M.G. Siegler about Apple’s AI competitiveness in Serious About Computing? You Should Build Your Own AI. — Spyglass

While they may look smart at some point for not pouring hundreds of billions into CapEx spend, that could come back to bite them in ways that are more tangential. Including, culturally, if the DNA of the company is never rewired to operate in the Age of AI. “People who are really serious about software should make their own hardware,” Alan Kay famously declared in 1982. What if the modern day version is something like: “People who are really serious about computing should make their own AI”?

I think we are still in the early days of generative AI. Siri’s failures apart, generative AI being so different than anything Apple did with Siri in the past, they can (and should) outsource their AI infrastructure. Eventually, they might bring it in, just like they did for Qualcomm modems, for example.

Forcing Micro.blog to rebuild the entire website to enable a design change is so slow; close to 30 minutes each time. 😳 One of the challenges is that not all design changes require a full rebuild; some do, some don’t, and I don’t always know in advance which ones do. Claude AI is not very good at predicting this either.

Since returning from vacation yesterday, I’ve been using Claude Code for different things, and yes, I can confirm that the credit consumption rate is much higher than it was a few weeks ago. 😳

I wonder what would happen if I tried ChatGPT Codex on one of my code base built with Claude Code. Could this bring some unforeseen areas of improvement? Could this be equivalent to human code review?

I’m currently on vacation but I keep an eye on the tech news and commentary landscape. With all the brouhaha surrounding Claude Code spotty performance lately, I’m starting to be wary of resuming my work on some of my projects when I get back home. 😟

OpenAI against the world:

Apple was seen as a laggard in AI development, and in fairness, they clearly saw themselves that way and invested billions in trying to catch up to the likes of OpenAI and Anthropic, but they failed miserably. And yet, they seem like they’re going to be doing just fine. You need a computer to do all this vibe coding on. You need a phone to talk to an AI agent. Who makes the best computers and phones? Apple does.

MP3 vs iPod playbook all over again? Or is it different this time? Not having to pay for all the necessary infrastructure to run AI might be beneficial for Apple’s future.

So, Claude Code source code has leaked, for real? 😳 Or is this a bad April Fool thing? Seems serious if true… how can they cope with this? Why did it leaked and how? It sounds like a disaster for the company.

Slash AI:

Personally, I couldn’t care less what you write on your /ai page. The same way I couldn’t care less if you use em-dashed. Words are cheap, easy to write, and they mean less and less. But your history, all the baggage you carry with you, all you have written and said, that is harder to fake, building it is time-consuming, but destroying it takes a second. If you start posting AI slop, my trust in you is gone in an instant, and no matter how you’ll try to justify it, that trust will not come back.

I can understand Manuel’s disdain of anything related to AI, but I feel his stance is a little too much “binary”; all or nothing. There must be a middle ground, isn’t it?

9to5Mac’s article “Vibe coding could mark the end of the App Store review process as we know it”:

Summary

The rise of AI-powered “agentic coding” has overwhelmed Apple’s App Store review process, with developers reporting review times of 3+ days to a week instead of the traditional under-24-hour turnaround. The influx of fully AI-generated apps from new developers has created a bottleneck for human reviewers, making it unfair for established developers whose update submissions are delayed. To address this, the author suggests Apple could implement separate review queues for established developers or automate updates while maintaining human review only for new submissions, though it may ultimately become necessary to reduce or eliminate full human review.

I don’t see the current review process at Apple as sustainable. I can imagine parts of the current workflow being automated (like finding instances of private API usage in application binaries). But, just for vibe coding, reviewing app submissions should be human-gated. An AI agent could even run the app in a simulator for testing.

I finally put together a video demonstration of my RSS Reader and Bookmark Manager. It’s a much longer video than originally anticipated, that is why it is being posted on my YouTube channel instead of Micro.blog. You’ll get to see both apps in action. I’m rather proud of thse apps, they are now essential for me.