Deployed: Where Physical AI Gets Real

John Giannandrea is the former Senior Vice President of Machine Learning and AI Strategy at Apple, and former head of Search and AI at Google. Earlier he worked at General Magic and Netscape, and was CTO of Tellme and of Metaweb, which Google acquired in 2010.

John joins Brad Porter, who worked for him at Tellme, to talk about ideas that arrive 20 years too early, the day he saw the web and quit his job, why skeptics are a good sign, and how to know when an AI system that is never 100% right is ready to ship.

The conversation ultimately points toward a future where the next leap in AI comes from better architectures and a deeper understanding of the models, and where the teams that win, in software or robots, set a launch bar, ship into real work, and keep climbing.

0:00 Intro
1:33 From Scotland to Silicon Valley: Inmos, Pixar, and SGI
4:15 General Magic: building the iPhone 15 years too early
5:18 Seeing Mosaic, quitting, and joining Netscape as employee #20
8:42 AOL buys Netscape and the move to Tellme
11:43 Open-sourcing the browser: the road to Firefox
13:32 Metaweb, Danny Hillis, and building the knowledge graph
17:54 How Metaweb disrupted Google Search
21:57 Google's AI second act and the $15K espresso machine
24:41 Leading Google Search: a policy job, not a technical one
27:23 Leaving Google for Apple: two opposite cultures
31:48 Hill climbing: how ML features get to launch
33:26 Beyond data and compute: the next architecture breakthrough
35:08 Physical AI and why more data isn't enough
36:56 Privacy, Apple Intelligence, and Private Cloud Compute
40:08 Shipping LLMs: Siri, Gemini, and ChatGPT's surprise hit
41:24 AI risk: what Geoffrey Hinton actually worries about
42:41 The Vatican cardinal and the "200-year experiment"
44:35 Was the internet a good thing? Building with humanistic intent
48:50 Young people, search, and the hunt for truth
50:52 AI is dual use: society will adapt and regulate
52:43 US vs. China and the global talent race
55:55 Overhyped short term, underappreciated long term

Swami Sivasubramanian is VP, Agentic AI & Emerging Technologies, AWS. He joined Amazon as an intern during his PhD, worked on Dynamo, and went on to launch DynamoDB, SageMaker, and Bedrock; his organization now builds Kiro, Bedrock AgentCore, Quick Suite, and AWS Transform.

Swami joins Brad Porter for a candid conversation about the gap between AI assistants and AI agents, what it means to be on the loop rather than in it, and how Amazon rewired the way it works: one agent for a company-wide Java upgrade, a demo in a week before the PR/FAQ, and no more version 30 of a doc.

The conversation ultimately points toward a future where the models are no longer the constraint and the organization is: the teams that pull ahead stay on the loop, decide in days instead of months, and put deterministic guardrails around probabilistic agents.

Brad Silverberg is a founder and advisor at Fuel Capital, and before that he spent the 1990s at Microsoft, where he led MS-DOS, Windows 3.0, Windows 3.1, Windows 95, and Internet Explorer, and grew Windows from roughly a $50 million business into a $3.5 billion one. He was PC Magazine’s Person of the Year in 1995, and he arrived at Microsoft by way of software research at SRI in 1978, the Apple Lisa team.

Brad Silverberg joins Brad Porter for a candid conversation about why the Lisa and the original Mac were brilliant and still struggled commercially, how he shipped Windows in deliberate steps with Windows 95 in his head from day one, and how Internet Explorer came from two or three percent share to winning the browser before Microsoft gave the lead away.

The conversation ultimately points toward a rule that outlives any single platform: platforms are won not by the best idea but by whoever sequences the hard problems correctly, and they are lost more often to the politics inside the building than to anything a competitor does.

Erik Nieves is the co-founder and CEO of Plus One Robotics and a 25-year veteran of Yaskawa Motoman, one of the industry’s largest robot makers, where he ran the technology group. Since founding Plus One in 2016, he has focused on vision-guided robots doing real parcel work for the big couriers, a fleet that has now passed two billion picks.

Erik joins Brad Porter for a candid conversation about the distance between a robot demo and a real deployment, what it takes to sell automation to companies a thousand times your size, and why a human in the loop still keeps the robots working.

