Boris Sofman

Think cutting-edge means owning every asset? Waymo’s latest deal says otherwise. Yesterday, Waymo signed a deal that shows scaling robo-taxis is more about ops partnerships than flashy lidar. Here’s the story: - Waymo taps Avis Budget Group to manage, charge, clean, and service its entire robotaxi fleet in Dallas. - Fleet: 200+ all-electric JLR (Jaguar) I-PACE SUVs ready to roll by year-end. - Avis staff will handle 24/7 turnarounds, while Waymo keeps its focus on autonomy, routing, and rider UX. - This mirrors Waymo’s Phoenix model, where an external partner slashed downtime by >20%. Why it matters: Waymo stays asset-light by outsourcing fleet logistics to Avis, focusing on core AI while turning a legacy rental giant into the backbone of tomorrow’s robotaxi network. What about Avis? Avis lands a foothold in next-gen mobility, future-proofing a rental business that’s been squeezed by ride-hail for a decade. For founders, the message is clear: 1️⃣ Partnerships can accelerate scale faster than capital. Asset-light beats asset-heavy when the tech is your edge. 2️⃣ Focus on your differentiator. Waymo owns autonomy IP; Avis owns fleet ops. Everyone stays in their genius zone. 3️⃣ Legacy incumbents are leverage, not obstacles. The right deal turns yesterday’s infrastructure into tomorrow’s distribution. The next wave of mobility, and many other industries, will be won by orchestrating partnerships that let each player compound what they do best. 🔗 Read more → https://lnkd.in/dkMYgyFj 💬 Founders: if you outsourced your lowest-margin activity tomorrow, how much faster could you ship your core value? #Waymo #Avis #MobilityTech #FounderStrategy #MyLittlePursuits ________ ♻️ Repost this to share it with others. I’m Dar, founder of Little Pursuits. I help founders turn bold ideas into traction through strategic and leadership coaching. Follow for more insights like this.

7

2 Comments

Waymo has logged over 200 million fully autonomous miles on public roads, from complex city streets to highways. And yet heavy-duty trucking autonomy trails far behind, even though intuitively it seems like an easier challenge. Turns out autonomous trucking is hard. Eyal Cohen, CEO of Humble Robotics, thinks the answer is a combination of a fresh start (taking advantage of vision language action models and today's perception stack) and a clean sheet design (with a truck built from the ground up for autonomy). It's a big, ambitious vision and we at Energy Impact Partners are excited to be along for the ride. I talked to Eyal about the autonomous electric trucking space, and what it might take to catch up to passenger AVs. Listen: https://lnkd.in/gt5i3nRw

49

2 Comments

How do hardware founders actually find their first manufacturing partner? In my experience, it’s rarely a straight line. Most don’t land their contract manufacturer from a Google search (sometimes they do). It usually comes from a mix of networking, testing, and a little trial by fire. Some of the common paths: ✅ Referrals and networks: trusted introductions from other founders or investors. The upside is trust, but the downside is your options are limited to who you know. ✅ Design firms and prototyping studios: they often recommend the CMs they’ve worked with. Helpful for handoff, but sometimes the recommendation is more about their relationship than your perfect fit. ✅ Industry events and trade shows: CES, CEDIA, HardwareCon, etc. Great to meet multiple manufacturers face to face, but overwhelming if you’re new. ✅ Online research: broad reach and easy to start, but filtering quality from noise can be tough. Cold outreach doesn’t always get replies. What I’ve noticed is most successful founders start with a referral, validate through an event or industry connection, visit and audit, and then scale with the partner that proves the most aligned. If you’ve built hardware before, which path worked best for you? #manufacturing #electronics #hardware #nearshoring

31

19 Comments

MLI portfolio company Robotruck continues to hit key milestones as it moves closer to bringing autonomous driving to the real world. https://lnkd.in/g3HSx4Db

