Rami's Readings #120 - Corporate Verticalization
The latest on AI, LLMs, Embedding Models, Claude Code, Corporate Vertical Integration, the Hanseatic League, Rust on GPU, and more.
Welcome to Rami’s Readings #120 - a weekly digest of interesting articles, papers, videos, and X threads from my various sources across the Internet. Expect a list of reads covering AI, technology, business, culture, fashion, travel, and more. Learn about what I do at ramisayar.com/about.
👋🏼 Welcome New Subscribers
Hello! A hearty thank you for subscribing to Rami's Readings! There are quite a few new subscribers this week, thanks to recommendations from The AI Ethics Brief, Global Fintech Insider, and The VC Corner. I am thrilled to have you on board! In this newsletter, I curate the best papers, tweets, and articles I have read during the week focusing on LLMs, AI, economics, business, and technology news. You can learn more about me on my website.
📈 Top Recent Editions According to Substack
🤖 AI Reads
Voyage-Context-3: Contextualized Chunk Embedding Model
Notes: This model produces vectors for chunks that capture the full document context. Outperforms OpenAI-v3-large and Cohere-v4 embedding models.
Google AI Edge
Notes: LiteRT looks interesting, I wonder how it compares with ONNX. Performantly runs JAX, Keras, PyTorch, and TensorFlow models on Android, iOS, web, and embedded devices, optimized for traditional ML and generative AI.
Claude Code Sub Agents
Notes: I’ll be exploring this new capability this week.
💼 Business Reads
From Tesla to Microsoft, Companies Are Going Vertical Again
Notes: Are companies truly going vertical? I find it hard to believe there’s a grand strategy of vertical integration. A quote from the article:
Microsoft by extending its business from office computing “downward” to servers and renewable energy to power the servers and “upward” to gaming; and Tesla by coupling software and hardware. Amazon moved into the server business to ensure security of supply (and created a thriving business selling its surplus capacity in the process) and Google moved into chip-making for the same reason.
In nearly every technology case cited in the article, and from what I’ve heard from friends who were involved, the product teams were focused on solving specific technical challenges tied to their core business. Accidentally, they would over-perform and build a tool or capability that could later be externalized or monetized. For example, Bing got really good at crawling and processing documents at Internet-scale. It was needed to build a great search engine. Today, we have an externalized & monetized Grounding API.
I would add that tech companies actively recruit top technical talent from top research universities; people who are often inclined to pursue intellectually interesting side-quests (I’m certainly guilty of this). It doesn’t help that many of the big HQs (Microsoft, Google, Meta, Apple) intentionally create a university campus atmosphere, which encourages exploration and hallway conversations. And when you’re operating with high profit margins, it’s easy to justify giving employees the freedom to spin up side projects. How many messaging apps did Google employees create again?
From the outside technology companies may look like they are trying to be vertically integrated, but in my opinion it’s largely unintentional. What we’re really seeing is technology companies inadvertently building ecosystems around themselves, more like chaebols, keiretsu, or modern conglomerates or a mix of each. They’re just really good at solving complex problems and the new businesses often happen as a side-effect. Happy to hear alternative points of view.
Seattle Corporate Landscape
Notes: A map of Seattle-headquartered companies I found on LinkedIn.
The Rise and Fall of the Hanseatic League
Notes: A long read about this coalition of cities covering their fascinating history of trade and defections. The league is also the reason why these cities have spectacular historical architecture. See Bruges.
🔀 Other Reads
Rust Running on Every GPU
Notes: Cool demo repo showing you how to run the same Rust code on every GPU platform: CUDA, SPIR-V, Metal, DirectX 12, WebGPU and CPU fallback.
Keep Pydantic Out of Your Domain Layer
Notes: Yes… It should only existing on the edges.
Signing off from Redmond, WA.