The conversation ultimately points toward a future where progress comes less from how robots move than from how well hands and eyes work together, built by engineers who have lived with the problem on the operator’s side. Also, check out the live stream response to Figure's Demo! https://bit.ly/4czQ37G

Russ Tedrake is a renowned roboticist, MIT professor, and founder and CEO of Walden Robotics. After decades of advancing robot locomotion, manipulation, and AI—including influential work with humanoids and the DARPA Robotics Challenge—he is now focused on bringing intelligent robots into real manufacturing environments.

Russ joins Brad Porter for a candid conversation about leaving the lab for the factory floor, the practical choices behind robots that can work today, and what the physical AI boom could mean for the future of manufacturing.

The conversation ultimately points toward a future where robots don’t simply replace individual tasks, they help factory workers and production engineers improve entire systems faster than ever before.

0:00 Intro
2:00 Lessons from Atlas: deforming legs and broken hands
6:30 Kung-fu humanoids: entertainment vs. real-world value
8:20 The real world is the ultimate eval — and the Walden pitch
11:30 Humanoid or not? Wheeled base, grippers, and safety cases
15:30 Why floors matter: flatness, wear, and messy facilities
19:20 Generality and embodiment: how much does morphology matter?
22:50 What data is actually salient? Filtering and noisy data
26:50 Challenges as evals: the Amazon Picking Challenge and its alumni
29:00 Capital, bubbles, and whether this time is different
31:00 Reshoring, export rules, and the logistics engine
33:40 Tips for founders and Chesterton's fence
36:10 Factories are dynamic: co-evolving with robots
37:20 The metronome effect and compounding automation gains
41:15 Vibe coding the factory floor and the Toyota Production System
45:30 The vibe at MIT, and closing

Mike McCue has spent 40 years building the tools that made the web more human, Paper Software, Netscape, Tellme Networks, and Flipboard. In this episode of Deployed, he sits down with Brad Porter, the founding engineer he once recruited at Tellme, to talk about what it actually takes to ship hard technology across the internet, mobile, and AI eras.

They get into the first-principles call that saved Tellme (walking away from a company-defining AT&T check), how to build teams that compound, why he shipped voice AI two decades before the world was ready, and two unforgettable Steve Jobs stories, including the day Jobs picked up an iPad and explained Flipboard's purpose better than Mike could himself.

Brad Porter sits down with Aicha Evans, CEO of Zoox and one of the few people who has actually put fully driverless robotaxis on public streets–no steering wheel, no safety driver, no one to take over. They get into what really happens when a robot car meets the real world.

Aicha traces her path from a childhood between Senegal and Paris–hacking the phone to call home because the bill ran a thousand dollars a month–to Intel, to the Zoox factory floor at the start of production. She dismantles the debate consuming physical AI, arguing that the models will commoditize but hardware-software integration never will, and that the real moat is people who can sit at the same table with equal decision rights.

In 2013, Teresa Carlson led a team of four to bid against hundred-person squads for the CIA’s first commercial cloud contract. Competitors called to laugh, and she figured she’d either get promoted or fired. She won, delivered in record time, and accidentally created the playbook that made “gov tech” and “defense tech” real categories where none had existed. Before any of that she spent five years as a speech-language pathologist at the Nuremberg Hospital, transforming Romanian orphans through therapy, nutrition, and relentless contextual cueing. It is the same listening-first operating system she later used to sell Amazon, and then the US government, on ideas neither was sure it wanted. Teresa sits down with Brad to trace the line from a Kentucky clinic to the CIA’s server room to the General Catalyst Institute (recorded just before she was named Anthropic’s first Global Head of Public Sector), where she built the bridge between venture-backed startups and the world’s largest, most conservative buyer, and what “shovel-ready capability” means for companies like Cobot bringing physical AI to government today.

Brad Porter sits down with Sidd Srinivasa, the roboticist he recruited to build Amazon’s Robotics AI org and now a professor at the University of Washington’s Allen School. They get into the gap between what robotics promised and what it actually delivered, why a robot that helps a paralyzed person eat is really a problem about dignity, why Sidd thinks the humanoid is the wrong default (“we humans are a local minimum in evolution”), why most robot demos are the easy 30 percent, and why moving the field forward takes uncommon sense: you have to be contrarian and correct.

