Sanjil Jain 🤖/acc

Presenting, @godrift_ai : Claude code for robotics simulations. We've been building robots for 10+ years and most of our time was spent fighting the tools and frameworks around it, not the actual robot use case. We didn't get into robotics to feel like an engineer debugging software and dependencies but because physics and the human brain fascinated us, to build things that move through the real world. At some point we stopped accepting that the setup was just "part of the job." It isn't. It's just friction nobody fixed. Yet.

People ask me why physical AI, why now. Because software ate the world. And the world still has a body.

@icanvardar Couldn't agree more!

listening to user feedback is one of the best things you can do as a builder

@TheGeorgePu 137x is the market saying the inference bottleneck is real and cerebras is one of the few bets that can actually move it.

Cerebras went public yesterday. Priced at $185. Opened at $350. Closed at $311. $510 million in revenue. ~$70 billion market cap. 137x sales. Nvidia trades at 26x. 86% of Cerebras's revenue came from two customers in the UAE. The market isn't pricing a chip company. It's pricing a belief system.

@Klajd_Lika the most cost effective solution is often the one closest to how the problem actually works. haptics just makes physical intuition transferable.

Is haptic teleoperation the most promising method to collect the right data for Physical AI? We don’t have an answer to this question but we know it works and is the most cost effective solution at the moment to program a robot! The Leader feels exactly what the Follower is feeling, and this is be transferred to a Physical AI. Complex trajectories over unstructured environments become much easier to program. We will be talking about this and other data collection methods and their future over the upcoming webinar with on 20th of May. The spots are slowly filling up. Make sure you subscribe ASAP so we prepare the ground. kinovarobotics.com/resource/relia… At Bota Systems AG we enable teleoperation, kinesthetic teaching, handheld devices and more with the most reliable force sensing solutions.

@XRoboHub 2,000 units in one order means this passed someone's ROI calculation. That's the real validation.

2,000-unit order: Zhejiang Humanoid just landed a major garment-manufacturing robot deal in China. 🧵 The buyer is Jack Technology, a major player in intelligent sewing equipment. The order covers 2,000 customized humanoid robots for apparel-factory scenarios. Garment manufacturing is brutal for robots because fabric never behaves like a fixed part. It bends, wrinkles, slips, stacks, and changes shape while the robot is touching it. For collars, pockets, and other cut pieces, alignment can be as tight as ±2 mm. Some cutting and sewing workflows need motion precision around 0.3–0.5 mm. That is why this order is worth watching. Zhejiang Humanoid’s NAVIAI is being pushed into template-machine work: opening templates, pinching fabric, separating layers, stacking pieces, and unloading material. The reported workflow has passed POC validation, with single-step operation under 10 seconds and multi-layer fabric alignment under 2 mm. Jack had already prepared the other half of the stack. It launched the Aitu brand, the Ai10 AI sewing machine, and showed a humanoid robot prototype before this deal. The signal is clear: sewing machines are getting AI, and humanoids are being added to handle the flexible parts machines still struggle with. For apparel factories, this is the real test. Small batches, many SKUs, short lead times, soft materials, and fewer skilled workers. If humanoids can handle fabric reliably, they stop being a demo. They become part of the line.

The robotics industry has been "5 years away" from mass adoption for 20 years. The difference now? simulation finally caught up.

"So many amazing founders get really undervalued because they are in the wrong market" - @nitinsharma1 At Embark 3, we are bridging the gap for immigrant founders building for the US through visibility, access, and impact. On Day 3, each of our 20+ founders got dedicated 1:1 time with the Antler team for an unfiltered and honest conversation about what it takes to win in the US market. See you on Day 4. 👀 @nitinsharma1 @jasnoorgill @GowriShankarNag @telljeeves @rschoo @NicolasSamDuval @nikhilkr97 @sanjil @experientia @ashishpiplani @AntlerGlobal

SF taught me one thing - everyone here has failed at something. And nobody is embarrassed about it. That's the actual culture shift.

Want to find out who actually supports you? Tell people you're building something.

