You're basically describing HackerRank, and I've worked with some pretty terrible engineers who were very proud of their HackerRank credentials.

Write such a test. Give it to all the engineers whose work you know well, and see if it can accurately predict what they brought to your team. I would be very surprised if you could both spot good talent and dead weights reliably.

We had some fairly standard interview questions that I wrote - problem solving type questions asking someone to sketch out a pseudocode answer. One of the smarter things I did was 'interview' several of our existing employees to get an idea of where people would fall on the scale. It was a great help when we then started giving them to candidates, and prevented us hiring some people because no, the good people we had really could solve these problems with only a bit of nudging.

This is more or less what we do on my current team. We write out a doc explaining all the common solutions, and what a good candidate should consider and be able to answer, and have a rubric, and we ensure people don't judge candidates until they've shadowed enough that we should have a pretty consistent scale and have made a reasonable effort at being objective.

The downside is some people have barely scraped by in the interview and turned out to be absolute rockstars and day 1, and some people have aced the process and driven everyone else mad with incompetence. So it could be improved.

And maybe that's just on us not writing the perfect questions yet, but a bigger problem is that once we've used a question enough someone will post it online and suddenly a suspicious number of candidates are able to show up and rattle off all 3 optimal solutions in 15 minutes. Weird! I suspect a standardized test approach to software engineers would get this problem even worse.

It doesn't have to be perfect, it just needs to be better than the tech part of the current interview system. I very much doubt that whiteboard interviews can accurately predict what people bring to a team; they're not even designed to assess that.

This system wouldn't replace the personality interviews, etc., but only measure technical abilities.

Is there much actual whiteboarding going on in tech interviews right now? It's been years since I did anything but remote coding interviews (where you have a fairly realistic editor and REPL environment) and remote behavioral interviews. I liked doing whiteboards where someone would walk us through a former project and architectural decisions, etc. with diagrams. I kinda miss that, actually. But I haven't seen literal whiteboard coding since long before COVID even.

But what would that test contain?

If I go by mainstream, then I guess it would contain some coding task, where you might have to solve algorithm stuff, that you will probably never encounter on the job, or some fake OOP "everything is a class" or "I must use [hot web framework of the month] for trivial things" kind of thing. None of that really tells much about a devs true skill.

I think the problem might be, that jobs are so divers as developer. For some jobs maybe the hot web framework of the month is really all you need to know, while other jobs might require you to think about how a system needs to be redesigned to run on some kind of multi server environment.

I guess there could be multiple such tests, for each area of expertise one. It would for sure be a lot of work to put them together. One would also need rotating problem sets or questions.

Also: Who is going to create these tests? I hope the idea is not big tech or some kind of government participation. Big tech would quickly make it a filter for only themselves, making their tech stacks assumed knowledge in the tests, and government would probably come up with something hopelessly backward like testing your Cobol skills or demanding you write an HTML 4 website, while you have gotten used to more semantic HTML 5 years ago.

I think it is a nice idea, but there are many questions about how it could be usefully implemented.

This could never work without on-site invigilated exams. Otherwise you will never be able to prove that when Jack applies for a job, Jack took the test. You'll just get an industry of Johns taking the tests for people.

This is how language certification works (think TOEFL, IELTS), and that industry still suffers from a massive amount of fraud. If you make a certificate like this too expensive, it serves as a barrier for entry for candidates. If you make it too cheap, the people administering the test will start pushing through candidates as quickly as possible without any regard for academic honesty.

I assume any certification exams where the results can be the difference between getting a good job and not face a huge amount of people who would cheat on them if they could. I know the ones I'm somewhat familiar with have a lot of safeguards in place to the degree that I gather some people found them pretty onerous with remote test taking in the pandemic.

You would need cartel-like control over the industry to make that work--as exists in the medical industry or legal industry. Without that hiring people just under a generally accepted cut off like 90 would be a form of gaming the system (hiring someone good and cheap but unavailable to bigtechco) that would kill the system.