February 11, 2026
Ep#62: PolaRiS: Scalable Real-to-Sim Evaluations for Generalist Robot Policies
RoboPapers
Key Takeaways:
- Generalist robot policies need robust, scalable evaluation. The shift is from bespoke, real-world-only testing to a hybrid real-to-sim approach that leverages modern 3D reconstruction and minimal sim data to create highly correlated, reproducible benchmarks.
- Builders should adopt PolaRiS's real-to-sim environment generation and "sim co-training" methodology. This allows for rapid, cost-effective iteration on robot policies, ensuring that improvements in simulation translate directly to real-world gains.
- Over the next 6-12 months, the ability to quickly and reliably evaluate robot policies in simulation will be a critical differentiator. PolaRiS provides the tools to build diverse, generalization-focused benchmarks, moving robotics closer to the rapid iteration cycles of other AI fields.
See full notes
February 11, 2026
The Autonomous Driving Race Is Already Over w/ Kyle Reidhead
Milk Road AI
Key Takeaways:
- Tesla's core identity shifted from EV maker to autonomous AI and robotics. Its cars are devices for deploying its advanced AI brain; competitors miss this.
- Tesla's 8 million cars collect real-world driving data. This massive dataset, combined with in-house AI processing, creates an unparalleled moat impossible for competitors to replicate.
- This convergence creates an abundance of labor and transportation, driving down costs. Robo-taxis and humanoid robots automate tasks, making goods and services cheaper, even as Tesla's profitability soars.
See full notesFebruary 11, 2026
Ep#62: PolaRiS: Scalable Real-to-Sim Evaluations for Generalist Robot Policies
RoboPapers
Key Takeaways:
- Robotics is moving towards generalist policies that need broad, diverse evaluation. PolaRiS enables this by making it easy to create and share new, correlated benchmarks, cultivating a community-driven evaluation ecosystem similar to LLMs.
- Adopt PolaRiS for rapid policy iteration on pick-and-place and articulated object tasks. Use its browser-based scene builder and existing assets to quickly create new evaluation environments, then fine-tune policies with a small amount of unrelated sim data to boost real-to-sim correlation.
- Investing in tools like PolaRiS now means faster development cycles and more reliable policy improvements. This accelerates the path to robust, real-world robot deployment by providing a scalable, trustworthy intermediate testing ground.
See full notesFebruary 11, 2026
Ep#62: PolaRiS: Scalable Real-to-Sim Evaluations for Generalist Robot Policies
RoboPapers
Key Takeaways:
- PolaRiS enables a shift towards LLM-style generalization benchmarks, where models are tested on unseen environments and tasks, accelerating robot capabilities.
- Use its browser-based scene builder and Gaussian splatting to quickly create diverse, real-world correlated evaluation environments, significantly reducing the cost and time of real robot testing.
- Cheap, reliable robot policy evaluation in simulation, with strong real-world correlation, means faster development cycles, more robust generalist robots, and a path to crowdsourced, diverse benchmarks that will push the entire field forward.
See full notes
February 12, 2026
Rivian’s Roadmap to AI Architecture and Autonomy with Founder and CEO RJ Scaringe
No Priors: AI, Machine Learning, Tech, & Startups
Key Takeaways:
- AI is forcing a fundamental architectural change in automotive, moving from fragmented, rules-based systems to vertically integrated, neural network-powered platforms. This technical reality dictates market survival, favoring companies that control their entire software and hardware stack to build a continuous data flywheel.
- Invest in or partner with companies demonstrating deep vertical integration in AI hardware and software for mobility. Prioritize those with a clear path to mass-market data collection and rapid iteration cycles.
- Autonomy will be a must-have feature in cars within the next few years. Companies without a software-defined architecture and a vertically integrated AI stack will struggle to compete, creating a market share shift towards those few players who can deliver true self-driving at scale.
See full notesFebruary 12, 2026
Rivian’s Roadmap to AI Architecture and Autonomy with Founder and CEO RJ Scaringe
No Priors: AI, Machine Learning, Tech, & Startups
Key Takeaways:
- The automotive industry is undergoing a fundamental re-architecture, moving from hardware-centric, rules-based systems to software-defined, AI-powered platforms. This shift favors companies with deep vertical integration and proprietary data flywheels.
- Invest in companies demonstrating full-stack control over their vehicle's software, hardware, and AI training data. This verticality is the moat against commoditization and the engine for rapid, continuous improvement.
- Autonomy will be a non-negotiable feature by 2030, making software-defined vehicles the only viable path for mass-market automakers. Companies that fail to build or acquire this capability will face market irrelevance.
See full notesFebruary 11, 2026
The Autonomous Driving Race Is Already Over w/ Kyle Reidhead
Milk Road AI
Key Takeaways:
- Tesla's core business is AI and autonomous robotics. This means its value comes from its software and data moat, not just vehicle sales.
- Tesla is sunsetting Model S and X production to convert factories for humanoid robots. This signals a full commitment to autonomous devices beyond cars.
- Unsupervised FSD is expected in select US states by Q2. This will enable cars to operate without human oversight, unlocking the robo-taxi network.
See full notesFebruary 12, 2026
Rivian’s Roadmap to AI Architecture and Autonomy with Founder and CEO RJ Scaringe
No Priors: AI, Machine Learning, Tech, & Startups
Key Takeaways:
- The automotive industry is undergoing a fundamental architectural shift from fragmented, rules-based systems to vertically integrated, AI-driven neural networks.
- Invest in companies demonstrating deep vertical integration in AI compute and data acquisition for autonomy, or those actively licensing next-gen software-defined vehicle architectures.
- The next 6-12 months will see accelerated divergence between auto players.
See full notesFebruary 12, 2026
Rivian’s Roadmap to AI Architecture and Autonomy with Founder and CEO RJ Scaringe
No Priors: AI, Machine Learning, Tech, & Startups
Key Takeaways:
- The automotive industry is undergoing a core architectural change, moving from fragmented, rules-based systems to vertically integrated, AI-native software-defined vehicles. This transition will consolidate market power around a few players who control their entire stack, from silicon to data.
- Invest in companies demonstrating deep vertical integration in AI hardware and software, particularly those with proprietary data collection and training pipelines. These are the players building defensible moats in the future of mobility.
- By 2030, self-driving capabilities will be a non-negotiable feature in every car. Companies that haven't fully embraced AI-native architectures and vertical integration will struggle to compete, making this a crucial moment for market share and survival in the auto industry.
See full notes
