<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Ant Newman on Medium]]></title>
        <description><![CDATA[Stories by Ant Newman on Medium]]></description>
        <link>https://medium.com/@ant_48348?source=rss-72d24c2910d------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*Dy9R2kF80HMCi4GC</url>
            <title>Stories by Ant Newman on Medium</title>
            <link>https://medium.com/@ant_48348?source=rss-72d24c2910d------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 01:13:32 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ant_48348/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Humanoid or not, physical AI is on the job | Spectro Cloud]]></title>
            <link>https://medium.com/spectro-cloud/humanoid-or-not-physical-ai-is-on-the-job-spectro-cloud-4addc9b9cabd?source=rss-72d24c2910d------2</link>
            <guid isPermaLink="false">https://medium.com/p/4addc9b9cabd</guid>
            <category><![CDATA[robotics]]></category>
            <category><![CDATA[edge-ai]]></category>
            <category><![CDATA[physical-ai]]></category>
            <dc:creator><![CDATA[Ant Newman]]></dc:creator>
            <pubDate>Tue, 20 Oct 2026 11:00:46 GMT</pubDate>
            <atom:updated>2026-10-28T12:30:10.953Z</atom:updated>
            <content:encoded><![CDATA[<p>If you’ve taken a ride in a Waymo, you’ve already put your life in the hands of physical AI. Except, of course, there weren’t any actual hands.</p><p>There <em>was</em> a car. It had to do astonishingly difficult things to get you through traffic and deliver you to the right place, in one piece. Waymo says it gives <a href="https://techcrunch.com/2026/03/27/waymo-skyrocketing-ridership-in-one-chart/">about 500,000 paid rides a week</a>. That’s a lot of people literally trusting their lives to an AI application and then, presumably, sitting doomscrolling on their phones like it’s the most normal thing in the world.</p><p>On the silver screen — since long before the days of <em>Terminator</em>, <em>I, Robot</em>, <em>Ex Machina</em> or my personal favorite <em>The Mitchells vs. the Machines</em> — an intelligent robot almost always gets a human-shaped body to help them take their shot at saving/destroying the world.</p><p>When we see a bipedal machine with a face, we immediately recognize what it is and feel the shiver of sci-fi potential… which helps explain the terrifying appeal of a Tesla Optimus or Boston Dynamics Atlas on a real keynote stage.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*Fr-TuNLADa3JYMpJ.jpeg" /></figure><p>Sending up the humanoid keynote launch: <em>The Mitchells vs. the Machines</em> (2021).</p><p>But the familiarity of the glossy white-and-black humanoid robot has encouraged a peculiarly distorted ideal of judging physical AI progress: we’re always waiting for a convincing mechanical person, while in reality machines that look <em>nothing</em> like us are already doing the real work.</p><p>A warehouse robot doesn’t need arms. A vision-guided agritech sprayer doesn’t need fingers. And if the job is inspecting a roof, why give the robot legs when it could simply fly?</p><p>We define physical AI as any equipment that uses learned models to perceive, decide and <strong><em>act</em></strong> in the world, whatever shape it takes. Building an affordable multipurpose humanoid robot would be an extraordinary way to act on the environment — but it certainly isn’t a prerequisite for a world filled with physical AI.</p><p>In fact, we expect much of the growth in physical AI to come through machines like these: specialized, often simpler and more robust equipment that becomes more capable as its perception and control improve.</p><h3>Eighteen years down the mine</h3><p>In April 2026, Komatsu announced it had <a href="https://www.komatsu.jp/en/newsroom/2026/20260422">commissioned its 1,000th autonomous ultra-class haul truck</a>, a 930E-5AT with a 290-metric-ton payload, at Barrick’s Nevada Gold Mines. Customers using the company’s FrontRunner system had moved more than 11.5 billion metric tons of material since its commercial introduction in 2008.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*lEPP-mTruNdSuRsw.jpeg" /></figure><p>Strictly, FrontRunner is a precedent for today’s physical AI more than an example of it. Its trucks were navigating with precision GPS and radar-and-laser obstacle detection long before anyone trained a robot foundation model: just like a giant version of the original Roomba, which found its way around with bump and cliff sensors rather than learned models. That doesn’t make FrontRunner less impressive. Moving an enormous load through an operating mine, repeatedly and safely, is precisely the kind of dangerous job we <em>should</em> be automating.</p><p>The mine also shows how much engineers can gain by strictly defining a known and knowable operating environment. They can design traffic procedures around a fleet and plan where vehicles operate. A Waymo has to cope with whatever turns up on the road. The mining system can be enormously valuable without ever acquiring the ability to deal with novel situations.</p><p>For a business deciding where to invest, a repetitive job in a manageable environment is the right place to start.</p><h3>A forklift is a pretty good design</h3><p>At Walmart’s distribution center in Brooksville, Florida, FoxBot autonomous forklifts used <a href="https://corporate.walmart.com/news/2024/04/11/a-fork-in-the-road-walmart-bets-on-associates-automation">AI-powered machine vision and dynamic planning</a> to unload pallets and carry them to an automated storage system.</p><p>This is distinctly not the sci-fi approach: there’s been no teaching a general-purpose humanoid to climb into a conventional forklift and operate its controls. Putting intelligence directly into the familiar forklift form-factor avoids a whole additional machine.</p><p>After a 16-month proof of concept, Walmart announced in April 2024 that it was rolling out 19 machines across four distribution centers.</p><p>But scaling has apparently been slow since. There’s been no public word of a bigger Walmart fleet, and when <a href="https://www.therobotreport.com/symbotic-acquires-autonomous-forklift-maker-fox-robotics/">Symbotic bought Fox Robotics</a> in February 2026, its CEO said most of Fox’s roughly 25 customers were still running pilots.