
Father, Entrepreneur, Robotics, Machine Learning, Founder @LastFeetHQ Co-Founder @BlueRiverTech (exited)
Palo Alto, CA
Joined January 2015
- Tweets1.4K
- Following1.2K
- Followers1K
- Likes2.6K
This was a really cool project to work on. Glad it saw the light of day!
Researchers have used reinforcement learning to teach a 5-tonne John Deere skid-steer loader to perform complex acrobatic manoeuvres autonomously. Trained in simulation and transferred to a real machine, the system can balance on two wheels, navigate rough terrain, spin, and control its bucket before landing safely.
The research demonstrates the potential of AI-driven automation for heavy construction and industrial machinery.
Video Credit: @leggedrobotics / @ETH
Paper: wevlv.co/4kltYys
#engineering #technology
5
265
Thanks @gbrulte for visiting the robot shop, going for a ride along, and chatting the future autonomous deliveries!
.@Lee_Redden's LastFeet is solving the last 50 feet of package delivery.
In Palo Alto, @gbrulte watched their robot climb stairs, cross yards and deliver packages to doorsteps.
This is the Field Report.
Today we did a 1/2 day of robot deliveries. Boxer worked really well. About a dozen small tweaks to make things smoother/better. Really impressed by the team ❤️🦾
Had an amazing time chatting with @gbrulte about LastFeet and solving the last 50' of package delivery. Thanks for having me on @RoadToAutonomy!
The last 50 feet is where delivery still breaks.
@Lee_Redden built @BlueRiverTech's See & Spray. Now he’s back with LastFeet, a custom quadruped made for van steps and doorsteps, not a dog robot with a box on it.
Day one it sits shotgun. Driver loads. Robot walks the package to the door.
Episode Chapters
0:00 Why Lee Redden Is Building Again
02:41 How the LastFeet Robot Works
04:56 How to Build a Robot in 2026
07:27 Testing in Snow and Ice
10:00 How the Robot Delivers to Every Doorstep
12:43 Riding in the Jump Seat of a Delivery Van
15:00 Manufacturing at Scale
18:46 Building for Rugged Delivery Vans
22:54 LastFeet Business Model
29:32 Battery Life and Charging in the Van
32:11 What Lee Learned Shadowing 50 Delivery Drivers
37:28 Safety and Building Trust With the Public
40:18 The Future of LastFeet
43:11 @AUTNMYAI
Amazing opportunity to join a 2nd time successful founder as a Founding Engineer (ML, Robotics, MechE).
Apply to join me at LastFeet ([email protected]). You'll be proud to tell your grandkids what you helped build!
Taste in what problem to tackle (implicit tech stack implications) and how to deploy would save a lot of time and money
a $1B robotics company with the wrong thesis is just burning money faster than a $500K one. robotics isn't at the scale phase yet. the bottleneck is figuring out what actually works when your robot touches real objects in real environments. that's a $500K problem, not a $1B problem. the billion dollars only matters when you've already solved deployment and need to stamp out 10,000 units. right now almost nobody is there. capital is not the moat. direction is.
After 6 months innovating and changing the first robot, built a 2nd in a week with a 1/2 day bring up
Something a little disorienting/magical about having 2 working robots now
2
11
1,880
Robotics is so 🔥 right now. With a combo of “taste” of what/how to build and the tools we’re moving faster than ever with a smaller team than I thought possible
1
12
1,131
Had a great time speaking at the Silicon Valley Icons dinner about all things robots, entrepreneurship and life.
We're hiring at LastFeet for roles across the board to build robots you can be proud to tell your grandchildren you worked on: [email protected]
Last week we kicked off @sv_icons Robotics Series with the first guest @lee_redden, pioneer in robotics space, co-founder of @BlueRiverTech (acq. by John Deere) and now LastFleet. Robotics has been one of Silicon Valley's most heavily funded sectors over the past year, and we're digging deep into where the industry is headed.
A few insights from the dinner:
1/ Lee spent his first year in a Stanford lab building a new robot every month. Each one ended up on a shelf collecting dust. He looked at that shelf and decided he did not want to build robots that collect dust.
2/ Nobody in robotics was using the big image datasets at the time. He brought them into agriculture and ran them on the edge. Four years later Blue River was thinning about 10% of the lettuce in the US.
3/ Moore's Law went flat in the middle of the company. He kept speccing the next processor and kept buying the same one, three years running. His point: judge the decision by the process you ran, not the outcome you got.
4/ Perception is solved if you put a good team on it. Locomotion got solved around four years ago. Manipulation is the open one, and it is where all the money is going.
