-
product · OpenAI
Rapidly scaling online storage to serve over 1 billion ChatGPT users
OpenAI evolved Habitat from a Python library into a globally distributed storage platform. The system now serves over 1 billion ChatGPT users and handles 22 million requests per second.
This gives engineers building large-scale AI services a concrete reference for designing storage architecture that scales to tens of millions of requests per second.
-
industry · Hacker News
Nvidia is the central bank of AI
A post on Hacker News describes Nvidia as "the central bank of AI." The title highlights Nvidia's outsized influence over the AI industry.
It points to how dependence on a single dominant GPU supplier shapes infrastructure planning and cost risk for developers.
-
release · Hacker News
OpenAI’s Navier-Stokes release included a Lean 4 formal proof
OpenAI's release related to Navier-Stokes reportedly included a formal proof written in Lean 4. Discussion of the release is underway on Hacker News.
Pairing AI-generated math results with formal proof verification points developers toward new tools for checking the trustworthiness of AI outputs.
-
research · Google Research
ToolGrad: Efficient tool-use dataset generation with textual "gradients"
Google Research introduced ToolGrad, a method for efficiently generating tool-use datasets using textual "gradients." Further details were not disclosed.
This offers developers building tool-use capable models a new, potentially cheaper approach to constructing training datasets.
-
product · Simon Willison
Quoting Boris Cherny
Anthropic's Boris Cherny says production code written by Claude should meet a higher bar than human-written code. The company relies on guardrails including lint rules, tests, Claude-driven end-to-end tests, daily Claude-powered fuzzers, automated code reviews and security reviews, and automated code refactoring.
Teams adopting coding agents like Claude Code should factor in that without these layered guardrails, agent-generated code can become an unmaintainable mess.
-
industry · TechCrunch AI
OpenAI’s Sam Altman says it would be ‘ill-advised’ to go public in 2026
OpenAI has filed confidentially for an IPO, but CEO Sam Altman says the company won't go public this year. He called it "ill-advised" to pursue an IPO in 2026.
A delayed OpenAI IPO timeline affects expectations around the company's funding capacity and long-term stability for developers building on its platform.