First day, first show of @AntlerGlobal 𝗘𝗺𝗯𝗮𝗿𝗸 𝟯💥 We hosted @gilbenartzy from @upwestvc, backer of iconic companies in the Israel US corridor — SentinelOne, HoneyBook & more. Immigrant founders who cracked the US market. Gil was 1/8 people we met today. More soon :)

Some of the most ambitious immigrant founders in the world are in San Francisco this week. 🇺🇸 Yesterday was Day 1 of Embark 3, and it was a blockbuster! → 8 world-class speakers, including @GilBenArtzy from @upwestvc, backer of generational companies in the Israel-US corridor: @SentinelOne, @HoneyBook, & @Stampli and more → 20+ founders from India, Australia, SEA, MENA & Korea → Sessions on the mindset shift required to win in the US market and hyperscaling in the AI era. And this is just Day 1 of many. Stay tuned. 👀 @nitinsharma1 @jasnoorgill @GowriShankarNag @telljeeves @rschoo @NicolasSamDuval @nikhilkr97 @sanjil @experientia @ashishpiplani @AntlerGlobal

I visited the same Starbucks at Stanford I went to 7 years ago. They spelled my name wrong back then. Walked in last week, ordered again just to check and they still spelled it wrong. Some things are just consistent.

we <“> robots

People keep asking when robots will take over. I'm sitting here watching them struggle with tasks a 5 year old does without thinking. The gap between what robots demo and what they actually do in the real world is enormous. That gap is exactly what we're building @godrift_ai for.

In the first three days in SF I’ve been to two major robotics events - a pitch and a panel. Things move fast here and we simply ended up at both cause of “we're doing something on robotics, you should be in the room” I think when you're genuinely deep in a problem, you stop having to chase the rooms, the rooms start finding you.

7 years ago I visited Stanford on my first US trip. Today, I am here as a speaker and it feels unreal. Life comes a full circle.

We just won a robotics pitch competition at @UCBerkeley. But that's not the interesting part Before the judges even gave us scores, one of them - Matt Berggren, NeuroCAD’s CEO said he'd already used @godrift_ai. That meant more than the win honestly

now drifting in SF 🇺🇸 making sure robotics devs deserve the best with @godrift_ai wild how fast things move when you're locked in with the right person.

Actually, AI/Robotics will mean everyone can have a penthouse if they want. The output of goods & services will be several orders of magnitude higher than today’s economy. Read the Iain Banks Culture books for the best imagining of how it will be. That said, what is the future you want? Amazing abundance seems the best to me.

The product driven acquisition + retention is going to be substantial given distribution it getting smarter cause of better targeting. True on your part and also to add, it’s not about the best beer but about the person who is most likely to vote it as best gets pulled in smartly to try it.

@sanjil @r0ck3t23 When exactly was the moat only product? What’s the objectively ‘best’ beer? Or trainers? Or phone? The idea the moat becomes ‘just product’ ignores how markets actually work. People don’t buy the ‘best’ product. They buy the one they know, trust, and feel something about.

Mark Zuckerberg just quietly executed the marketing profession. He didn’t announce it. He described it. Zuckerberg: “The AI is actually probably going to be able to find who is going to be interested in your product better than you can.” He is not pitching a better ad tool. He is telling you that your intuition about your own customer is now inferior to his algorithm. For twenty years, the internet was built on demographics. Age. Location. Income. Founders built careers defending customer personas in boardrooms. Zuckerberg is telling you to burn them. Zuckerberg: “Don’t constrain who we’re going to reach.” When you tell the system who to target, you are not helping it. You are crippling it. Every audience filter a marketer applies is a ceiling disguised as a strategy. Human targeting is no longer a competitive advantage. It is a tax. And the takeover does not stop at the audience. It swallows the creative itself. Zuckerberg: “We’re going to be able to come up with like 4,000 different versions of your creative and just test them and figure out which one works best.” No agency on Earth can test four thousand variations in real time. The machine does not have taste. It has math. And math does not lose to intuition at scale. For a generation, distribution was the skill. When every company on Earth plugs into the same omniscient ad engine, distribution stops being a weapon. It becomes infrastructure. Like bandwidth. Like cloud compute. Nobody wins because they have better access to AWS. Soon, nobody will win because they have a better media buyer. Which leaves exactly one variable on the board. Zuckerberg: “You just focus on building the best product.” For a decade, mediocre products survived on superior distribution. Bad products with great funnels printed money. That arbitrage is dead. When distribution becomes a utility, the product has nowhere left to hide. Marketing was the mask. Zuckerberg just pulled it off. Most companies are about to find out they never had a product. They had a campaign.