Pat Shanahan has deployed some of the most complex systems ever built. Thirty years at Boeing on the B-2 bomber, the V-22 Osprey, the F-35, the 777, and the 787. Then Deputy and Acting Secretary of Defense under Secretary Mattis, leading a department of more than 3 million people. Then the turnaround of Spirit AeroSystems. He was elected to the National Academy of Engineering in 2019, and he helped Brad and the team at Scale AI.

In this episode of Deployed, Brad Porter and Pat Shanahan get into what it actually takes to put hard new technology into the hands of the people doing the work. Pat tells the story of betting the company on designing the 777 entirely on a computer, and the moment on the factory floor when everyone was sure it would fail. They cover his rule of "only three miracles at once," why every deployment is a people problem before it's a technology problem, the difference between AI "safety" and product integrity, and why a man who has shipped at the largest scale there is wishes he could go back 20 years in his career with these tools.

Brad turns the tables on his former MIT professor, David Mindell. David is a historian, deep-ocean robotics pioneer, founder of Humatics, and now an investor in industrial transformation through his firm Unless. They trace the thread from James Watt’s steam engine to robots two miles deep in the ocean to the warehouse floor: why the demo is never the hard part, why industrial transformation runs on “human time” and takes about 30 years, why the real automation problem is too few robots rather than too many, and why — from Thinking Machines to OpenAI — there’s no technology story that isn’t a human story.

Brad reconnects with Scarlett Koller, who he first met as an MIT sophomore for Amazon’s original Prime Air intern cohort, back when drone delivery was a “60 Minutes” segment, not a product. She’s since delivered on SpaceX Crew Dragon, NASA Perseverance, and Amazon Kuiper, and now runs Mithril , a deep tech startup reshaping space sensing. They trade stories on why NASA’s shuttle veterans cared so much about a simple handle, why people are harder to engineer than radiation, and how an antenna the size of a volleyball court, controlled to the thickness of a sheet of paper, could keep us from losing low Earth orbit.

Brad reconnects with Ross Fubini, the two met as Netscape interns when being a nerd wasn't cool. Ross went on to accidentally make the most valuable introduction in Silicon Valley history, connecting Palantir's first employees, then built XYZ Venture Capital around the network that followed. They trade notes on what the browser wars actually teach us about the AI wars, why only 40 people in America know how to sell tech to the Pentagon, and whether OpenAI ends up like Netscape: world-changing, then gone.

Brad Porter sits down with Bilal Zuberi: MIT PhD, former Lux Capital GP, and the investor behind Applied Intuition, the most profitable physical AI company you've probably never heard of. They dig into why the smartest money is leaving mega-funds, what separates visionary founders from execution founders, and why the entire enterprise software market is smaller than most people think. Bilal makes the case that AI's real impact isn't in chatbots, it's in a physical world that's stayed dumb for decades.

Everyone says humanoid robots will win because the world was built for humans. But was it? Sequoia Capital partner Alfred Lin joins Cobot CEO Brad Porter for a conversation spanning their shared history, from the Zappos deal that almost convinced Amazon robots would never work, to why Alfred's father's journey from farmer to bank president is the best case for AI optimism. They dig into why the best robot form factor might not have hands at all, what the "ChatGPT moment" for robotics already is, and why every technology wave demands a combination of new hardware and software that we haven't yet seen on the consumer side for AI.

Sign up for exclusive insights from leaders in Physical AI and Robotics: https://cobot-deployed.com/RSS2

Jeff Wilke scaled Amazon from a startup to the most sophisticated automation operation on earth. Now he's focused on rebuilding American manufacturing. Brad Porter and Wilke go deep on what enterprise AI looks like in practice, from warehouse automation and autonomous mobile robots, to the Physical AI deployment lessons that only come from running operations at massive scale. If you're building, investing in, or deploying robotics and automation, start here.

Sign up for exclusive insights from leaders in Physical AI and Robotics: https://cobot-deployed.com/RSS1

Deployed Episode One Drops 3/24 9AM PST.