5

Less Artificial More Intelligence Driving the Autonomous Future In the backdrop of the GenAI Week, I reflected on the experience of what the future of transportation feels like. From the driver’s seat of a Tesla equipped with Autopilot to the passenger seat of Waymo’s fully autonomous taxi, it felt like being behind the wheels of tomorrow. Autonomous vehicles are perhaps the most powerful application of AI in the physical world today. No pedals. No human steering the wheel. Just you, the car, and a system that has seen and learned from more scenarios than any one human ever could. These cars handle highways with seasoned confidence. They navigate intersections, adapt to weather conditions, and react to the unexpected, all without showing fatigue or hesitation. It was eerie at the start but as it navigated itself beautifully, it was humbling and inspiring at the same time. What struck me most wasn’t the absence of a driver, but the presence of intelligence. The Financial Crime Perspective As a founder building AI products in Financial Crime Management, the technology in autonomous vehicles felt oddly familiar. What these cars do on the streets, we do behind screens: deciding in milliseconds whether to approve a transaction, flag a terrorist risk, or identify suspicious behavior. We are not dodging pedestrians or navigating intersections, but we are navigating risks, fraud, compliance and human behaviors. The ‘road’ our models travel is built from transaction histories, user intent, regulatory landscapes, and trust signals. The edge cases we face aren’t weather conditions or pedestrians, but synthetic identities, money laundering attempts, terrorist financing and sudden shifts in behavior due to life events that are unpredictable and dynamic. Autonomy in financial crime technology is quieter, but no less complex and the challenges are similar… - Handling edge cases - Real-time decision-making under uncertainty - Balancing automation with trust The lesson I learned from both these technology driven futuristic vehicles? Trust is earned through predictability and safety. Whether navigating traffic or transactions, AI systems must… - Understand context - Adapt to the unexpected in real-time - Deliver reliable outcomes - Provide transparent explanations for their actions In the end, I realize that any Artificial Intelligence system must feel less artificial and more intelligent. Thank you Satheesh for generously sharing our conversation on trust in AI and also trusting me with your car. #AIinAML #LessArtificialMoreIntelligence #FinancialCrimeCompliance #AntiMoneyLaundering

80

What if the biggest bottleneck in autonomous driving isn’t more data—but using it smarter? Rohan Bhasin, Senior Solutions Engineer for Sensor Simulation at Foretellix, and Daniel Gural, Head of Technical Partnerships at Voxel51, break down how neural reconstruction and scenario-driven data curation help AV teams save time, target critical edge cases, and scale safety faster. 🎧 Listen to the episode: https://apple.co/4cbvbnS

1

The biggest engineering problem that companies are trying to solve w/ large scale model training? It's all about how to make something work on thousands of machines as efficiently as it can work on two. Ceramic.ai's founder Anna Patterson shares her thoughts with Prateek Joshi on the Infinite Curiosity podcast, where she discusses opportunities in long context training in industries like pharmaceuticals and banking, and why the current training stack needs to be rebuilt. You can listen-in and watch the full episode here: https://lnkd.in/ggvQxN6h cc: Laura N Simon, Ph.D. Autumn Yuan, Riley Munks, Maggie Philbin, Hannah Colson

7

Uber Is Investing $100M to Power Robotaxis Uber is putting $100M+ into building a dedicated autonomous vehicle charging network, starting in the Bay Area, LA, and Dallas. It’s partnering with EVgo Inc. in the U.S. and Ionity in Europe, while expanding robotaxi operations with Waymo and WeRide. This isn’t just about EV charging. It’s about securing the infrastructure layer of autonomy and especially as Tesla Inc. ramps up competition. Autonomy is becoming an infrastructure game. #AutonomousVehicles #Robotaxi #EV #FutureOfMobility

29

4 Comments

My (biased) takeaway from Jensen Huang's comments at GTC is that software tools will continue to be valuable. If you're designing chips you will use CAD tools. If your AI agent produces an image and not a CAD asset, it's useless. You can't verify it, you can't modify it. The same holds for most of knowledge work. We give experts a compressed representation of high complexity information so they can verify it and modify it.

