Discover more from Young Reacts
Young Reacts #226
I am hesitant to try new things. To mask my discomfort, I often say something like, “My time is worth more than that.” But sometimes, the best way to grow and learn is to step outside of my comfort zones.
That’s what I learned when I decided to take over the maintenance of my spa and pool. I had always hired a service company to do it, but I have become disappointed with the service. The company was often unresponsive, and my spa and pool water were never quite clean enough.
So, I decided to learn how to maintain my spa and pool myself. I was surprised that it wasn’t as difficult as I thought. I actually enjoyed learning about it and fixing things with my own hands. I realized I could do this and finally gave the service company the notice.
This experience made me realize how much I resist getting out of my comfort zone. Whether it’s maintaining a pool at home or coding in C# at work, I dread doing them before I first try and see what it’s actually like. But after overcoming that initial hurdle, I’m more confident about taking care of the house. That new scary thing is often not as bad as I expect it to be, and I hope to bring this experience to my work.
Software Engineering ⚙️
Here I thought prompt engineering was the future of software engineering. And, of course, LLMs are coming for those jobs too. In all seriousness, a research team built a model that, given an output, will create an instruction to make another LLM model produce a similar output.
In Chrome 113 (which is the beta version for desktop Chrome), you will be able to override a request header either via UI or code. This feature will be most useful when I need to test new infrastructure.
My coworker and I have been working to make our Typescript code stricter to avoid nullability-related bugs. Since our code has a few hundred files, fixing all issues in one go was impossible. We checked out some prior arts (one with Jest and one with a package structure) but were uncomfortable with the added developer friction. So instead, we decided to run a script to mask all existing failures with the directive ts-expect-error and fix them over time.
Engineers at Discord learned that Cassandra no longer fit their needs because of its operational toil, so they migrated to ScyllaDB. This doesn’t necessarily mean their first decision to use Cassandra was bad. Circumstances do change. The scale has increased, and technology has evolved. But it’d be good to revisit the decision-making process.
It’s unclear whether generative AI will assist humans in doing things better or replace them by doing things cheaper. This research finds the latter, unfortunately for us human workers.
When things aren’t working, you can turn around by focusing narrowly, getting small wins, and building momentum.
Even if the trained models do not generate music similar to its input, I think the music's copyright owner should decide if their music can be used to train a model, just as I should be able to determine the same for my personal information.
I am just amazed at how quickly Stripe adopted this technology in their business.