Rami's Readings #101 - ✨ Google & OpenAI Releases 🚀
The latest on AI, LLMs, Gemma 3, OpenAI's Agents SDK, Mistral, Anthropic, H Mart, and more.
Welcome to Rami’s Readings #101 - 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.
Apologies for the late-night delivery—I'm traveling in Europe this week, and jet lag got the best of me.
🤖 AI Reads
🔥 Google Launched Gemma 3
Notes: Multiple multimodal models released (try to say that 3 times fast haha!) with improved vision understanding abilities and 128K tokens. Read the technical report.
🔥 OpenAI Released Agents SDK
Notes: With companies like IBM, Microsoft, Amazon, and Anthropic rolling out their own SDKs, it's clear that 2025 will bring significant improvements in developer tooling for building AI agents. OpenAI’s SDK focuses on the fundamentals of orchestration between models, tools use, streaming, and handoff. Developer experience is critical for success as highlighted in #100. Official blog post.
OLMo 2 32B: Outperforms GPT 3.5 and GPT 4o mini
Notes: From hometown ❤️ Seattle, the Allen Institute for AI released the next version of their fully open source model.
Mistral Launches Mistral OCR
Notes: PDFs are everywhere! Every OCR model improvement unlocks more PDFs to AI workflows.
Latitude: Open Source Prompt Engineering Platform
Notes: I haven’t tried Latitude yet, but many of the important components are present, including supporting LLM and human evaluation of prompts.
GenAI & LLM System Design: 500+ Production Case Studies
Notes: Shared by a subscriber. The case study selection is pretty good!
LLM-Reasoner: Make any LLM Think Deeper
Notes: This is an open source project from MIT. It’s fairly basic, you can understand the entire codebase in a couple of hours. It’s an easy dive for those starting up in this space.
Fully In-Browser Graph RAG with Kuzu-Wasm
Notes: From University of Waterloo 🇨🇦, WebAssembly-based. Further evidence of a broader shift back to native code.
💼 Business Reads
How America Got Hooked on H Mart
Notes: We ❤️ H Mart! We were so happy when they opened a location in Redmond!
The ChatGPT Edge: Why Regular People Are Embracing $200 AI Bills
Notes: I previously discussed this development in #92, Prof Tucker is right: Value pricing is the correct strategy.
Is Zapier Obsoleted by Anthropic’s MCP?
Notes: A provocative post from VC, Tara Tan (Formlabs, IDEO, MIT).
🔀 Other Reads
A 10x Faster TypeScript
Notes: TypeScript is being rewritten in Go! In #100, I discussed the importance of engineering fundamentals and the shift back to native code. This is yet another example!
Succinct Data Structures
Notes: Y’all can probably tell that my current reads are shifting back into algorithms and CS fundamentals. Bit vector data structures are critical for AI.
Signing off from Greece 🇬🇷