5

🚗 Waymo’s Autonomous Driving: A Technical Overview & Safety Impact Waymo Driver has accumulated 96 million rider-only miles as of June 2025, delivering compelling safety improvements over human drivers - 91% fewer serious injury or worse crashes, 79% fewer airbag deployments and 80% fewer injury crashes overall, validated through rigorous comparisons using NHTSA’s Standing General Order crash data. At the core of Waymo’s technology lies an advanced AI and machine learning framework that integrates: • Real-time sensor fusion: Combining lidar, radar, and a 360° multi-camera system to construct a rich, high-fidelity 3D environmental model • High-definition mapping: Detailed geospatial data of roads, signals, and key driving landmarks to enhance localization accuracy • Predictive modeling: ML algorithms anticipate behaviors of vehicles, pedestrians, cyclists, and other road users for proactive decision making • Trajectory planning: AI computes safe and efficient navigation paths (perception, prediction, planning) dynamically in complex and evolving scenarios This combination enables Waymo’s fully autonomous system to operate safely without human intervention, using continuous learning from billions of simulated and real-world miles to adapt and improve. #ai #artificialintelligence #machinelearning #autonomousvehicles

3

Today at the Robotics Summit in Boston, Brian Gerkey (Open Robotics, Intrinsic) opened with a keynote called "An Open Foundation for the Age of AI-Powered Robots." 👀 His argument, paraphrased: proprietary stacks introduce vendor lock-in right when the hardware itself is commoditizing. The future of robotics looks more Android than iPhone — a horizontal foundation, not a single sealed box. This is the thesis we're building on at vfrog. ⚡ The hard part of physical AI isn't the model. It's the loop around the model: ✅ Collect → Train → Eval → Deploy ✅ Every team rebuilds it from scratch ✅ Quan Vuong (Physical Intelligence): "There wasn't a company that offered this kind of services… so we ended up writing a lot of the software at Pi ourselves." That quote isn't a complaint. It's a market signal. Sculley's 2015 Google ML paper put engineering time on plumbing at 90%. TDWI's 2026 survey puts it at 61%. Eleven years, billions in ML infra spend, and the tax barely moved — because Physical AI is a new generation with new infra needs. 🔧 The horizontal layer — the managed data loop that runs on your stack, exposed identically via API, MCP, and CLI — is where vfrog lives. 🎯 The question isn't "is physical AI happening." That's settled. The question is: who builds the Android of robotics infrastructure before every team rebuilds it again? If you're shipping vision models and tired of being the in-house ML infra team, we'd love to talk. → vfrog.ai

2

𝐖𝐡𝐚𝐭 𝐦𝐨𝐬𝐭 𝐚𝐮𝐭𝐨𝐧𝐨𝐦𝐲 𝐡𝐞𝐚𝐝𝐥𝐢𝐧𝐞𝐬 𝐦𝐢𝐬𝐬 𝐚𝐛𝐨𝐮𝐭 𝐰𝐡𝐞𝐫𝐞 𝐫𝐞𝐚𝐥 𝐯𝐚𝐥𝐮𝐞 𝐢𝐬 𝐛𝐞𝐢𝐧𝐠 𝐜𝐫𝐞𝐚𝐭𝐞𝐝 The autonomy conversation is still anchored around public roads, robotaxis, and regulation. That’s where the headlines are, but it’s not where most of the commercial value is being created today. The industry is over-indexing on what is technically hardest, and underestimating what is commercially happening now. Private-site operations are not a simplified version of autonomy. They are where autonomy is already delivering real-world outcomes. For example: 🚚 Distribution centres 🏗️ Industrial yards 🌲 Mining and forestry environments These are active, revenue-generating environments where autonomy is being deployed, measured, and scaled. The question in these settings is not, “Can this work in theory?” It’s, “Does this improve throughput, reduce cost, and operate safely every day?” That changes and influences how systems are built. After years of deploying in these environments, one thing is clear, progress in autonomy is not just about solving the hardest technical problem, it’s about solving the problems that matter commercially…first. At NuPort, that has shaped our focus from the beginning. 🚛 Real deployments ⚙️ Real operations 📏 Real constraints Why? Because autonomy doesn’t need to be perfect to be valuable, it needs to be reliable where work is already happening. Not because they are easier, but because they are where autonomy becomes useful. #Autonomy #AI #Logistics #SupplyChain #IndustrialAutomation https://lnkd.in/eYf8EVsZ