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*PRO5U3QV1f5w2xgN.jpeg" /></figure><p>Other logistics operators have been more ambitious about scale, often in tandem with novel greenfield deployments. Amazon last year deployed its <a href="https://www.aboutamazon.com/news/operations/amazon-million-robots-ai-foundation-model">millionth robot, across more than 300 facilities</a>. It has reimagined the way stock is held and moved: a self-contained world built to fit robots, not a robot inserted into traditional logistics. Ocado’s <a href="https://www.ocadogroup.com/about-us/our-technology">smart platform for supermarkets</a> relies on a 3D storage grid that looks unlike a conventional warehouse. Ambition doesn’t guarantee payback, though. In November 2025, <a href="https://www.newstribune.com/news/2025/nov/20/kroger-closing-automated-fulfillment-centers-as/">Kroger said it would close three of the eight automated warehouses</a> it had built with Ocado, because order volumes couldn’t justify their fixed costs.</p><h3>Three seasons of weed control</h3><p>Agritech has been one sector enthusiastically adopting physical AI, literally out in the field, using the adaptability of AI models to handle the messy reality of fruit crops and weeds.</p><p>Researchers at the University of Arkansas System Division of Agriculture studied a prototype of John Deere’s See &amp; Spray system over three soybean-growing seasons. Cameras and machine learning identified weeds so the equipment could apply herbicide selectively. In the trial, <a href="https://www.uaex.uada.edu/media-resources/news/2025/march/03-10-2025-see-and-spray-research.aspx">herbicide use was 43% to 59% lower than with broadcast spraying</a>.</p><p><a href="https://aureaimaging.com/">Aurea Imaging</a>, a Dutch startup, also uses tractor-mounted computer vision in orchards, analyzing each tree in real time so spraying can be targeted. The company pitches it as a way to cut waste and raise yields.</p><h3>From the roof to the kitchen floor</h3><p>Some physical AI equipment is already operating much closer to home. Camera-equipped Roombas, including the j7, use <a href="https://homesupport.irobot.com/articles/en_US/Knowledge/31161">object recognition to avoid hazards such as cords and pet waste</a>. It’s a small domestic ambition (although anyone who’s contemplated the consequences of getting the second part wrong might disagree). It hasn’t been a lucrative one either: iRobot <a href="https://www.sec.gov/Archives/edgar/data/1159167/000115916726000006/irbt-20260122.htm">went through Chapter 11</a> in December 2025 and is now owned by Picea, its Chinese manufacturing partner.</p><p>Above ground, Skydio’s X10 drone uses <a href="https://support.skydio.com/hc/en-us/articles/18919821734683-Skydio-X10-Frequently-Asked-Questions">computer vision and adaptive flight planning</a> to capture inspection data under a human’s high-level direction. The pilot can specify the work while the aircraft handles parts of the flying. A robot needn’t be independent of every human decision to perform some of its work autonomously.</p><p>Each of these machines is shaped by its job. A low disc fits under furniture, a quadcopter reaches a roof, a sprayer is pulled behind a tractor. Engineers can spend their effort on the movement and perception the task needs.</p><h3>Why a factory might pay for legs</h3><p>Specialized machines are everywhere, but humanoids are making real progress too, and the most instructive example so far comes from a car plant.</p><p>At BMW’s Spartanburg plant in South Carolina, the assignment for Figure 02, a humanoid built by California startup Figure AI, was more Bruce Springsteen than Isaac Asimov: ten-hour shifts loading sheet-metal parts into welding fixtures.</p><p>During a ten-month pilot, the humanoid moved more than 90,000 components in roughly 1,250 operating hours and helped support production of more than 30,000 X3 SUVs, <a href="https://www.press.bmwgroup.com/united-kingdom/article/detail/T0455870EN_GB/bmw-group-to-deploy-humanoid-robots-in-production-in-germany-for-the-first-time">according to BMW</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*6MWKvO0u-OGhzNs1.jpeg" /></figure><p>BMW chose a body shop with extensive automation and teams accustomed to integrating new equipment. Production IT, safety and logistics staff were involved early. The robot could do its job because a great many people had already done theirs.</p><p>The company called it a pilot, and its February 2026 account didn’t publish a comparable cost per component or a human-intervention rate — so we can admire the work without pretending we know whether a humanoid was the cheapest way to do it. The hardware has already moved on. Figure <a href="https://www.figure.ai/news/production-at-bmw">began retiring the 02</a> in November 2025, BMW says it’s evaluating the Figure 03, and its first humanoid pilot in Germany, at Leipzig, uses Hexagon’s AEON instead.</p><p>While robot arms have been a fixture in auto factories since the last century, there is a credible reason to bring humanoids into the picture. Factories contain workspaces and tools already designed around people. A machine that can use them, then move between assignments, might avoid expensive brownfield facility reconstruction, or even any line downtime at all. But buyers have to weigh that flexibility against the cost of a complicated robot body. For a repetitive task at a fixed station, a conventional arm may be perfectly adequate.</p><p>At home, the ambition is more seductive. One biped to do the ironing, carry the laundry upstairs, load the dishwasher and clean the bathroom? We’d like that too. But unfortunately for our domestic bliss, these tasks ask far more of a robot than unloading another pallet in a familiar warehouse.</p><h3>Who’s supervising the air fryer?</h3><p>In its April 2026 account of π0.7, Physical Intelligence described a robot <a href="https://www.pi.website/blog/pi07">trying to cook a sweet potato in an unfamiliar air fryer</a>. Left to itself, it made a reasonable attempt but didn’t finish after a few false starts. Step-by-step language coaching got it through, and PI then used that coaching to fine-tune the robot’s high-level policy until it could do the job without help.