You would be foolish to completely avoid AI, but you are also right to be careful. AI is best used for boring stuff that's non-critical, and for stopping you from having to reinvent the wheel.
For example for a game, you could use AI to finish a sprite sheet based on a couple frames you provide, or to give you a pathfinding algorithm.
AI is a resource, and a startup has a better chance of success the better it leverages the few resources it has.
Frankly I think serving dynamic HTML at all is a problem in an SPA. Use a service worker to cache structure, control, and styling, then use an API to fetch dynamic data in a more efficient manner.
Attention to details that matter and knowing what to pass on, is what's important. If they did poorly on the important stuff but then point out the typo, clearly they were putting their efforts in the wrong places
Sadly I've seen this kind of interviewing from lots of places; maybe it's a technique they teach in college or something.
Interview questions like this are useless as they are not an accurate approximation of a real work scenario. I've frozen up at a whiteboard in an interview before, and then figured it out on the drive home.
The best approach I've found is to do a send home quiz, and then ask them questions about the quiz to test understanding. The test also shows coding habits, which tells you a lot right there.
Bonus: put some typos in the quiz that don't matter. Red flag for the people that get hung up on that.
Any framework that requires me to learn custom syntax, is a problem in my opinion. I agree with the author that the browser is already the framework, and a powerful one at that. I don't really struggle with views and dom thanks to my own libraries, so maybe those without appreciate thw guilded gardens of these frameworks. Problem is that if these frameworks fall out of favour or stray from their original quality, you're stuck with worthless knowledge and maybe code as well
I've been running this web server for over a year with no stability issues like that. The app itself shutting off was the biggest issue, but the alarm plugin reboots it whenever that happens. Perhaps you had a bad batch of phones but in my case, it's been very reliable.
Author here. I'll admit that I have made limited use of AI for helping in areas that I'm less familiar with (like FTP protocols), but I haven't paid a dime; all free plan. I'll use AI to help make the project possible, but it's not trustworthy enough to make things readable and maintainable.
If you want help getting this setup for yourself, let me know; happy to help.
Author here. Qik is the name I've given to the project. It's pretty clear on all the pages that it will always be free, and that there's no reference anywhere to any crypto services. Give specifics if you think otherwise; there's lots of noise these days, so I'm not offended you figured this was another one.
Author here. All good points from this point and its replies, which to me says best to me that not all products and experiences are equal, so proceed based on your comfort level.
Note that the app is capable of monitoring both battery charge percentage and temperature. If one was inclined a USB to serial adapter with some simple circuitry could be used to monitor and control the state of charge. Good idea for a follow-up project and article!
Hold the submitter to a higher standard by introducing penalties for errors, and/or incentives for clean PRs. In my opinion a PR should be flawless once submitted, as the dev should have tested thoroughly. QA should only catch errors in a small percentage of cases
If you want to buikd and distribute something B2B or just to friends or colleagues outside of the app stores, it's a nightmare on iOS. Sideloading isn't as simple anymore for Android, but it's way easier. It's for that reason and for more open hardware and software that I exclusively work on Android now. Too bad about iOS as generally I like Apple hardware, just not interested in the hassle with everything else.
Personally if I see a PR that's not readable, I send it back immediately without much of a read-through. AI or not, the code should be readable or it's not maintainable by people or AI
Yes, very familiar. I think it might be ok though, as it's similar to a web search where you maybe need to refine it a few times.
Just assume that the first answer may not be fully correct or complete, until the AI has more context.
It would be a nice change if it would ask the follow-up questions automatically.
You may want to offer a page credits option instead, as "50 pdfs" for the free tier is a little vague. Is that 50 1-page pdfs, or 50 100 page pdfs? We generate our pdfs one page at a time, and then stitch them together into one at the end; I'm unsure whether your solution would be practical for us. Self-hosting is maybe the other option
Totally agree that creating something to reproduce the issue is a big turn off when it comes to reporting bugs; what else can one do though? In your case though, I guess it was a pretty simple thing
Well-written and valuable for insight whether you have similar personal experience or not.
As someone who does hardware and software as well, I relate to the challenges of making something you can hold; it's very easy to underestimate the challenge difference between the two.
Your Murphy's law references are spot on; I feel comforted reading I'm not the only one this happens to! Misery does love company, and it's important to hang on that I think, so that you don't lose hope :)
I do but I feel it's long overdue for replacement. I've been working on a new protocol that includes timestamp filtering, real-time notifications, and optional things like 'likes' and 'comments'. I use it everyday and I load it with content from RSS feeds. Not ready for sharesies yet though
A self-hosting ecosystem with apps and services to do everything from file sharing, to email, and dynamic DNS. It's Javascript based and runs in node, but I've also got it running under Android and in a regular web browser (the web browser one needs requests forwarded to it). The whole goal is to make self-hosting really easy, cheap, but also capable.
It's got a long way to go, but I've already posted a few articles on HN that are hosted on it.
I'm not sure I can accept the findings of this article. It seems to me that all the concerns and issues can be summed up to bad design. A mini framework should be designed to be easy to adapt and extend, and any layer on top should be just as easy to further adapt and extend.
That to me is the crux of this, "avoid frameworks that are difficult to adapt and extend".