That's true. Is the cost of firing really that high, though? And how good are the tech tests at determining that a software developer will be good at working on a product? It seems to me that you could be great at printing all paths in a linked list of numbers that sum to to sum of a different path in a binary tree, but really not all that good at working on a product.

I know that many companies are so concerned about false positives (hire a bad developer) that they are willing to tolerate a high incidence of false negatives (fail to hire a good developer). That's an odd position for an industry claiming a severe shortage of talent, but there's also a strong case that if you allow too high a rate of false negatives, you can make it so hard to hire a good developer that you end up hiring a bad one. In short, it's more likely that one bad developer will slip through the cracks than a good developer will make it through, if the good developers are a small percentage of the hiring pool and the false negative rate is very high.