons operate at a scale finer than human hands can manage. They're the warehouse picking systems that have finally, after decades of trying, figured out how to grasp unfamiliar objects reliably. They're the dairy-farm robots that milk cows on the cows' own schedule, gently, without human intervention.
None of these will trend on Twitter. All of them are genuinely changing the work they touch.
What AI brought to robotics
For decades, robots were programmed — every motion, every grip, every branch of logic hand-coded by engineers. It worked for repetitive tasks in controlled environments (welding cars on an assembly line), and nowhere else.
The recent wave of AI has changed two things:
Perception. A robot now knows what it's looking at. Modern vision models let robots identify objects they've never seen, understand scenes, and read instructions from humans. This was the bottleneck for sixty years, and it's mostly gone now.
Learning from demonstration. Instead of programming a robot to do a task, you can now show it the task — through human demonstration, video, or even language — and let it figure out the motions. The first generation of this is promising; the next is going to be transformative.
Put those two capabilities together, and you get robots that can work in environments they weren't specifically programmed for. Which is what we've been waiting for.
Where robots are quietly winning
Warehouses. Amazon, Ocado, and smaller players have built robotic fulfilment systems that now handle a substantial share of e-commerce. The best of these aren't fully autonomous — they combine robots for the structured parts (moving pallets, sorting bins) with humans for the unstructured parts (picking an irregular item off a messy shelf). The balance is shifting slowly, but the hybrid is here to stay for a while.
Surgery. Robotic surgery has gone from a curiosity to a standard option for many procedures. The surgeon is still in the chair, but the robot's precision, tremor-filtering, and ability to work through smaller incisions improve outcomes and recovery times. This is well-established, not cutting-edge.
Agriculture. Robotic weeders that distinguish crops from weeds using computer vision, and remove the weeds mechanically without herbicides, are rapidly growing in orchards and vegetable farms. Robotic milkers, as above, are mainstream in European dairy. Robotic harvesters for delicate crops (strawberries, grapes) are in early commercial deployment.
Logistics and last-mile. Not the sidewalk delivery robots (those are mostly theatre). The indoor delivery robots — in hospitals, hotels, restaurants — are genuinely useful and spreading fast.
Inspection of dangerous environments. Robots that climb wind turbines, swim through oil pipelines, or crawl through nuclear facilities. These save human lives by doing jobs we used to send people to do.
Where robots are still far away
General-purpose household robots. Despite what the marketing suggests, we are not close to a robot that can fold laundry, cook dinner, and clean the bathroom with anything like the reliability of a teenager asked to do the same. The combination of dexterity, common sense, and ability to operate in messy unstructured environments remains very, very hard.
Humanoid robots in production. A lot of capital is flowing into humanoid robotics companies right now. Some of it will pay off. Most of it, in my honest opinion, won't — or will pay off in different markets than the founders imagine. Humanoid form is often the wrong form for the task. A robot that looks like a person is often much less capable than a purpose-built robot for the same job.
Caregiving robots. The idea of robotic caregivers for elderly people, promoted for decades especially in Japan, has not materialised at scale. The social and physical complexity of care work is extraordinary. Technology can augment human caregivers; it cannot replace them, and we shouldn't want it to.
A note on "robots taking jobs"
The framing has always bothered me. Robots don't take jobs. Economic decisions by the humans who deploy robots reshape jobs.
Every serious study of robotic deployment in manufacturing shows that the jobs change — fewer repetitive-strain tasks, more supervision, more maintenance, more higher-skilled work around the automation. The total employment picture depends entirely on what the surrounding policy, training, and investment looks like. A society that handles automation well gets richer and more equal. A society that handles it badly gets neither.
The robots themselves don't care. This is a human choice.
Where I'd invest attention, as an operator
If you're running an operation that might benefit from robotics:
- Start with the task, not the robot. What specific task is expensive, dangerous, or hard to staff? That's your candidate.
- Hybrid is usually the right answer for now. Fully autonomous end-to-end systems are rare. Robot-plus-human workflows are mature and productive.
- The total cost of ownership is higher than the price tag. Integration, maintenance, training. Budget for it.
- Pilot in a corner. One task, one shift, measured against a specific baseline. Expand from what works.
Why I stay optimistic, carefully
The robots that have changed lives so far — the surgeon's assistants, the dairy-barn robots, the warehouse pickers — have done so in a very particular way. They haven't replaced the humans. They've let the humans do work that's more interesting, more skilled, and more valuable.
That's the version of robotics I believe in. If you're thinking about where robots fit into your operation, let's chat. We're fans.
