| It's one reason I am actually going to continue pursuing the 'physical engineering' side of my career rather than the software development (software 'engineering' side of it). I studied CS at UCL years ago but worked mostly in teams with electrical, chemical, and mechanical engineers and on projects relating to various things in relation to various different industries as different as pharmaceutical and ship building. During that time I got qualified as a CAD engineer among other things I was required to by my employer(s). Now, when I look at software jobs and the way software developers are being interviewed...I just can't. You would never ask eg a mechanical engineer in an interview 'so, you say you know SolidWorks? Ok, design me a sphere with a chamfered cylinder extruded from the centre of the top. It should be chamfered at a 46 degree angle to a plane passing through the centre of the sphere. Also, sweep a five pointed star around the sphere at the point that is midway above the plane and the top of the sphere'. This is effectively what Leetcode tests etc are asking. It is a waste of everyone's time because they will either do it and think 'ok, that was weird' or they will refuse and think the company is insane. There should just be accepted certifications for software development AND THEY SHOULD BE HARD. They should be very much failable and you should expect good people still to fail once or twice (like engineering or medical exams). Then just ignore the Leetcode stuff. It serves no-one. Then all it really comes down to is frameworks etc and if someone is familiar with that particular framework or not. Just let people qualify out of this stupid algorithmic vetting process because it's utterly daft in the context of most jobs. |
The problem I see here:
> This is effectively what Leetcode tests etc are asking. It is a waste of everyone's time because they will either do it and think 'ok, that was weird' or they will refuse and think the company is insane.
is that candidates, even excellent ones, who are in any kind of position less than fully employed and/or with f-u-money, are going to comply with this sort of mistreatment no matter how odious it seems.
I am reminded of Steve Albini's famous article on how vile the music industry's recruitment methods are.[1] (warning: ugly truth, vulgar language)
No, obviously, we're not that bad off. Yet? From what I'm told it's already effectively impossible to get a job as a junior programmer now, thanks to LLM coding assistants. Others with better pedigrees than mine have said many times here that the tech industry is eating its own seed corn with such short term thinking.
But this hiring paradigm described in TFA is making things far worse. I feel like very few people realize that tech is poised to drive off a cliff, and the few people who point this out are looked at like killjoys at best or Michael Crawford at worst.
> There should just be accepted certifications for software development AND THEY SHOULD BE HARD.
Agreed. What should these certifications test for? CompSci 101 concepts like variables, loops, good structure, maybe OOP? That's going to be useless for @koliber's Sr. Java Engineers screening problem. Heck, server jockeys like me with a smattering of shell and Perl and such under our belts could pass that.
Do we make it a dozen tests, each the equivalent of a final exam for a full semester of MIT senior-level courses? Well, now we've weeded out every possible junior starting their career; we might as well just give in to the AI assistant apocalypse right now and avoid the rush.
Do we have a certification specifically for folks who work with 3d graphics primitives and also need to be able to do matrix math in their heads?
> Then all it really comes down to is frameworks etc and if someone is familiar with that particular framework or not. Just let people qualify out of this stupid algorithmic vetting process because it's utterly daft in the context of most jobs.
I'd love to see somebody dump the HN database for the past 10 years or so and do a semantic search to count how many times this sentiment has been articulated here. If you told me it was over 10k I wouldn't be at all surprised and even if you said it was 50k I wouldn't be too skeptical.
And yet, who is actually moving the needle on correcting these systemic flaws that we all seem to recognize? Lots of folks talking about it and maybe a few companies trying, but not much progress or we wouldn't still be complaining about the same things year after year.
I have only a half-baked image of a solution in my mind, and it requires all parties to make painful sacrifices that they certainly won't do without force (ie. legislation and penalties). In another recent thread somebody mentioned that programmers are all special snowflakes who couldn't possibly be subjected to a guild or union or such. Well, we may be in the final days of the Old Times of computer programming. What happens when the tech industry has its own Quebec Bridge moment? [2] What comes next may be a mandatory Canada-style engineering culture and laws[3][4] for everybody who writes code, with prison sentences like for Practicing Medicine Without a License.
How many of us are willing to commit to that?
1. https://web.archive.org/20260228050755/https://thebaffler.co...
2. Some folks might say we've already had several, like Therac-25, the Snowden disclosures, the Cambridge Analytica fiasco, the vibe coder you know who dropped the Prod database because he gave Claude direct access to it, ....
3. https://web.archive.org/web/20260215000702/https://ironring....
4. https://hackaday.com/2023/12/18/when-is-an-engineer-not-an-e...