- WANG Shudao
It was a week of sandstorms, and it wasn't until Sunday that the air quality index dropped below 100. One of the main reasons I fled Beijing in the past was smog, but this time the sandstorm affected the entire northern region, and there was nowhere to hide.
I had a cup of Luckin coffee almost every day this week, alternating between the ice-sucked coconut latte and the orange C Americano.
Photo taken in June 2013.
This week has been developing new features for version 4.3.0.
Benefits of Refactoring
Before, I was worried that I couldn't implement the design of the new version, but after the designer gave me the new UI, I immediately concluded that it was feasible, because building UI with SwiftUI is very convenient, like building with blocks.
Thanks to the results of the previous refactoring, I saved a lot of time when writing the pages for the Pomodoro clock and the stopwatch this time. I thought it would take two or three days to finish, but in reality, it only took a few hours. Each reuse of the carefully designed refactoring before caused a rush of endorphins.
However, there are still many pitfalls with SwiftUI, and I spent a lot of time this week solving an animation problem. There is a view that should move randomly every few seconds on the screen, but it didn't move at all at first. I thought it was a timer problem, but through logs, I found that the timer wasn't even triggered. Finally, I found that the root cause was that the parent view was constantly being redrawn, causing all child views to be regenerated constantly. So I spent a lot of effort to refactor the structure of the entire page, separating the view with animation from the parent view that was constantly redrawn. This not only solved the animation problem but also eliminated obstacles for adding other animations later.
On Tuesday night, I had a two-hour video conference with the designer using Midjourney to create the background image for the new version. Midjourney is fast in generating images, and sometimes the effect is amazing, but the premise is that you can accurately describe the scene you want, and it's not easy to modify the details locally.
We tried for a night and came up with five images, three of which could be used directly, and the other two needed the designer to modify them in Photoshop. Midjourney still didn't solve the last mile problem.
Claude & Albus
Another reason why I started using Slack again after many years is that Claude can integrate with Slack and is currently free and unlimited to use. It is said that Claude's current level is equivalent to that of GPT-3.5. I tried it briefly and found that its code generation is not as good as ChatGPT, and it often confidently talks nonsense.
On Sunday, I also integrated Albus, a Slack application based on GPT-4. It offers a 7-day free trial (with a limit of 100 questions) and is ready to use next week.
From now on, all app-related issues, such as development, ASO, and translation, will be handled in Slack, and I can also communicate with others.