</p><p>As research, that’s encouraging, but as a household purchase, it’s a tougher sell. How much of dinner are we willing to spend supervising an expensive robot as it operates an appliance we already own? A domestic assistant will have to earn our trust to be left alone.</p><p>Nor does <em>understanding</em> all the steps involved in a task guarantee the ability to carry it out. DeepMind’s July 2026 <a href="https://deepmind.google/blog/gemini-robotics-2-brings-whole-body-intelligence-to-robots/">Gemini Robotics 2 results</a> included 92% success unscrewing a lightbulb, but only 36% screwing one in, in multi-finger tests on Apptronik’s Apollo 2 humanoid with SharpaWave hands.</p><p>DeepMind doesn’t explain the gap, beyond conceding that multi-finger dexterity “remains challenging”. Our guess is that screwing a bulb in demands alignment and contact pressure that unscrewing it doesn’t. But a one-in-three success rate at changing a lightbulb would get any janitor fired, human or robot.</p><p>Air fryers and lightbulbs might be challenging, but some enterprises have had more success in commercial kitchens (with robots that look nothing like chefs, incidentally).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*7VcVwSGzJOJO77wN.jpeg" /></figure><p>In September 2024, Spectro Cloud customer Chipotle <a href="https://www.nrn.com/fast-casual/chipotle-has-moved-its-automation-tests-into-restaurants">moved two collaborative robots into restaurant tests</a> in Southern California: Autocado, built with Vebu, which cuts, cores and peels an avocado every 26 seconds before staff mash it into guacamole, and an augmented makeline from Hyphen that assembles bowls and salads, which make up about 65% of Chipotle’s digital orders. Chipotle pitches them as a way to take the least popular prep work off its crews.</p><h3>How physical AI gets built today</h3><p>So far we’ve looked at where physical AI is already working and why the shape of the machine matters less than the humanoid hype suggests. For an enterprise deciding whether to build or buy, the next question is what separates a robot that works on an ordinary Tuesday from one that only works in the demo. Look past the body and there are four ingredients (none of them are a sweet potato).</p><h3>Sensors that suit the job</h3><p>Everything starts with what the machine can perceive. Waymo’s sixth-generation driver carries <a href="https://waymo.com/blog/2024/08/meet-the-6th-generation-waymo-driver/">13 cameras, four lidar units and six radar sensors</a>, giving it overlapping views that pick out objects up to 500 meters away, at night and in bad weather. A mine truck on a planned route can get by with far less. The sensors set a ceiling on what any model can understand, and they decide what data the machine collects to improve the next model.</p><h3>Models, plural, trained on the right data</h3><p>Most working robots run several models, each at its own speed. <a href="https://www.figure.ai/news/helix">Figure’s Helix</a>, for example, pairs a vision-language model with seven billion parameters, which reads camera images, the robot’s hand and finger positions and a natural-language instruction seven to nine times a second, with a much smaller policy that turns that intent into finger, wrist and torso movements 200 times a second. Both run on GPUs inside the robot. The <a href="https://aws.amazon.com/blogs/physical-ai/how-certis-achieved-autonomous-robot-security-patrols-with-aws/">Certis inspection dog that AWS describes</a> uses simpler parts in the same pattern: an object detector finds the target, a waypoint model plans the approach and a conventional controller does the steering. Installing the largest available model everywhere would be an expensive way to make some jobs harder.</p><p>Models of course are trained on data. Figure trained Helix on about 500 hours of teleoperated demonstrations. Certis fine-tuned its waypoint model on its own trajectory data, and Physical Intelligence turned its air-fryer coaching into training data. Training is expensive, particularly this kind of training, shadowing expensive experts. Generalist models should make this sunk cost more reusable: DeepMind ran one Gemini Robotics 2 checkpoint on three different robot configurations, and the <a href="https://robotics-transformer-x.github.io/">Open X-Embodiment project</a> has pooled more than a million robot trajectories from 22 types of robot. But a motion learned on one machine still has to suit the next one’s geometry, sensors and available force, so most deployments will still need fine-tuning for their own robot, site and job.</p><h3>Hands that don’t break lightbulbs</h3><p>The body has to deliver what the model intends. Motors, joints and grippers need the precision and force feedback to handle a lightbulb, a slippery bag or a sheet-metal part without dropping or crushing it. The <a href="https://www.cnx-software.com/2026/06/02/sharpa-wave-high-end-dexterous-robotic-hand-with-22-dof-high-sensitivity-dynamic-tactile-array/">SharpaWave hands</a> on DeepMind’s Apollo test robot each have 22 degrees of freedom and more than 1,000 tactile sensing points per fingertip, and the robot still screwed in a bulb only 36% of the time. Touch sensing helps, but models and hardware have to improve together.</p><h3>Reflexes that run locally</h3><p>Between a model’s decisions, control loops running hundreds of times a second keep a grip steady and a robot upright. That needs fast, predictable compute on the machine itself, which is why robot makers keep pushing more of it on board. NVIDIA’s <a href="https://www.therobotreport.com/nvidia-jetson-thor-brings-2k-teraflops-of-ai-compute-to-robots/">Jetson AGX Thor</a>, launched in August 2025, offers up to 2,070 teraflops of AI compute at low (FP4) precision in a 130-watt module, and Boston Dynamics and Agility Robotics are among the humanoid makers planning to use it.