iPhone 17 Pro Runs 400B LLM, Walmart's ChatGPT Flops, and Starlette 1.0 Drops

Apple just demonstrated something wild at the iPhone 17 Pro launch: a 400 billion parameter LLM running locally on the device. That's not a stripped-down version. It's the real deal.

Meanwhile, Walmart's experiment with ChatGPT-powered checkout just went south. And Starlette 1.0 finally dropped, which is a bigger deal than most Python developers realize.

TL;DR: Edge AI is accelerating fast with phone hardware leading the charge. Walmart's failed AI checkout shows real-world integration isn't solved yet. Starlette 1.0 is the silent heavyweight of Python web frameworks.

iPhone 17 Pro: The 400B Milestone

The demo showed the iPhone 17 Pro running a 400B parameter model with impressive inference speeds. To put that in perspective, GPT-4o sits somewhere in that range, and it's running entirely on a smartphone.

This marks a turning point for edge AI. We're no longer talking about quantized 7B models that fit in RAM. The entire model lives on the device, no cloud required, no latency, no privacy concerns.

What changes? Offline assistants, privacy-first AI apps, and a whole new category of mobile applications that don't need internet to be intelligent.

Walmart's ChatGPT Checkout: A Reality Check

Walmart tried embedding ChatGPT into their checkout flow. The result? It converted 3x worse than their regular website.

That's a brutal data point. Users didn't want an AI chat interface for something as transactional as checkout. They wanted speed, clarity, and predictability.

The lesson here is obvious but easily forgotten: AI doesn't automatically improve every flow. Sometimes users just want to scan, pay, and go. Adding a conversational layer to transactional tasks doesn't add value, it adds friction.

This is the same trap many products fall into. Just because you can add AI doesn't mean you should.

Starlette 1.0: The Foundation You Didn't Know

Starlette 1.0 dropped last week, and if you're a Python developer, you should care even if you've never used it directly.

Here's why: Starlette is the underlying framework that powers FastAPI. It's the engine under the hood. Now it hit 1.0, which means stability, a real API, and a signal that it's ready for production workloads that go beyond being a dependency.

The framework is lightweight, ASGI-native, and handles routing, templating, and async database operations cleanly. If you're building something that needs speed and doesn't need Django's weight, Starlette is worth a look.

The irony is that Starlette has always been widely used relative to its brand recognition. It was the quiet workhorse. 1.0 changes that story.

Sources: Hacker News, Simon Willison, Cursor Changelog, Hugging Face Blog