It probably works but the cost (and scalability) can restrict how fast/far it can be applied...

If a degree or formal education has been used as a barrier to restrict access for any time only the strongest and most driven of the degree-less are going to persevere and make it.

You can and should consider toy "beachheads" into larger and growing markets, but the idea of targeting a small/limited market is flawed IMO.

The examples presented are successful in what where HUGE markets with limiting factors. No one debated the size of the retail book business when Amazon started, the challenge was how to pay for things over the internet. Ebay provided market-making for individuals to sell stuff - anyone remember classified ads and the buy & sell? That was big business for newspapers.

I feel like there's a lot of selection bias and revisionist history in this post.

What? The entire point of vetting is to increase the quality of the remaining pool. You can argue how successful they are at doing this, but the intent of the process is pretty clear.

>> The way I see it, the vetting is more or less random.

Don't let YC or anybody else define you, but if you'd got in initially I bet you'd take it as a positive signal data point.

You also state that "not everything has to be binary" and then say something can be "worst", which is the same absolute you seem to be arguing against

I would however agree that:

* those who produce & then cull generate a lot more than we see in the end product (due to those who do little editing, probably more than 10:1)

* writers/programmers get better at their craft, reducing the chaff to wheat ratio as they practice (like the original author saw in his work)

* well understood domains or themes are also more efficient (Michael Crichton didn't throw away 90% of his output; your average CRUD app doesn't either)

Finally, the written word tends to remain unchanged once committed (revisions tend to be minor in the overall scope), while code that is used changes dramatically. Long-running projects that are successful will see their application increase and evolve accordingly, so it's likely less valuable to track change over time if you're looking for total vs. finished effort metrics.

* Whiteboard coding is awesome for software development shops that don't actually own computers or use punch cards to load programs.

* Shared coding environments with a time limit works well when you want to double screen for someone who can also diffuse suspicious packages that arrive at the office

* Riddles and puzzles are good when your workplace has a chicken coop, an office fox and you can't figure out how everyone can go for lunch without leaving the fox alone.

* Behavioral interviews are appreciated by candidates who took Psychology 100 back in school for an easy A.

* Take-home tests go over well with the huge population of talented software developers who can't find a job and have loads of time they want to spend decoding the operational cost of a bubble sort

The only thing I've seen that's at all realistic and effective is a very short, __paid__ project. I did a two-day one for Indeed (ironically one of the worst promoters of all this bullshit) and a 4-hour one for my current employer. The payment doesn't even have to be market, it's more important as a signal to candidates that the company values your time.

I ask for 3 things from perspective employers:

1. value my time like you value your own (both the number and composition of your interview steps)

2. keep me updated as to where we're at in the process and when the next stage/decision will be made

3. Move forward in the process in a timely manner. It should not take more than 2 weeks from when you initially contact me to the process concludes.

I've never gotten more than two of the above from a single organization, but I will someday, and I'm betting that a company that treats potential employees that well will treat actual employees better as well.

Actual Code? Not a chance. Portability, Durability, Efficiency - It's the worst of all worlds.