</p><h3>Training in the data center, inference on the machine</h3><p>Those ingredients come together in different places. Training a robot model takes clusters of data center GPUs, plus simulation to generate practice that would be too slow, costly or dangerous to collect in the real world. Fine-tuning for a site or task might run in the cloud, on premises or both. Inference, the point where the model drives the machine, happens on the robot.</p><p>Tooling to stitch that pipeline together is starting to appear. NVIDIA’s open-source <a href="https://developer.nvidia.com/osmo">OSMO</a>, for example, lets teams define training, simulation and hardware-in-the-loop testing as one workflow that runs across Kubernetes clusters in the cloud, on premises or air-gapped.</p><p>A model may be born in a data center, but it spends its working life on hundreds or thousands of machines in mines, fields and warehouses, usually far from an IT team. For anyone running a fleet, physical AI is edge AI.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*GFfZct7-oTpwGUgy.png" /></figure><h3>Why the robot can’t wait for the cloud</h3><p>When a load starts slipping from a gripper, sending a request to a distant service and waiting 100ms or more for the answer may be too slow. For time-sensitive perception and action, inference often needs to run on the machine or on nearby infrastructure whose delay can be controlled. A fast average response isn’t enough if an occasional long wait makes the movement unsafe.</p><p>Local execution also reduces dependence on a network connection, although a machine can operate offline only if its other essential dependencies are local too. And proximity alone doesn’t guarantee timing: <a href="https://control.ros.org/jazzy/doc/ros2_control/controller_manager/doc/userdoc.html#determinism">controller scheduling and jitter</a> still need engineering. Safety mechanisms must be designed and validated alongside the models.</p><p>Waymo provides a good working example. In a <a href="https://waymo.com/blog/2025/12/demonstrably-safe-ai-for-autonomous-driving/">December 2025 engineering account</a>, the company described training large teacher models and distilling them into smaller versions, including models optimized to run onboard. The driving system combines camera, lidar and radar information with driving-specific vision-language reasoning. A separate onboard layer validates the proposed trajectories.</p><p>Developers can use substantial computing resources for training and simulation, then deploy an efficient model close to the sensors. The car doesn’t have to carry all the computing power used to develop its driving ability. It does have to execute its part of the system in time.</p><p>That fits the <a href="https://www.spectrocloud.com/blog/hybrid-ai-enterprise-operating-model-part-1">seven-stop hybrid AI spectrum we’ve described</a> before. Teams choose both the models they need and where to run them. Cloud services can support training and fleet analysis; a site’s servers can coordinate local work; onboard hardware can handle the immediate response. The split depends on the job, including its connectivity, privacy and cost constraints.</p><h3>Who comes running when the robot stops?</h3><p>One of the huge appeals of physical AI is that it runs autonomously, even continuously. But that doesn’t mean people are completely out of the picture — they just might be doing different jobs.</p><p>Amazon, which describes itself as the world’s largest manufacturer and operator of mobile robots, says the shift is already visible in its buildings. Since 2019 it has upskilled more than 700,000 employees, many to work with advanced technology, and its next-generation fulfillment center in Shreveport, Louisiana, needs <a href="https://www.aboutamazon.com/news/operations/amazon-million-robots-ai-foundation-model">30% more people in reliability, maintenance and engineering roles</a>.</p><p>Waymo <a href="https://waymo.com/blog/shorts/advice-not-control-the-role-of-remote-assistance/">describes its remote-assistance staff</a> as providing contextual advice when requested by the driving system. They don’t drive the vehicles remotely, and the system can accept or reject their input. Sometimes that doesn’t produce the best outcomes, as San Francisco residents have observed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*XNRZLcvPpQ8c2lwf.jpeg" /></figure><p>For a warehouse, an occasional unplanned human intervention may be easy to absorb. Multiply it across a busy fleet, however, and staffing becomes part of the economics. A robot that works beautifully while someone watches it has demonstrated a skill. Whether it saves anybody’s time remains an open question.</p><p>Someone will still have to clean the camera and inspect the gripper, and they’ll need the right spare part when it breaks. It’s possible to replace repetitive manual labor with better technical jobs; it’s also possible to leave a smaller crew chasing complex faults all shift long.</p><p>Whether you build or buy, it’s vital to rehearse failure before it happens instead of leaving it to the vendor’s support line. Decide what a safe stop looks like for each machine and who’s allowed to restart it. Keep enough trained people and manual equipment to run the process by hand for a shift. Work out how a stuck robot gets recovered, how long spare parts take to arrive, and whether the logs and sensor data from an incident will reach the people who can fix the model. And plan for the failures robots have and workers don’t: a dropped network, a bad model update or a compromised controller can stop every machine in a fleet at the same moment. The safety standards for your class of machine, such as <a href="https://www.controldesign.com/industry-news/news/55268769/iso-10218-update-makes-functional-safety-requirements-more-explicit">ISO 10218</a> for industrial robots (revised in 2025 with new cybersecurity requirements) or ISO 3691–4 for driverless trucks and mobile robots, are an obvious place to start the checklist.</p><h3>A new model every Monday</h3><p>Once AI is deployed in machinery, updating it becomes an operational decision. A better perception model may change what the system detects, which can in turn change what it does next.