@reconbot Even still, I'm impressed by the speed! Even Epic Games said they needed a week to launch Fortnite after the ruling

What about frameworks like React Native? Looking at Spotify’s job postings, that goes right out the window. They hire iOS developers for actual iOS developer skills, and not for React Native. The only bonus skill they list is C++ experience, presumably because they share code between platforms using a C++ binding.

That only leaves one possible conclusion, which is in the post.

Also, I went an entire post and an entire note without confusing “Spotify” and “Shopify.” Please clap.

For some background, this order came on April 30th, and 1 day later on May 1st Spotify shipped their update. This is unusually fast for a major corporation; you’d expect that they’d need product, design, legal, developers, QA testing, legal review, etc. So over a 24-hour period, you’d want someone to be present from each discipline. It seems tight for this to be done, although it’s hard to argue with the fact that a talented mobile developer likely could develop the screens in that time

How did Spotify ship a release 24 hours after a judge ordered Apple to loosen its restrictions on alternative payment methods within the app store?

https://www.clientserver.dev/p/how-did-spotify-ship-that-ios-payment

@reconbot I've seen these graphs before but I never knew what it was called before today. It looks like it would have been the way to go, though

The fix took us through shredding a backpack, play with BIOS and Windows Registry settings, building a custom battery pack to prevent the machine from being power throttled, and finally a tour through distributed queuing. And we had given up on fixing it, until I suddenly had a flash of inspiration after three months.

When I had only been working in the software industry for a couple of years, I worked on an augmented reality training simulator for DARPA that required an incredible number of fixes.

https://www.clientserver.dev/p/war-story-i-fixed-this-bug-after

Bazel has taken some strides towards integrating with the rest of the world. For example, it’s easier to integrate with language-native build systems in C/C++ and frontend development than it used to be. However, it feels like it remains targeted at large organizations.

This is because Bazel wants everything done The Bazel Way, and the rest of the world has diverged from The Bazel way. Bazel wants everything spelled out perfectly: every dependency and every output should be clearly identified. But the rest of the world has settled on Convention over Configuration, layers of compilers, and command runners. As the years go on, it gets progressively more difficult to integrate new stacks with Bazel.

Bazel recently had the 10-year anniversary of its announcement.

I’ve used Bazel a lot, both personally and professionally. I also used Blaze when I was at Google, and it was magical. I always hoped that Bazel would become a viable build system for everything, so that I could recapture the magic of Blaze. But unfortunately it never really came to pass. I wrote a newsletter about why that is and how Bazel is still improving towards the ideal anyway.

https://www.clientserver.dev/p/bazel-is-too-pure-for-this-world

@reconbot TIL! How is everyone taking it?

Gaming the promotion criteria means that you pathologically do the things that are called out in your career ladder, and avoid doing anything else. If you’ve worked in Big Tech for long enough, you will certainly meet many employees who do this.

There are common playbooks of ways that you can get ahead quickly, but new “use AI” directives like this throw them into the trash. So this post asks: if you were trying to game the system described in the memo, what would your pathological behavior be?

Shopify recently announced that they expect their employees to use AI as part of their day-to-day work. They’re serious about it: they’re going to even include it in their peer feedback cycle, and it is now mandatory to use AI in the prototyping phase of projects.

I decided to have some fun. If you were a Shopify employee, and the announcement memo is the only information that you had, how would you game the promotion criteria?

https://www.clientserver.dev/p/helping-shopify-employees-game-their

@reconbot I've noticed a bunch of junior devs on subreddits getting discouraged by the rapid pace of AI, so sometimes it feels worth repeating. Especially for people who don't yet know how engineering happens inside of companies!

In this newsletter, I talk through these arguments on a deeper level: what would it look like to stop learning to code now? What activities are difficult to replace? How many times can I use a block quote? All of this and more in the latest issue of my newsletter.

There are still a whole host of activities that inform the code, and are informed by the code. So even if coding agents were able to start doing the majority of coding over the next few years, there is a tremendous value in being able to place the coding in the context of an organization.

Several CEOs have declared “you shouldn’t learn to code” in several different ways. These proclamations range in severity from “coding and all related activities will be done by AI systems” to “you should not learn to code now.”

It may be true that AI will take over coding and all related activities, but the stance “you should not learn to code now” is irresponsible. In many ways, coding is the easy part of being a software engineer

https://www.clientserver.dev/p/you-should-obviously-still-learn

War Story: The Hardest Bug I Ever Debugged, by @jakevoytko:

https://www.clientserver.dev/p/war-story-the-hardest-bug-i-ever

#debugging #javascript #google #v8 #chrome #browsers

I believe that we should do the same for LLMs: Allow Cursor / Copilot / ChatGPT etc in your interviews. Let people use them. Hold them to the same standard and make them prove that their solutions are correct.

In this issue of my newsletter, I give examples of cheating and give a more in-depth argument about why you should prefer allowing people to use generative coding tools in interviews.

When I was at Etsy, we had an interesting way of addressing people cheating on their interviews by Googling the answers or looking on Stack Overflow. We just told them that they could use external resources, and we designed our interviews so that the entire thing couldn't be answered just by searching for the answer. This allows people to interview like they are working, instead of interviewing them with one hand tied behind their back.