Tiny edge transitions are critical when interacting with parts. We slid a Kyber fingernail over thin shims and it reliably felt the “ka-chink” as it dropped onto the table, including relative magnitude. For paper, films, foils, and assemblies, that tactile state change helps the robot find edges and manipulate and assemble parts more accurately. Want a hand? Wait list is open! lnkd.in/eJCMzuMz

Check out our recent work accepted to #RSS2026! We enable a robot to learn a flying knot from a single human demonstration and less than 10 trials using Task-Level Iterative Learning Control: flying-knots.github.io/

HAPPY 1 YEAR ANNIVERSARY @noxmetals A mission that started out as a summer time high school job at my dad's shop, to the past 365 days of nothing but thinking about metal. >22 US manufacturing jobs created >hundreds of deliveries to American machine shops >more weekend and emergency orders than I can count, to keep American manufacturing going >brought life back to an empty Detroit factory >enjoying every damn second of it more saws. more metal. more code. more factories. more truckloads. more jobs. more reindustrializing america.

Walking the Optimus production line in Fremont

We had some fun getting the hand to quickly pick up and set down four small berries without crushing them, dropping them, or launching them across the table. Soft fruit is a fun test because you need enough contact to move fast, but not so much that you squish it. Our hardware makes that surprisingly simple: the compliance keeps contact forgiving, and the torque transparency helps the hand naturally settle into a useful grasp even with very simple controls.

In Sep 2024, I left a job I loved to start @proceptionAI with one obsession: to give robots hands that actually work. 20 months. 20 people in Mountain View. A lot of Coke Zero. Today we ship.

By the way a huge shout out to @GoingBallistic5 for calling it two years ago: injection molded plastic body panels ftw! Also, I was listening to Scott and my discussion from October 2024 after we rode in the cyber cab for the first time and realized that Elon had been a little bit cautious and said it would likely be the end of 2026 or even 2027 before anyone got to ride in the cyber cap. Well, it turns out he actually might’ve been a little too conservative in that case because we might all be riding in them relatively soon now. So for those who say Elon‘s always overly optimistic, maybe for once he actually was not.

A quick follow-up demo showing the hand safely interacting with people while it works. Even while threading, the hand is only using the force it needs for the task, so someone can interrupt it with their own hand without it hurting. This is a big part of how we think about manipulation: not building a Terminator-strength hand, just a capable, force-aware one that can do useful work around people. Want a hand? Wait list is open! lnkd.in/eJCMzuMz

🦅🦅🦅🦅 Chicago was fire! So many great memories made and awesome people met! With my fellow robot guy @GoingBallistic5 at @standardbots May the robots be with you! 🫡

A robotics company just landed a $2.5B public listing, without ever filing for an IPO. How does that even work? That’s the SPAC playbook. Agility Robotics is the latest to use it, and it’s a good moment to explain how these work and why companies pick them. How a SPAC works • A SPAC uses a “blank-check” company, a shell that goes public first with no business at all. SPAC = Special Purpose Acquisition Company • It IPOs, raises a pool of cash that sits in trust, and lists on the exchange as an empty vehicle with one job: find a private company to merge with. • When it finds one, they combine, and the operating company takes over the shell’s existing listing, usually under a new ticker. • The target never runs its own IPO. It inherits one. Agility’s deal, as the example • Merging with Churchill Capital Corp XI (Nasdaq: CCXI), a SPAC from serial sponsor Michael Klein (behind the Lucid and Oklo listings). • Values Agility at ~$2.5B, with $620M in proceeds: Churchill’s trust cash plus a ~$200M PIPE anchored by Foxconn. • Will trade as AGLT. Why companies choose it • Faster and cheaper, because a shell has almost nothing to scrutinise: no business to value, no years of financials to audit, a thin prospectus. • Valuation is negotiated privately with one counterparty rather than set by a live, volatile market, so more speed and more price certainty. • The PIPE brings strategic backers (like Foxconn) in at the same time. • The catch: the hard work doesn’t vanish, it just gets deferred to the merger. This post is for informational purposes only and not a recommendation one way or the other. Do your own research.

Publishing a new Robotics Business breakdown in the next 24 hours Summary in image Time to settle the debate: Humanoids or special-purpose robots -- who actually wins? Engage with this post and I'll DM you a sneak peek (must be following so I can DM)

A deceptively tricky wet lab demo: measured solids transfer between vials, using table salt as a stand-in. Solids can mean powders, flakes, granules, sticky gels, and more, which is exactly why general-purpose manipulation matters. The goal is to handle the same messy variety a human technician can, as one step in larger automated lab workflows. Thanks to Eikon Therapeutics for working with us on this!

Scott calls out a trend of designing humanoids to be easy for isaac lab to manage instead of fixing the simulator or whatever Interestingly this was a problem for a while with a certain humanoid that I have worked with due to large closed loops in its legs Simulator tech is clearly going to bound the quality of robots we can build at some point