</p><p>A fleet operator needs to know which combination of model, sensors and software is running on each machine. Engineers need a way to test changes under representative conditions and introduce them gradually, with a route back to a known configuration if performance worsens. Updating thousands of machines is also a question of maintenance windows and people available to respond. We’ve argued for years that edge fleets should treat everything from the operating system to the model as one versioned, immutable image, so every machine runs a known configuration and a bad update can be rolled back in one step. Physical AI fleets are no different.</p><p>Aurea, the orchard-spraying startup, deals with this on working tractors. <a href="https://www.youtube.com/watch?v=qCyDeCGavB8">At KubeCon + CloudNativeCon Europe in Amsterdam in March</a>, Aurea engineer Jordan Karapanagiotis and Spectro Cloud’s Mauro Morales described how each tractor’s NVIDIA Jetson computer runs <a href="https://kairos.io/">Kairos</a>, an immutable Linux distribution for edge Kubernetes, and takes whole-image A/B updates, so a unit that can’t boot a new version falls back to the last one that worked. (Disclosure: Spectro Cloud is the lead sponsor of Kairos, a CNCF Sandbox project.)</p><p>We expect companies that can maintain reliable fleets with regular software stack iterations to have an advantage over companies that can only demonstrate a capable machine.</p><h3>Who gets the gains?</h3><p>Removing people from repetitive lifting and dangerous work in mines, warehouses or wheatfields is an appealing prospect. Whether their working day improves also depends on what employers do with the additional capacity that physical AI can unlock.</p><p>A <a href="https://journals.sagepub.com/doi/10.1177/00197939251333754">2025 study in ILR Review</a>, which compared Amazon’s robotic and conventional fulfillment centers, associated robotics with roughly 40% fewer severe injuries, but 77% more nonsevere injuries. The researchers attributed part of the latter increase to faster work. Those are associations, but they complicate the assumption that automation improves every aspect of a job.</p><p>Amazon’s own figures look rosier. It reported that its <a href="https://www.aboutamazon.com/news/workplace/amazon-workplace-safety-2025-injury-reduction">global lost-time incident rate fell 70% between 2019 and 2025</a>. But that’s self-reported and company-wide, and Amazon credits a mix of safer design, inspections, training and $2.5 billion of investment, with robots only one part of it. On recordable incidents, the measure Amazon leads with, the improvement was 43%.</p><p>A worker spared heavy lifting may welcome the machine and still worry about a faster pace, fewer colleagues or closer monitoring. The second worry has evidence behind it. In October 2025, The New York Times <a href="https://www.nytimes.com/2025/10/21/technology/inside-amazons-plans-to-replace-workers-with-robots.html">reported internal Amazon documents</a> suggesting automation could let the company avoid hiring more than 600,000 US workers by 2033 while selling twice as many products. Amazon said the documents reflected one team’s perspective and didn’t represent its overall hiring strategy. Businesses have choices here. Savings can fund training and safer work, or simply a higher output target (and of course this applies to all AI discussions, not only physical AI).</p><h3>The next ordinary day</h3><p>For an enterprise developing or piloting a physical AI use case, the challenge is to follow it all the way through with a holistic perspective. How often does the task get completed acceptably? What happens to the exceptions? What does it cost once setup, supervision and maintenance are included? Compare the proposed robot with the equipment and workflow that would otherwise do the work.</p><p>Sometimes, as with BMW’s example, a humanoid will justify its complexity because the building, tools and changing assignments favor a human-shaped machine. In many other cases, we expect businesses to improve a forklift, fit better perception to an arm or send a drone where a person would struggle to reach. More adaptable models should widen those possibilities without making every machine resemble its owner.</p><p>We’re still in the early days of physical AI, and there’s plenty left to invent. For all the demos of robots performing tai chi or running obstacle courses, general-purpose dexterity remains difficult, and making physical AI affordable and dependable across suboptimal environments will take years of engineering. But the suggestion that we’re waiting for robots to enter the real world is getting harder to sustain. They’ve been hauling ore since before the current AI boom. People are already riding in them.</p><p>The passenger in a Waymo doesn’t judge the car a failure because it can’t also fold laundry. Getting across town safely, comfortably and on time is plenty to ask. If engineers can bring that degree of usefulness to more of the work around us, we’ll have ample reason to be impressed. We may also spend remarkably little time thinking about glossy white robots that look like us.</p><p><em>Originally published at </em><a href="https://www.spectrocloud.com/blog/physical-ai-already-on-the-job"><em>https://www.spectrocloud.com</em></a><em> on October 20, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4addc9b9cabd" width="1" height="1" alt=""><hr><p><a href="https://medium.com/spectro-cloud/humanoid-or-not-physical-ai-is-on-the-job-spectro-cloud-4addc9b9cabd">Humanoid or not, physical AI is on the job | Spectro Cloud</a> was originally published in <a href="https://medium.com/spectro-cloud">Spectro Cloud</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[KubeCon Chicago: familiar faces, fresh conversations]]></title>
            <link>https://medium.com/spectro-cloud/kubecon-chicago-familiar-faces-fresh-conversations-aaf3bf495b86?source=rss-72d24c2910d------2</link>
            <guid isPermaLink="false">https://medium.com/p/aaf3bf495b86</guid>
            <category><![CDATA[bare-metal-cloud]]></category>
            <category><![CDATA[platform-engineering]]></category>
            <category><![CDATA[kubecon]]></category>
            <category><![CDATA[kubernetes]]></category>
            <dc:creator><![CDATA[Ant Newman]]></dc:creator>
            <pubDate>Wed, 22 Nov 2023 09:03:09 GMT</pubDate>
