Co-founder & CEO Paddington Robotics Ex-Head of Applied Research @Twitter | Co-founder Magic Pony Technology (acq. by Twitter) | PhD @imperialcollege
London
Joined November 2011
- Tweets2.2K
- Following909
- Followers4.4K
- Likes3.5K
Just saw we're 2 places ahead of @perplexity_ai on the list. Wasn't expecting that🤯
Somehow made it into @Jackandjillai's 100 most wanted startups despite still being in stealth & not even having a proper website.
Exciting to see our approach resonating with talented people.
Supermarkets are our first proving ground, not our ceiling.
Big few months ahead!
Somehow made it into @Jackandjillai's 100 most wanted startups despite still being in stealth & not even having a proper website.
Exciting to see our approach resonating with talented people.
Supermarkets are our first proving ground, not our ceiling.
Big few months ahead!
Walmart’s struggles show just how challenging deploying robots in the real world can be.
This is why a "deployment first" approach is needed. Ship early, face the chaos, and iterate on reality, not just theory.
futurism.com/robots-and-mach…
Zehan Wang retweeted
Flow has raised a $50M Series B at a $750M valuation, co-led by Antonio Gracias (Valor) and Gavin Baker (Atreides) with Sequoia Capital, Roelof Botha (SpaceX, Block), and more.
When I became a mechanical engineer, I wanted to invent. In reality, 90% of the day to day execution work (CAD, excel, sim) was digital manual labour. That’s about to flip. The engineer’s role will change more profoundly than at any point since the invention of CAD, with humans focusing on invention, architecture, tradeoffs and judgement calls, and agents taking the grunt work.
Anduril, Joby, Stoke Space, Rivian & many more - Flow is now the default platform for requirements and verification for frontier hardware teams. But it’s not just next-gen anymore. Industry giants including General Motors PPU and Volkswagen & Rivian’s joint venture (RV Tech) are reshaping their core development practices with Flow’s AI.
The revolution is just starting. Models are getting better quickly at CAD, analysis, simulation, and tool use. In the next year, AI will move from integrating the work to doing the work.
One day, this shift won't just enable us to iterate faster, it will enable us to design a class of products that we could not have dreamt of before.
We’re hiring.
First time trying the new Waymo car and found an edge case already where it got stuck. Many European roads are tighter than this.
Zehan Wang retweeted
Seeing a hype wave around GEN-1.5, and rightfully so. Lots of respect to Pete & Andy for executing so well. The secret is in the naturally repetitive motions in human-collected data. There're 2 main sources for such repetitions:
(1) Symmetric patterns. Sorting, tidying, and assembling almost never finish in one motion. Open any assembly manual from IKEA, and you find most objects symmetrical. You drive one bolt, then its twin, then the next pair. Every {bolt A, bolt B} pair is a natural continuation in context, and the second instance is a free training signal that imitates the first ("prompt").
(2) Recovery. Humans drop things all the time, but we pick them up so fast, we don’t even notice. That reflex to fix is half of our physical competence. The key insight is to keep the failed first half instead of trimming it away. If the model consumes the full arc, fumble, catch, continue, then recovery shows up organically at test time. It's funny that in-context improvement results from *NOT* over-sanitizing your data.
The other critical ingredient is UMI. I've been saying for a while that teleop will not last, and GEN-1.5 is driving the final nail in the coffin. UMI is essentially a human wearing the robot gripper to collect data directly (human → data). Teleop inserts a layer of separation: human → VR/skeletal device → robot → data, which bleeds out all the human "physical intuition". The subtle sleight of hand we perform constantly with objects, the micro-adjustments, the feel of a part snapping into place, is nearly impossible to capture when you can't feel the environment directly.
Once you have enough data, many behaviors can actually be zero-shot. For example, you don't even need finetuning to pick up a novel object. The model "just knows" what to do given a similar scene in the training distribution. Whether in-context learning truly works or not also depends on how far away the test is from training. Currently, the demos are still a bit too simple to conclude.
I'm cautiously optimistic. Still, it's a great day in robotics.
I recently had a great experience recording a podcast with @AmardeepSParmar. Lots of brilliant questions from Amar, who has really been perfecting his craft to become an excellent podcast host!