22

🚗 Mobility, AI & Systems Digest From 🤖 robotaxi scale to 🧬 regenerative biology and 🌍 infrastructure shifts — this week shows something important: Scale is replacing speculation. 💬 Quote of the Week “We are called to be architects of the future, not its victims.” — R. Buckminster Fuller 🚖 Future of Mobility 🚕 Waymo CEO predicts 1 million weekly rides by the end of the year. – The Verge 🔗 https://lnkd.in/eV8nVMPA 🏎️ Ferrari Luce: a first look at the Jony Ive and Marc Newson-designed EV – Wallpaper* 🔗 https://lnkd.in/ej9FbFvZ 🚦 Glydways to ‘redefine mass transit’ with autonomous EVs – Electrek.co 🔗 https://lnkd.in/erf5EhFE 🚗 Hyundai is reportedly looking to supply Waymo with 50,000 IONIQ 5 for robotaxis – Electrek.co 🔗 https://lnkd.in/eFr8x5hj 🚖 Pony AI and Toyota roll out mass-produced robotaxis in China – South China Morning Post SCMP 🔗 https://lnkd.in/ecTwkSmJ 💻 Software & Platform Shift 🧠 Rivian was saved by software in 2025 – TechCrunch 🔗 https://lnkd.in/eukcnyFy 📊 Geotab Is Moving From Tracking Fleet Vehicles To Tracking Everything – Forbes 🔗 https://lnkd.in/e9X2dGpa 🌐 Google Chrome ships WebMCP in early preview – VentureBeat 🔗 https://lnkd.in/ebMgRHY8 🔘 Human Factors 🔘 Physical Buttons Are Making a Comeback in EVs – Business Insider 🔗 https://lnkd.in/eMnVk_Yk 🚪 Carmakers Rush to Remove Chinese Code Under New Federal Rules – Autoblog 🔗 https://lnkd.in/ePMqzDCv 🧬 Science & Systems 🧠 Scientists Grew Mini Human Spinal Cords, Then Made Them Repair After Injury – ScienceAlert 🔗 https://lnkd.in/e4Ek6p4h 🌲 Forests are changing fast and scientists are deeply concerned – ScienceDaily 🔗 https://lnkd.in/eyfzWcZH ⚛️ Europe Launches Euro-Q-Exa Quantum Computer – Quantum Computing Report 🔗 https://lnkd.in/eYPcMXt7 📊 Stat of the Week 📈 Global equity funds saw a net inflow of $25.54 billion last week, led by record inflows into European equity funds. Reuters 🔗 https://lnkd.in/eNjzcgsD

4

Is Waymo’s growth in California flattening? 🤔 The data sure looks like it.. In our latest piece on The Driverless Digest, we examine Waymo’s growth trend in California based on the recently released trip data from the CPUC, and a couple of things stood out: 📉 California weekly ride growth slowed sharply between January and March 2026, with weekly ride additions falling from +14.8k to just +2.7k rides/week 📊 That represents an 82% deceleration in the pace of growth from January to March. 🌎 Meanwhile, growth outside California appears to be accelerating, with non-California weekly ride additions jumping from -448 in January to +11.6k rides/week in March 🚕 Waymo now operates across 11 US cities with roughly 3,000 robotaxis and more than 500,000 paid rides per week ⚖️ The slowdown could point to several factors: fleet constraints, expansion trade-offs, a strategic shift toward newer cities, or challenges moving beyond the early adopter phase 📍 California may still be growing, but the data suggests it’s entering a more mature phase where gains come more incrementally than explosively Check out the full story here 👉 https://lnkd.in/gwPsbqbE

10

5 Comments

Putting Antioch in the hands of customers and partners over the past year has made one thing clear: closing the feedback loop — so AI can improve systems autonomously — changes everything. In software, we take this for granted. Cursor, Claude Code, and Codex are part of how we work every day. In the physical world, it hasn’t been possible. Until Antioch. Alex captured a lot of what we’ve learned in this post. Worth a read.

11

1 Comment