            <atom:updated>2023-11-22T09:04:57.793Z</atom:updated>
            <content:encoded><![CDATA[<p><em>This blog was originally written by </em><a href="https://medium.com/u/30da521f1f79"><em>Cornelia Davis</em></a><em> and posted at </em><a href="https://www.spectrocloud.com/blog/kubecon-chicago-recap"><em>https://www.spectrocloud.com/blog/kubecon-chicago-recap</em></a></p><p>It’s been nearly half a year since I shared with you <a href="https://www.spectrocloud.com/blog/why-did-i-join-spectro-cloud-because-devex-matters">my excitement at being back in the cloud native space</a>. Doesn’t time fly?</p><p>Now that I’m back home from <a href="https://info.spectrocloud.com/kubecon-na-2023"><strong>KubeCon Chicago</strong></a>, you could say my reintegration is complete.</p><p>But while I still have the benefit of my outsider’s eye, it’s time for some observations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*CpWDfNQ6H0Ap9YwY.jpeg" /></figure><h3>A community busy building the future</h3><p>My last KubeCon was in Los Angeles in 2021, back in the muted and masked-up days of Covid. Fair comparison or not, what a difference two years makes. Chicago was <strong><em>hopping</em></strong>!</p><p>We put together a little video to give you a sense of the atmosphere, for those of you who weren’t there:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F-lDSFoDRbWM%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3D-lDSFoDRbWM&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F-lDSFoDRbWM%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/d69adf25a9d75e07ab18a34421fbf974/href">https://medium.com/media/d69adf25a9d75e07ab18a34421fbf974/href</a></iframe><p>The pre-conference day, home to the colocated events, felt as busy as any I’ve seen. There were half- and full-day tracks on graduated CNCF projects like Istio, Cilium and Argo, hot techs like Wasm and Backstage, and still-evolving spaces such as AI and edge (we again were Diamond sponsor for Edge Day, and shared <a href="https://www.youtube.com/watch?v=eMIWKwfno3k">some interesting new research data</a> as well as demos of our <a href="https://www.spectrocloud.com/blog/two-node-edge-kubernetes-clusters-for-ha-and-cost-savings">2 node architecture</a>).</p><p>This year’s structuring of the pre-conf day, where a full conference pass gave attendees entrance to any of the colocated events, seemed to have increased the cross-pollination — which can only be a good thing for learning and innovation.</p><h3>Hustle, bustle and sell-out sessions</h3><p>The rest of the week, I spent time on the show floor reconnecting with my industry colleagues, and the crowds never once let up. The hallways were full of people queuing for oversubscribed sessions.</p><p>I had the great honor to support one of those sessions. In fact, it was so popular that we not only filled the first room, we also sold out an “encore session” on the Thursday, with over 900 people in total attending.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JCyhg2oe5DoZf1Si3cDpAA.jpeg" /><figcaption>The first of two sell-out sessions…</figcaption></figure><p>What was it that grabbed the community’s attention so dramatically? Well, a real-world user journey, of course.</p><p>Justin Head, VP of DevOps at Super League Gaming, told the story of replatforming SLG’s Minecraft hosting service from bare metal, to the cloud and Kubernetes, and then back to bare metal. (In these days of cloud bill shock, it’s a story that many of you may find relevant).</p><p>The clever part is that SLG didn’t give up any of the goodness they had gained in the cloud on their way back to bare metal. They kept their containerized architecture, API-driven provisioning and lifecycle management of Kubernetes, and resource elasticity, all while saving 65% on their monthly infrastructure spend.</p><p>A moment of surprise for me came when I queried the audience as to their familiarity with two of the key technologies that power this solution:</p><ul><li>Bare metal as a service (BMaaS) providers like Equinix Metal, Cox Edge and PhoenixNAP, which support the Canonical MAAS API protocol.</li><li>The <a href="https://www.spectrocloud.com/blog/cluster-api-and-kubernetes-cluster-management">CNCF Cluster API (CAPI) project</a>, which turns API-served machines into Kubernetes clusters, via an open source <a href="https://github.com/spectrocloud/cluster-api-provider-maas">MAAS CAPI provider</a> (which we’re proud to maintain here at Spectro Cloud).</li></ul><p>Based on show of hands, only about <strong>20%</strong> of the audience reported having any experience with either of these technologies — the Kubernetes community is far, far from being done with innovation. And of course new people are entering the fold all the time. More than half of the attendees at the show were first-timers. Can we blame them for not knowing every project in the vast CNCF landscape?</p><p>If you’d like to watch the session with Super League Gaming, check it out on <a href="https://www.youtube.com/@cncf/videos">the CNCF YouTube channel</a>.</p><p>Or catch the slides here: <a href="https://kccncna2023.sched.com/event/1R2lz">https://kccncna2023.sched.com/event/1R2lz</a></p><h3>Enterprise ♥️ OSS</h3><p>KubeCon Chicago marked a new chapter in the evolving relationship between open source and the enterprise.</p><p>Cloud-native has had enterprise involvement from the earliest days, at least on the tech vendor side. Of course, Kubernetes itself was donated by Google, and even as far back as 2016, KubeCon was sponsored by big guns like IBM, Samsung, Cisco and Intel.</p><p>It took a few years for a wider range of enterprises to truly get on board as users of the cloud native ecosystem. But those that did really adopted “Open Source”, with capital letters. They drank from the upstream firehose, they built their own tooling, they rolled up their sleeves. Organizations like retailer Zalando were building their own <a href="https://www.youtube.com/watch?v=1xHmCrd8Qn8">cluster lifecycle managers</a> in 2018.