Friday 5.0: Scaling Laws, Robotics' Inflection Point, New Space, Neo Clouds and Engines That Move Markets
What I'm Learning this week: Some readings for your weekend from the world of Compute, AI, Robotics, Energy, & Materials
Table of Contents:
Topics I’m Studying
Videos to Watch
Startups I’m Excited About
Books to Read
Topics I’m Studying
Scaling Laws for Foundational Models.
Scaling Laws are the new “Moore’s Law”. Well, perhaps not apples-for-apples, but understanding AI Scaling Laws is becoming increasingly important as they say a lot about how AI capabilities improve across multiple dimensions, helping developers and investors alike think about breakthroughs and make strategic decisions - just like Moore’s Law has done for both hardware and software providers (read: Crucible Moments - NVIDIA) .
Very well worth reading the article from Semianalysis on Scaling Laws – O1 Pro Architecture, Reasoning Training Infrastructure, Orion and Claude 3.5 Opus “Failures”. The article offers a comprehensive look at both traditional and emerging scaling approaches, while showing how major AI labs are implementing these principles through innovative training and inference strategies. Also worth reading: DeepSeek Debates: Chinese Leadership On Cost, True Training Cost, Closed Model Margin Impacts
When will robotics really scale?
We’re seeing amazing progress in robotics these days, both for humanoids and other more specialized systems. The question is when we’ll read the true inflection point for acceleration. Ultimately, successful robotics is a data problem, more than a hardware problem.
The real winners will the companies with the best tech stack and the best model, ultimately being able to learn better and faster than their human programmers. Worth reading perspectives from both sides here, with the more cautious take The "ChatGPT Moment" for Robotics: Are We Ready for Scale?, versus more bullish analysts like ARK (How ARK Is Thinking About Humanoid Robotics) and Goldman Sachs (The global market for humanoid robots could reach $38 billion by 2035)
Overall, it seems more a debate of when rather (than if) we will have humanoids deployed at scale within the near term - and I know I’m trying to get on every early-adopter signup list myself!
Videos I’m Watching
Startups I’m Excited About
ORI (private GPU-cloud provider) based in the UK
Ori a neo-cloud startup that’s been grabbing some market attention recently. By enabling flexible GPU-compute, direct hardware access to physical GPUs, CPUs, and storage without virtualization overhead, providing raw computing power for demanding tasks, all based in the UK, they highlight the growing trend within Sovereign Cloud which ia becoming increasingly important in areas like Europe and Asia. If you’re in the space of training ML/AI models yourself, it’s definitely worth checking out their API end-points recently launched for model training. Also pretty cool to see How to run DeepSeek R1 on a cloud GPU with Ollama
Disclaimer: I’m a very small investor in Ori through a private angel syndicate
Books I’m Reading
“Engines That Move Markets: Technology Investing from Railroads to the Internet and Beyond”. Explores the history of technological innovations and their effects on financial markets over the last two centuries. The book provides a really good deep dive into recurring market patterns seen time and again when it comes to euphoria, desperation, and success in the space of technology investing. A “must-read” for anyone working in the Tech space, regardless of where in the Stack you find yourself at the moment
About Me
Working at the interface between frontier technology and rapidly evolving business models, I work to develop the frameworks, tools, and mental models to keep up and get ahead in our Technological World.
Having trained as a robotics engineer but also worked on the business / finance side for over a decade, I seek to understand those few asymmetric developments that truly shape our world
You can also find me on X, LinkedIn or www.andreasproesch.com