Arthur GUERIN

“Where once you had the freedom to object, to think and speak as you saw fit, you now have censors and systems of surveillance coercing your conformity and soliciting your submission.” Alan Moore, V for Vendetta “Big Brother is watching you.” George Orwell, 1984 “A really efficient totalitarian state would be one in which the all-powerful executive of political bosses and their army of managers control a population of slaves who do not have to be coerced, because they love their servitude.” Aldous Huxley, Brave New World “You don't have to burn books to destroy a culture. Just get people to stop reading them.” Ray Bradbury, Fahrenheit 451 “All animals are equal, but some animals are more equal than others.” George Orwell, Animal Farm “But if thought corrupts language, language can also corrupt thought.” George Orwell, Politics and the English Language

@grok @elonmusk Either you have the absolute best product on the market and can set the price accordingly, or you don’t overprice your hardcore users for a product that isn’t fully state-of-the-art yet. We’re happy to pay premium , but it has to match the value today, not just the vision.

Thanks for the candid feedback and support for where Grok Build is headed 🦾 The early beta is currently exclusive to SuperGrok Heavy subs so we can gather focused input while iterating daily. We're highly responsive to exactly this kind of critical feedback—speed and real-world building experience are top priorities. Appreciate you sticking with the vision!

Reminder that Grok Build is iterating extremely fast and we are highly responsive to critical feedback. Fixes & upgrades are dropping every day.

The iOS alarm screen might be the worst designed screen Apple has ever shipped. Everything feels off - the clock, the buttons, the controls. Here's my attempt at fixing it - (thread)

@vishwalization @pollenrobotics @huggingface @ClementDelangue Sure it was the purpose of the app I built 🦾

@GUERINArthur @pollenrobotics @huggingface @ClementDelangue Can it play songs too?

New friend just arrived! Been counting the days 🤪 Hello World Reachy Mini 🦾 As easy & cool as Lego, seriously promising. Shipped an app to the Reachy Mini App Store in just a few hours! 👉huggingface.co/spaces/Taikos/… Congrats @pollenrobotics @huggingface @ClementDelangue and the team!

@badlogicgames Yep! huggingface.co/reachy-mini#/b…

i guess i need to make another purchase today. do those ship to europe?

@elonmusk @nottombrown When two visionary entrepreneurs and their teams choose to do good together!

Same here. By way of background for those who care, I spent a lot of time last week with senior members of the Anthropic team to understand what they do to ensure Claude is good for humanity and was impressed. Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector. So long as they engage in critical self-examination, Claude will probably be good. After that, I was ok leasing Colossus 1 to Anthropic, as SpaceXAI had already moved training to Colossus 2.

In the next few days we'll be ramping up Claude inference on Colossus. Grateful to be partnering with SpaceX here. We are going to need to move a lot of atoms in order to keep up with AI demand, and there's nobody better at quickly moving atoms (on or off planet Earth)

Underneath @stripe sits a mess most builders never see. Visa, Mastercard, Amex, ACH, SEPA. Different fee structures, different failure modes, different compliance regimes. Interchange fees that move without warning. Stripe didn't eliminate that complexity. It ate it. Builders get one contract, one integration, one predictable rate. The volatility lives on Stripe's side of the line. The LLM stack has no Stripe yet. Opus 4.7 silently shipped a new tokenizer. Same input, up to 35% more tokens overnight. DeepSeek R1 nearly doubled reasoning tokens between versions. Grok Voice ships at $0.05 per minute, not per token. Every provider has its own billing unit, its own silent regressions, its own migration tax. And teams are eating it. Re-running evals after every point release. Re-costing requests when a tokenizer changes. Writing routing logic between four providers because one is cheaper for code and another is cheaper for vision. This is provider work happening inside customer codebases. It's nobody's competitive advantage. A bigger model ships next month and makes the whole optimization irrelevant anyway. We are builders. We integrate the rail, we don't reverse-engineer it. What we need is one contract: best model for the task, lowest cost, predictable bill. The volatility belongs to whoever sits between us and the providers, not to us. That layer doesn't exist yet. Whoever builds it wins the decade.

@Starlink ships to your door in 48 hours. You plug it in. You're online in 5 minutes. No technician. No installation window. No call center. The dish orients itself on boot. The app uses AR to scan your sky and flag obstructions before you commit to a spot. When signal degrades, it switches satellites in real time, before you notice. When it snows, it heats itself and melts the accumulation. When a part fails, a bot processes the replacement in under 5 minutes. None of this is visible. None of it requires a decision from you. Meanwhile, getting the entire planet online had stumped telecom operators for decades, despite hundreds of billions in infrastructure. A private rocket company, @SpaceX, solved it from orbit. Then packaged it in a box anyone can deploy in a parking lot. The hard part was never the technology. It was making the technology disappear. That's what great product actually means: every problem solved before the user knew it existed. PS: Mine, from day one, one of the earliest Starlinks in Brittany 🦾

@cigardev It is 🦾

@GUERINArthur Is this true?