</p><p>Fast forward to now and enterprises from all the supposedly slow-moving, risk-averse industries are at KubeCon <em>en masse</em>. I personally spoke with folks coming from telco, pharma, healthcare, and manufacturing. For example, after seeing him on the main stage on day one, I had the pleasure of spending an hour catching up with <a href="https://www.linkedin.com/in/michaelpatrickbowen/">Mike Bowen</a>, Director of the Open Source Program Office at BlackRock. I saw another dear friend and former colleague, <a href="https://www.linkedin.com/in/damani-corbin-4329089/">Damani Corbin</a>, who launched Boeing’s Open Source Program Office in the last year or so.</p><p>Suffice it to say, when investment houses and aviation manufacturers are running OSS offices and sending big-hitters to the event, the CNCF’s work to bring end-user organizations to the show has succeeded in style.</p><p>These enterprises are not only adopting open source, but increasingly see it as strategic and are building programs around it. But that no longer means a whole lot of tooling development in-house. They are now actively looking at commercial offerings that allow them to get the best of OSS, while relieving them of the significant burden that DIY approaches require.</p><p>Thanks to their investment in OSS over the last few years, enterprises are older and wiser — with tougher expectations of vendors. Namely, they are looking for commercial offerings that stay true to the principles of the OSS community that these enterprises have helped to shape.</p><p>These organizations have become cynical enough to spot when a vendor is embedding OSS, but have modified it without contributing back to the upstream, effectively making their offering proprietary. End users are okay with a vendor delivering value in addition to the OSS they embed, but they are absolutely not okay with a solution that locks them into using only that vendor’s solution. That’s not truly OSS.</p><h3>A new chapter in the platform engineering story</h3><p>Searching for the term “platform engineering” in both last year’s North America schedule and this year’s European event shows a half dozen or so sessions. Chicago’s schedule? No fewer than <strong><em>thirty</em></strong>.</p><p>Okay, so I’m cheating a little here because there is now a dedicated track for platform engineering, but this is, in fact, my point — platform engineering is getting a lot of attention.</p><p>I <a href="https://www.spectrocloud.com/blog/why-did-i-join-spectro-cloud-because-devex-matters">wrote about this trend</a> earlier this year and not only is it demonstrated with the new track, but also <em>everyone </em>at the conference was talking about the two white papers that the <a href="https://tag-app-delivery.cncf.io/">App Delivery TAG</a> has produced in the last year: the <a href="https://tag-app-delivery.cncf.io/whitepapers/platforms/">CNCF Platforms White Paper</a> and the <a href="https://tag-app-delivery.cncf.io/whitepapers/platform-eng-maturity-model/">Platform Engineering Maturity Model</a>.</p><p>In another nod to the elevation of the topic, platform engineering and developer experience permeated many of the keynotes. On opening day, a <a href="https://sched.co/1R4Rw">panel</a> of active CNCF community members from large organizations paid homage. In one of the closing keynotes, one of the founders of Kubernetes, Tim Hockin, spent significant time talking about the developer experience while talking about “<a href="https://sched.co/1R4e1">the things we’ll need for the next decade</a>.”</p><p>His themes had all of us at Spectro Cloud nodding along. Stuff like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/899/0*c0PFMXa45wkb1GIV.png" /></figure><p>…is, frankly, the reason we’re here. When <a href="https://info.spectrocloud.com/report-kubernetes-2023">56% of Kubernetes adopters have more than 10 clusters</a>, you need that higher level management capability. You need a platform that makes serving developers with clusters repeatable, consistent, and easy — not risky.</p><p>The CNCF community has been talking about the need to serve the developer experience for some time, but until now the App Delivery TAG yielded little. The two well-received papers and the buzz around this topic in Chicago has me hopeful that the time has come.</p><h3>New frontiers, new opportunities</h3><p>I can’t end this writeup without talking about AI, of course.</p><p>AI was all over the conference, including a full day unconference called <a href="https://sched.co/1UklN">AI Hub</a> that I will definitely prioritize in Paris.</p><p>But one of my favorite sessions was Priyanka’s <a href="https://sched.co/1R4Qy">opening keynote</a> where she live-demoed AI inference on her laptop.</p><p>Well, almost.</p><p>She <em>attempted </em>to demo it live, but after waiting on her machine for a bit she fell back to a recorded demo. The inference had happened on her machine by the time she had finished showing the recording.</p><p>Why did I love this so much? Because she was showing AI on smaller compute — something that is at the core of the emerging market category of <a href="https://thenewstack.io/edge-ai-how-to-make-the-magic-happen-with-kubernetes/">Edge AI</a>.</p><p>But more than that, I loved the conversation that she kicked off around Kubernetes itself evolving to serve this new category of workloads.</p><p>Kubernetes was originally built when microservices were on the rise, and was optimized for those use cases, but with the work that is happening in the ecosystem now, Kubernetes is showing itself to be quite adaptable to this new world.</p><p>Oh, and I just <strong><em>love </em></strong>live demos — even when they don’t work as planned. We’ve all been there, Priyanka.</p><h3>See you in Paris?</h3><p>We’ve already secured our sponsorship for KubeCon Europe in Paris next Spring, both for <a href="https://events.linuxfoundation.org/kubecon-cloudnativecon-europe/">the main event</a> and for <a href="https://events.linuxfoundation.org/kubecon-cloudnativecon-europe/co-located-events/kubernetes-on-edge-day/#call-for-proposals">Edge Day</a>. Keep an eye on our socials or <a href="https://info.spectrocloud.com/subscribe">subscribe to our newsletter</a> for all the details as they emerge. Hopefully see you there?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aaf3bf495b86" width="1" height="1" alt=""><hr><p><a href="https://medium.com/spectro-cloud/kubecon-chicago-familiar-faces-fresh-conversations-aaf3bf495b86">KubeCon Chicago: familiar faces, fresh conversations</a> was originally published in <a href="https://medium.com/spectro-cloud">Spectro Cloud</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A lot happens in two years in the world of Kubernetes]]></title>
