So I do believe that web dev has changed dramatically in the last year or so and that one probably needs to expand their skills and a front end web developer is probably less of a safe long term career than I believed in the past due to AI. This is a shame as I loved the craft and challenges of front end development.

when I see online everyday how AI is going to remove the need for devs I do tend to end up believing this is the case. But then in my day to day job I see that this just isnt the case.

there have been a few recent examples. So people believe you can one shot things. Like the recent adobe reverse engineering. It looks great at first sight as the UI looks like adobe. All of a sudden you believe adobe will be out of business but when you try it out it full of bugs and doesn’t Really work, essentially smoke and mirrors. Also you look at the code and realise it’s completely unmaintainble due to bad decisions made up front by people who aren’t really experts in web development.

there is something I was thinking about also, when non technical people say “oh but the app already works, what do you actually do at work all day”. It kind of sounds true but in reality it is so far rrom the case - there are always bugs to fix, features to add, packages to update. sure, ai can help with this, but in a big app for example a package update can break something, Claude wont neccessarily catch this, but it will be us Mere mortals who will get a call and need to fix this.

also I have seen vibe coded projects where non experts in managerial roles who are creating software, which they are so proud of despite having no idea how it actually works. Then things start going wrong and things are breaking, it’s not the manager who is fixing it but the devs have to. and since the code has no owner and it’s a complete mess everyone is just one shotting the fixes and adding to the pile of 💩

this post is a little scattered but I’m hoping it makes some sense and people can see what I’m trying to get at

Just got handed a vendor integration where the entire “documentation” is a 60-page PDF. Half the endpoints are in screenshots, and one of the auth examples was wrong, which I found out the hard way.

My process is the same every time: read it, retype everything into Postman, write my own types, and hope for the best. It takes forever, and I’m never fully sure I got it right.

What bugs me even more is that vendors update these docs every few months and don’t tell anyone. You only find out when something breaks in production.

So how do you all handle this? Do you paste it into an LLM and hope for the best, build the spec by hand, or ask the vendor for something better (and do they ever actually reply)? And how do you even know when they’ve changed something?

I’m also curious if this is just me, or if everyone has a graveyard of half-finished Postman collections.