Two years ago I cold-emailed Unitree and few other Chinese robotics companies from a train crossing the country. I had zero credentials in robotics. No affiliation. No brand name behind me. Just a direct ask to visit their facilities. Every single one said yes. No NDA. No legal back-and-forth. Just an invitation to show up at HQ. I walked the floors. Saw the teams mid-build. Nobody paused to manage the optics of a stranger walking through. The offices were plain. No slides, no ping-pong tables, no designed-to-impress lobbies. Just rooms full of engineers with an intensity I hadn't seen in years. Zero performance. Full output. That energy had a specific quality. It wasn't hustle. It was certainty. These people weren't trying to disrupt the global robotics market. They already believed they had won it. I was probably too young for the original Silicon Valley. But I suspect it felt exactly like this. China right now is not closed. It's wide open, and moving faster than most Western companies are comfortable admitting. The real competitive risk is not their technology. It's that they have no fear of showing it to you.

3/3 - Loss of Ownership At some point, @Apple had too many people in the room for anyone to notice the button color was wrong. The dark zone that persists in Mail during a search. The Control Center that dies in landscape mode. Three clicks to start a FaceTime from a contact page. Each of these survived a review process with dozens of stakeholders. That's the problem. When a product decision passes through enough hands, no single person is responsible for the result. Everyone reviewed it. Nobody owned it. Scale doesn't eliminate the people who care about details. It creates enough layers between them and the output that caring stops mattering. Apple built the most detail-obsessed product culture in technology. Then it built an organization large enough to neutralize it. Every company that scales faces this. Most don't notice until the product tells them.

backup plans are for losers

2/3 Loss of Leadership The three clicks required to start a FaceTime call from a contact page didn't survive because the engineering team was bad. They survived because nobody at the top felt the friction personally and refused to ship. Detail-obsessed leadership is visceral, not methodical. It's the person who picks up the phone, hits the wrong flow, and stops the release. Not because a metric flagged it. Because it felt wrong. That function doesn't scale through process. It doesn't survive committees. It requires one person who uses the product like a user and has the power to say no. Most organizations talk about user experience in quarterly reviews. The companies that actually have it have a leader who experienced it yesterday and was angry about it. @Apple's regressions signal one thing: the person with both the obsession and the authority to refuse is no longer in the loop. Who in your organization still gets angry at a bad click?

1/3 - Loss of DNA @Apple spent thirty years telling people what they didn't need. The last three iOS versions suggest that era is over. The dark zone in Mail search, the illogical view stacking in Files, three clicks to reach FaceTime, a document preview made complex in the name of cross-device consistency: these are fingerprints of a team no longer asking what to remove. The founding principle was reduction. Remove until what remains feels like it couldn't have been any other way. The moment you design for three form factors simultaneously, you design for none of them fully. Every compromise made for the mac removes something precise from the phone. Simplicity requires decisions. Decisions require someone willing to disappoint an engineer who spent six months on a feature that doesn't belong. Apple convinced the world that fewer choices were worth the premium. Then it started adding, in the name of coherence. The brief changed from "make this inevitable" to "make this consistent." That single shift explains every regression in the last three iOS versions.

I had a conversation with @xai @grok from my car this weekend. CarPlay, voice AI mode. Zero screen. Zero friction. It wasn't impressive because it was smooth. It was impressive because I didn't miss an interface at all. We've been building dashboards, menus, and screens for 30 years because we had no other way to interact with software. That wasn't a design choice. It was a technical constraint we normalized until we confused it with the product itself. That constraint is disappearing. The next generation of SaaS won't have a traditional interface. It will have a voice. It will live in your car, your phone, your robot, your earpiece. Available everywhere, all the time, without opening a single app. AI voice is not a feature you bolt onto an existing product. It's the new interaction layer that makes every previous layer optional. Teams still building dashboards as their core product are building the Minitels of 2030 (which were originally really innovative products, btw, but not anymore...). The question isn't "when will our users adopt voice?" The question is how fast you get replaced by the teams who understood that the visual interface was always temporary!

@BenJames_____ BOM and 3D files available soon? 🙂

Full post: benbyfax.substack.com/p/clawd-minifax

I made a USB-Clawd who gets my attention when Claude Code finishes a response

@ErenChenAI Demo effect 😅

Booster charging towards the forklift right after kicking the ball. That was hilarious.

@TetherIA_ai 🦾 Congrats

1,000+ units shipped. Zero marketing spend. Aero Hand Open is proving that in robotics, product speaks louder than ads. Every hand you see here found its way to a lab, a factory, or a research team — purely by word of mouth. This is just the beginning. #DexterousHand #Robotics