            <link>https://medium.com/spectro-cloud/a-lot-happens-in-two-years-in-the-world-of-kubernetes-193d299a22b6?source=rss-72d24c2910d------2</link>
            <guid isPermaLink="false">https://medium.com/p/193d299a22b6</guid>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[platform-engineering]]></category>
            <category><![CDATA[cloud-native]]></category>
            <category><![CDATA[edge-computing]]></category>
            <category><![CDATA[devops]]></category>
            <dc:creator><![CDATA[Ant Newman]]></dc:creator>
            <pubDate>Thu, 03 Aug 2023 09:35:59 GMT</pubDate>
            <atom:updated>2023-08-03T09:35:59.156Z</atom:updated>
            <content:encoded><![CDATA[<p>We last posted here more than two years ago. Eek. Time flies when you’re changing the world of Kubernetes, right?</p><p>What’s been happening? Well, put it this way: we’ve been busy.</p><p>The Spectro Cloud team has grown to around 150 of the best and brightest, and our cadre of amazing <a href="https://www.spectrocloud.com/partners">partners </a>has swelled, too. We’ve picked up our share of <a href="https://www.spectrocloud.com/awards">awards </a>and we’re proud to be serving some of the <a href="https://www.spectrocloud.com/customer-stories">leading lights in Kubernetes.</a></p><p>We’ve been touring the world, sponsoring and speaking at hundreds of <a href="https://www.spectrocloud.com/events">events </a>large and small, from <a href="https://info.spectrocloud.com/kubecon-na-2023">KubeCon </a>to <a href="https://www.meetup.com/seattle-kubernetes-meetup/">local meetups</a>, sharing expertise.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5Vpcik5wIq6s78gJGPu-HQ.png" /></figure><p>Open source is still <a href="https://www.spectrocloud.com/community">at the heart of what we do</a>. Have you heard about <a href="https://kairos.io/">Kairos</a> and <a href="https://localai.io/">LocalAI</a>? (Do yourself a favor and take a look.)</p><p>And we’re proud to have continued our stream of detailed market research into the <a href="https://info.spectrocloud.com/report-kubernetes-2022">State of Production Kubernetes</a>, with our third annual report coming later this summer (it’s gonna be a good one).</p><p>Our killer product, Palette, hasn’t stood still either.</p><p>We continue to double down on tackling the daily challenges of running Kubernetes in the enterprise, from <a href="https://www.spectrocloud.com/security">security </a>to <a href="https://www.spectrocloud.com/day-two-operations">cluster lifecycle management</a>, with a goal of <a href="https://info.spectrocloud.com/esg-validation-study">saving your team precious time </a>every day.</p><p>Now more than ever, Kubernetes is multi-cloud, multi-distro, multi-integration, multi-workload and multi-environment. Multi-everything, really. Our mission is to make that multiplicity easy for you.</p><p>So we continue to build out our list of supported <a href="https://www.spectrocloud.com/integrations-and-environments">integrations</a>. We’ve tackled the challenge of running Kubernetes on <a href="https://www.spectrocloud.com/solutions/bare-metal">bare metal</a>. And most recently we’ve added the ability for you to <a href="https://www.spectrocloud.com/solutions/vms-on-kubernetes">bring your VM workloads</a> to Kubernetes, too (don’t forget to <a href="https://www.brighttalk.com/webcast/19922/589882?utm_source=SpectroCloud&amp;utm_medium=brighttalk&amp;utm_campaign=589882">sign up to the webinar</a> to learn more).</p><p>For many organizations, <a href="https://www.spectrocloud.com/solutions/edge">edge computing</a> is the future — particularly for AI workloads. A <a href="https://www.spectrocloud.com/news/spectro-cloud-launches-the-secure-edge-native-architecture-sena">string</a> <a href="https://www.spectrocloud.com/news/new-spectro-cloud-palette-edge-platform">of</a> <a href="https://www.spectrocloud.com/news/spectro-cloud-extends-its-palette-platform">innovations</a> has made Palette the <a href="https://www.spectrocloud.com/news/spectro-cloud-named-a-2022-gartner-cool-vendor-in-edge-computing">first choice</a> of organizations running Kubernetes at the edge: simply, securely and at scale.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FPB8TdTGmvKA%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DPB8TdTGmvKA&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FPB8TdTGmvKA%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/f586f8aba413499074ab0ba8c143a3a6/href">https://medium.com/media/f586f8aba413499074ab0ba8c143a3a6/href</a></iframe><p>We’ve added a <a href="https://www.spectrocloud.com/product/pde">developer environment</a> to our core ops-focused experience, in the form of Palette Dev Engine and its ‘App Mode’. So if you’ve been looking at IDPs and devex, we have something for you.</p><p>And with our <a href="https://www.spectrocloud.com/news/spectro-cloud-announces-palette-vertex-for-government">new VerteX edition</a>, we even have a comprehensively ‘FIPSified’ secure version of Palette for <a href="https://www.spectrocloud.com/solutions/government">government and highly regulated industries</a>.</p><p>So when I say we’ve been busy… I wasn’t joking.</p><p>Keep checking back here for more posts (I promise not to wait two years for the next one). But if you want to get the absolute latest, make sure you <a href="https://twitter.com/spectrocloudinc">follow us on social</a> and <a href="https://info.spectrocloud.com/subscribe">sign up to our updates</a>, for invitations to our webinars and regular newsletters.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=193d299a22b6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/spectro-cloud/a-lot-happens-in-two-years-in-the-world-of-kubernetes-193d299a22b6">A lot happens in two years in the world of Kubernetes</a> was originally published in <a href="https://medium.com/spectro-cloud">Spectro Cloud</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>