Perhaps Jason's approach may come out of his intuition, I'd like to explain it from a more "academic" perspective. In economics, the labor market is often suffered from information asymmetries where the employer has little means to determine the productivity of prospective employees. Therefore, if the employer is willing to pay average wage, it will obtain below average workers, as workers who has higher than average productivities won't accept the offer. Such a phenomenon is also called "the Market of lemons" in the context of used car market.
The root cause of such a market failure is because workers' productivity is difficult to measure. A certain measure has to be introduced to indicate the worker productivity indirectly (often termed Signaling in economics). For decades HR/Recruiters have been addressing this problem by using different metrics as the signals/indicators. Popular signals include education (GPA, university prestige).
Programming may be a bit more challenging as multiple factors can affect programmers' productivity, e.g., intrinsic intelligence, problem solving skills, the speed of learning. Our hard-working recruiters/interviewers have introduced some new signals --- brain teasers, coding tests, etc.
What Jason proposed in this blog post is that we don't need all those signals, they are all inaccurate and can be fooled around by a well-prepared interviewee, why not directly measure their performance by working with them for a short period, say, three weeks.
That does sound like a good idea. IMHO, that's basically what internship does for students, but not sure if that will work for full-time employees, as it will incur extra opportunity cost for them.
The root cause of such a market failure is because workers' productivity is difficult to measure. A certain measure has to be introduced to indicate the worker productivity indirectly (often termed Signaling in economics). For decades HR/Recruiters have been addressing this problem by using different metrics as the signals/indicators. Popular signals include education (GPA, university prestige).
Programming may be a bit more challenging as multiple factors can affect programmers' productivity, e.g., intrinsic intelligence, problem solving skills, the speed of learning. Our hard-working recruiters/interviewers have introduced some new signals --- brain teasers, coding tests, etc.
What Jason proposed in this blog post is that we don't need all those signals, they are all inaccurate and can be fooled around by a well-prepared interviewee, why not directly measure their performance by working with them for a short period, say, three weeks.
That does sound like a good idea. IMHO, that's basically what internship does for students, but not sure if that will work for full-time employees, as it will incur extra opportunity cost for them.