>"Many people don't have this. And why would they: it's hard to evaluate a candidate via bits of random code and half finished projects which took their fancy, that they worked on on their own."
I do agree with everything here except "why would they". I have tons of old, unfinished, half baked code /designs lying around. I wouldn't show any of it to a future employer. What I would show them is something that I took real time and effort to do. Often time that's something for my self, an open source contribution or refactoring a side project to make it "usable". This takes a lot of time. But if you ever want to be in the position of choosing who you work for this is needed. That's all that I'm suggesting. The alternative is 6+ hours of spec work for every job you want to get. Which is mostly not reusable because it's company specific.
>"That's right. It's fucking impossible to evaluate someone based on an interview (there have been hundreds of posts to HN on this topic in the last few years). Existing work is hit-and-miss."
I mostly agree. I think multiple interviews with multiple people and talking through problems and design/development problems can greatly reduce the gap of it being hit-and-miss. I'm also not suggesting this is cheap to do (for both employee and employer). But hiring someone or starting a new job should require a lot of investment on both sides.
> " ...with the caveat that _we pay for it_. If you're serious about interviewing this person, paying their day rate for a day of their time is a reasonable way to ask people to do this."
I wish more folks did this. You obviously take hiring seriously. I bet you have many sold folks at your place.