Notices by JP Simard (jpsim@hachyderm.io)

I initially wrote off Apple’s integration with ChatGPT as an admission of defeat, that they couldn’t develop an LLM competitive with GPT-4o or Gemini or Claude despite having near infinite resources, powerful ML co-processors in their hardware lineup going back years and some very bright people.

But now I’m beginning to see that Apple’s strategy is actually kinda brilliant in unexpected ways.

(1/3)

1. Offer decent local & server LLMs on their private cloud that maybe aren’t industry leading for generic tasks but are super well tuned to Apple-specific workloads.

2. Offer free access to other companies’ AI models for more general requests FOR FREE, offloading the compute costs *and* deflecting some liability away from them.

(2/3)

Apple is letting the rest of the industry burn money and duke it out while providing a ton of value for their customers. This has echoes of its approach to integrate 3rd party search providers, with a 2024 AI craze twist.

Apple is playing the long game here.

(3/3)

Anyone else running into stack overflows very easily with Swift Concurrency? Its cooperative thread pool appears to have a relatively tiny stack size of 512K on iOS. A few large structs and things go kablooey💥

This is a leading cause of crashes in our app and it’s been hard to fix holistically.

Hat tip to @kabiroberai for figuring out the stack size from spelunking in the libpthread sources.

@inthehands you can’t just waltz in here and post slurs, buddy, there are RULES on this platform

I was laid off from Lyft today, so I'm looking for work!
If you're hiring remote or in Montreal, hit me up.
12 years iOS experience, 9 years Swift.
I've worked on low level stuff (networking & databases), product stuff (built transit, bikes & scooters into Lyft), tooling stuff, open source stuff and lots of other stuff.