Finally got my first Apple Watch App 🎉
Used two GLM-5.2‘s: one for controlling & tool approval, one for implementing.
Absolutely straightforward apart from the typical Apple stuff (had to accept a new license, had to prepare the Watch for dev mode.)
GLM implemented the scaffold (like companion and Watch app working, but only showing fake data) in about 20-30 minutes … fully unattended! 🤯
When it finished, I had a working project including screenshots from Watch app and Companion app.
Solved all issues interactively with getting it installed afterwards.
In a second unattended run over maybe 10-15 minutes, it added live data from my agent system.
Now I can already see the latest sessions on my Apple Watch 🤩🎉😎
I think I would use Claude Fable 5 for review only. All in all, working with GLM-5.2 for about a week or so: it feels more like Opus 4.8 than GPT-5.5 and currently, I prefer it’s output over both GPT-5.5 and Opus 4.8.
I guess it totally depends on your project and setup and my experience can’t be generalized. Still the very first real competitor to OpenAI and Anthropic, IMHO!
„One developer said: I think of myself as [a] mediocre engineer, and I feel this AI reset is giving me a chance to build skills that will bring me closer to excellence.“
Very interesting post on the upcoming seismic shift in software engineering.
https://t.co/Vv5paRRZaj
@DevDminGod@BenjaminNorton The ban could be worse. For instance, it could be banned on both play and apple stores worldwide. Leaving users unable to install the app and resort to the Web only. Android could sideload, Europe with alternative iOS stores could also have access. We just have to wait and see.
There is a Marburg outbreak in Rwanda.
Marburg and Ebola are very similar in their clinical and epidemiological features. Rwanda does have a strong health system (stronger than most african countries) with good leadership and so this helps.
Nigeria should begin surveillance.
I wanted to avoid this but ended up just building a html parser for my own specific needs. It only works with utf-8 encoded content. But overall performance is pretty alright. Handles Wikipedia 1MB content in 16ms. https://t.co/DcDPGA0Dp9
#buildinginpublic.