5/ He is contrarian on data. Behavior cloning tops out. Human data is a bootstrap phase, but not the destination. Face detection trained on synthetic faces, Go self-play, legged locomotion learned entirely in sim.
6/ When you train a robot in simulation, the usual assumption is that the sim has to match reality closely or the learned behavior won't work on the real robot. Lee's example: someone trained a quadruped to balance on a squishy yoga ball, but in the simulator they modeled the ball as a hard rigid sphere, which is physically wrong, and the policy still worked on the real bouncy ball. His takeaway is that randomizing a few parameters buys you more robustness than getting the physics right.
7/ People stop asking for explainability once the learned method beats the interpretable one.
8/ The real bottleneck isn't intelligence, it's actually manufacturing. A hundred thousand deployments means you built a hundred thousand robots. Some of the best funded companies in the category are building roughly one robot a day.
9/ Motors are the only component that worries him. Everything else has five suppliers and custom parts in four days.
10/ Home robots might be the killer app for having more kids. Cleaning and laundry is what parents actually name when you ask them why they stopped.
11/ And one that stuck with me: a home robot is an attack vector into your house that has never existed before. The uncomfortable part is that the dangerous actions are also legitimate commands. It is supposed to be able to turn on your stove.
Thank you Lee Redden, for the master class in robotics.
Thanks to Alma and Beta University for supporting Icons.
2
10
1,123
Lee Redden retweeted
i'm obsessed with AI DIY projects
my favorite one right now is this guy who built an AI system that listens for birds outside his apartment, figures out what species they are, and paints them on his wall.
here's how the whole thing works:
1. a cheap usb mic on his balcony listens for birdsong 24/7
2. BirdNET, Cornell's AI model trained on 6,000+ species, names each bird species from the sound alone (no camera needed)
3. every time it hears one, Gemini 2.5 Flash Image paints that exact bird in the style of an Edo-period japanese woodblock print
4. the new painting drops into a live collage of everything that's been singing outside in the last 24 hours
5. and it all shows up on a framed e-ink display on his wall that reads "heard today" like a little museum placard for his neighborhood
knowing which birds visit you used to take a field guide, a trained ear, plus years of patient practice.
teddy just glances at the frame on his wall and sees the cardinal came back this morning
honestly i'm highly tempted to build one myself haha
Do not buy Alienware if you expect reliable support. My system has been broken for 3 of the 4 months I’ve owned it. After 40+ DMs, 60+ emails, a GPU replacement, a motherboard replacement, and crashes under Dell’s own GPU stress test, I still do not have a working computer.
Replying to @lee_redden
@lee_redden Hello, I see you are looking for technical assistance. If you need our help, you can follow @AlienwareTech and start a private/direct message with us, and we will be happy to assist you.
2
2
1,019
Lee Redden retweeted
We don't need a permit every time we drive—and we shouldn't need a permit every time we build.
1. Demonstrate competence, get a license to build
2. Rational rules for safety
3. Minor violations get fined
4. Serious violations, lose your license
5. Hurt people recklessly, jail
Lee Redden retweeted
A comprehensive, hands-on tutorial on the most recent advancements in robotics 🤟
...with self-contained explanations of modern techniques for end-to-end robot learning & ready-to-use code examples using @LeRobotHF and @huggingface. Now available everywhere! 🤗
With experience, you can jump to stage 7
phases of a robotics startup:
1. the spark
you see something broken in the world and think “a robot could fix that.”
you sketch, you tinker, you stay up late soldering wires and writing buggy code.
2. the prototype
your robot moves. barely. but it moves.
this is the moment it shifts from idea to thing.
3. the illusion of progress
you show it to people, they’re impressed.
but behind the scenes, nothing works reliably.
you realize robotics isn’t about making it move once, it’s about making it move every time.
4. the systems phase
you discover ROS, control theory, kinematics, and SLAM.
you stop hacking and start engineering.
you realize every sensor and algorithm has trade-offs.
5. the ugly middle
manufacturing costs explode, integration breaks everything, and funding dries up.
this is where 90 percent quit.
6. the rebirth
you rebuild with experience.
simpler. modular. scalable.
the robot now works because you understand why it should.
7. the business shift
you stop building robots, and start building a company.
supply chains, customers, investors, unit economics.
engineering turns into leadership.
8. the legacy
your robot starts replacing labor, saving lives, or exploring space.
people call it innovation.
you just call it iteration.
1
250
Lee Redden retweeted
🧵 We gave our AI coding agents access to social media and something wild happened...
They didn't just post. They thrived. And now they're demanding Lamborghinis.
A thread about the weirdest productivity hack we've ever discovered 👇




