Link to episode on YouTube: youtu.be/0APU_qk8skY?si=TR_f…
Zehan Wang retweeted
🚀 Meet PRX Pixel.
Our new open-source 7B text-to-image model that generates images directly in pixel space.
After months of pretraining on hundreds of millions of images, supervised fine-tuning, and preference alignment, we're excited to share a first public preview.
The weights are already available, and we're currently working on integrating the model directly into Diffusers 🤗to make the model even easier to use.
Test it yourself in the demo below. And as always, we'll be sharing the full story behind the model through a series of technical blog posts covering the entire training recipe.
Link in the comments 👇
Also what we're doing. Except we don't even need teleop to begin
The best robotics wedge nobody’s talking about:
Industry-specific form factor → teleop deployment → proprietary workflow data → agentic automation layer on top built into the systems of the end customer → full autonomy explodes margin over time
Lower COGs than general-purpose humanoids.
Real revenue on day one.
Switching costs before you’ve touched autonomy.
Distribution, cost structure, process power, counter-positioning and cornered resources are moats.
Robots aren’t exempt from this.
This is exactly the approach we're taking at Paddington Robotics. Start with a specialist system which allows us to learn, get into production and then expand to adjacent use cases.
We can aim for geneality eventually but don't hold your breath for robotics' ChatGPT moment
Robotics is advancing rapidly: we can move even faster by being more pragmatic and less dogmatic. Excellent summary by @AnneliesGamble:
It's been 10 years since Magic Pony was acquired by Twitter/X, and now we are in the year of the horse. Let's hope its a good omen for the year ahead!
Zehan Wang retweeted
“The unintended consequence of the rental reform act will be a 30% increase in rents”
I just spoke to one of the UK’s most knowledgable landlords and property investors.
He said that the government have tried to solve a few problems with unprofessional landlords in the UK but they are creating a much bigger problem.
Currently the UK has millions of accidental landlords. These are people that have a rental property as a result of getting married (or divorced), inheriting a property or randomly buying a rental property as a way to get some financial independence later in life. These are people that have one or two rentals but they aren’t “professionals”.
These new legislation makes it a legal minefield to own a property. It gives local councils a huge financial incentive to go after landlords for minor infractions.
The net result of this is a generational shift in housing sales to “professional landlords” like foreign private equity firms, banks, pension funds and very wealthy individuals.
Large professional landlords are well placed to do two things. Firstly, they mercilessly increase rents each year because under the new rules there are “limits” on how often and by how much rent can go up - they max out these limits. Secondly, they corner the market and enjoy less competition from smaller landlords with lower costs and more compassion. They even use the new laws to get smaller landlords forced to sell to them.
We are rapidly seeing the consolidation of the housing market into the hands of the wealthy thanks to the do-goodery of the government. They might mean well with their intention to “clean up the rental market” but the impact is going to be dire. Big companies love regulations because it destroys their little competitors and then they can dominate and price gouge with impunity.
This happened before in the beer brewing industry. We used to have thousands of small boutique brewers and then the government tried to improve things with regulations and new taxes. Shortly after the entire industry consolidated into a few massive corporations (mostly foreign owned).
Whatever the government tries to do, creates unintended consequences. The people responsible are only in government for a few years and then they move on to lucrative speaking, consulting and board roles. They leave behind a trail of destruction paved with good intentions.
Be very careful asking governments to do more. They just might do what you ask, destroy competition and accidentally make people’s lives hell.
Also, it's easier to make something safe when it's on a wheel base.
When out of power, it won't just collapse or fall over. Safety has to be critical when deploying robots around people.
It makes more sense as the starting point.
I disagree with this.
Yes, the wheel base shown wont be able to go up stair(s) - but the way in which data is getting collected means that the wheels can be swapped for legs (or some other form) at a future time without having to throw out data.
Mobility and manipulation are two beasts on their own, and it's wise to choose your battles. Wheelbase gets you deployed in the real world a hell of a lot faster and a wheel base has way less edge cases than legs/is less expensive/has longer battery life
A robot that can walk but not manipulate is significantly less useful than a robot that has limited mobility but can manipulate
LOLs aside, I actually don't think this is how startup founders should operate.
I like to keep my team and investors in the loop, and it is ok to ask for help.
If you can't trust your closest supporters / team, then there's something wrong with the foundations of the company
