jaradcannon

Robots & AI | CTO @ Humanoid | Previously CTO @ Brain Corp

[Humanoid @TheHumanoidAI]

Today, we're announcing that we’ve raised $152 million in our Series A at a $1.35 billion post-money valuation, becoming Europe's first pure-play humanoid robotics unicorn just two years after founding. This funding brings the total raised to date to $270 million.

The round was led by @PrimeMoversLab, with participation from @SchaefflerGroup, @BoschGlobal, Fubon Financial Holding Venture Capital, and Aglaé Ventures.

The capital will support development and launch of the next-generation humanoid robot platform, the roll-out of Beta version in Q4 2026, start of mass manufacturing of wheel-based robots and the continued development of KinetIQ, our AI brain.

Humanoid's valuation reflects rapid execution and the technology's shift from research to commercial deployment, demonstrating that Europe can build and scale globally competitive Physical AI companies.

Read more here: https://t.co/3p9AYMS0Ca

[jaradcannon @jaradcannon]

@zhengyiluo You can segment the thor's GPU into 3 logical GPUs

[C. Zhang @ChongZzZhang]

this curve is important and adds significant value to real2sim2real

[Humanoid @TheHumanoidAI]

Today we’re introducing KinetIQ Ascend: our reinforcement learning approach designed to reach 99.9% manipulation reliability at human speed and beyond.

It enables our robots to learn from real production tasks and improve through trial and error:

• Machine feeding: throughput +42%

• Item handover: throughput +85%, success from 80% to 98%

• Tote handling: throughput more than doubled, success from 78% to 99%

More in the blog post below 👇

[jaradcannon @jaradcannon]

Full technical blog here: the general RL formulation (PPO on flow-matching VLA policies), the reward design, and generalization results that surprised us. Strong signs it's not just overfitting: we trained on one set of objects, and we saw +40% on objects it never manipulated.

https://t.co/NSKtORDjdw

We're hiring across London/Boston/Vancouver/San Diego. Come join us!

[jaradcannon @jaradcannon]

A big open question in robot learning has been whether the "RL scales predictably with experience" story transfers from LLMs to real-world manipulation. We've now shown that it does for these industrial tasks on our humanoids: end-to-end from vision, with sparse generic rewards and no privileged info.

The curves are still climbing. Proud of what @hr0nix and the team have pulled off here.

[hr0nix @hr0nix]

Today's robots are too unreliable and slow for real industrial work. Reinforcement learning changes that.

Meet KinetIQ Ascend: our robots practice real production tasks and improve predictably towards industrial reliability standards.

How it works 🧵

[jaradcannon @jaradcannon]

@JayLiStanford Really curious to hear more

[jaradcannon @jaradcannon]

@bilawalsidhu Ok this is even cooler than the first time you posted about this. Being able to ask that type of question when it's genuinely ambiguous and get a visualization like this is scifi

[jaradcannon @jaradcannon]

@tankots @WisprFlow Love it on my work machine, but don't want it having full access to my screen on mobile so I don't use it there. Please solve this privacy issue

[jaradcannon @jaradcannon]

@standardbots @RoboStrategy @generalcatalyst Congrats @evanbeard and team! Awesome to see your team's progress through the years!

[jaradcannon @jaradcannon]

@graphtheory @0interestrates fffff

[jaradcannon @jaradcannon]

@0interestrates this is the way

[jaradcannon @jaradcannon]

@Andercot Time to rebuild beauty and aspire to build a future.

[jaradcannon @jaradcannon]

@saranormous at least on the notes side is this not just claude code + obsidian? ask claude code to create a transcription service that just auto-transcribes if you click and drag. everything else is already handled?

[jaradcannon @jaradcannon]

@srush_nlp @ericzakariasson @drawitpoorly @drawitpoorly me

[jaradcannon @jaradcannon]

@xiaolonw @Meta Congrats to you and your team!

[jaradcannon @jaradcannon]

@Scobleizer Thanks, this is always a moving target, useful to have this as an up to date reference

[jaradcannon @jaradcannon]

@kepano This is fantastic, thank you for building a great product!

[jaradcannon @jaradcannon]

@kepano appreciate what you do. I've been using obsidian (and sync!) for years. The instant llms showed up it felt like this is the obvious path forward. Been using llms & now agents in my obsidian notebook for years.

[jaradcannon @jaradcannon]

@kepano Is it possibly to run these headless? Ideally being able to run sync headless would be very useful.