Rami's Readings #122 - OpenAI Released GPT-5 and GPT-OSS 🚀
The latest on AI, LLMs, OpenAI, GPT-5, GPT-OSS, Open-SWE, Vintage Cars, Framework Desktop, Local-First Linear, and more.
Welcome to Rami’s Readings #122 - 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
Vacation Notice: I’ll be on holiday from August 25 to September 21, 2025. You may get a few surprise newsletters during this time, but I won’t be following my regular publishing schedule.
🤖 AI Reads
Introducing GPT-5
Notes: OpenAI finally released GPT-5, supposedly their one model to rule them all? As part of the release, OpenAI removed all their other models in ChatGPT, but then backtracked. The user backlash is unsurprising once you start playing with GPT-5, it’s mostly alright 😞 (not particularly 🤯). Polymarket now believes Google will have the best AI model at the end of the year.
Personally, I would prefer not to trigger reasoning on some of my long-running threads (cough… cough… my calorie tracking chat), so I too was annoyed that the model selector was removed (see the Reddit AMA for more outrage). That said, I think is right, more users will discover reasoning and come to expect it. This will raise the bar for all other experiences and ultimately benefiting OpenAI.
Also, what was going on with livestream benchmark diagrams? Makes me wonder if their marketing team used ChatGPT at all… For what it’s worth, GPT-5 fixed it.
OpenAI Released Open AI (GPT-OSS)
Notes: We can’t joke that Meta is the only one releasing open AI anymore… OpenAI launched two open-weight models (20B and 120B) on HuggingFace (available also on Ollama, LM Studio, etc). I never thought we’d see the day!
What I find most interesting is the quantization format MXFP4 (a fairly new floating point format) and the model’s tool-calling performance. Simon Willison has a good write-up on GPT-OSS, which also uses Harmony. Another interesting take: GPT-OSS is just Phi-5 (Sebastien Bubeck joined OpenAI 11 months ago). Nathan Lambert on GPT-OSS.
langchain-ai / open-swe
Notes: Open SWE is an open-source cloud-based asynchronous coding agent built with LangGraph.
Kitten TTS
Notes: Optimized for real-time speech synthesis, 15B parameters, CPU-optimized, and the entire model is 25MB.
🔀 Other Reads
Breaking the Sorting Barrier for Directed Single-Source Shortest Paths
Notes: Wow. This is a significant improvement for a foundational computer science problem.
Linear Sent Me Down a Local-First Rabbit Hole
Notes: I also went down this rabbit hole because the latency blew me away. We started using Linear at work.
How We Made JSON.stringify More Than Twice as Fast
Notes: SIMD for the win!
Introducing platformatic / mcp
Notes: A production-ready MCP server from Matteo Collina who led fastify development.
The Framework Desktop is a Beast
Notes: Birthday present anyone? 😊
💼 Business Reads
How Samsung Missed the AI Moment
Notes: Yep…
Carry Traders Ramp Up Bets in Emerging Markets as Fed Cuts Loom
Notes: Highly technical read, but interesting nonetheless.
So You Bought a Fancy Vintage Car. Now Who’s Going to Restore it?
Notes: My friends know I am obsessed with vintage Porsche 911s.
Signing off from Redmond, WA.