To be frank, my first reaction on reading this idea was negative.
I think part of my reaction is because I'm not seeing a useful goal: "interesting things" is certainly a goal, but strikes me as being too vague from which to extract useful information. What are interesting things? If you don't know what they are, how will you know if they happened? Are you set up to measure interesting things? If so, can you quantify them, and what sort of sensitivity/specificity, dynamic range, etc. do you expect your instrumentation to provide? What is the duration/endpoints of the experiment? Does the experimental design support the objective, and with what power does the experiment have to inform?
I think that the other part of my reaction is that, as an experimental scientist, I am specifically trained against doing something so unstructured. Practical considerations like safety and budgetary waste aside, I am the person I trust least: it's very easy to fool oneself about the significance of a result with an ongoing experiment that one is conducting. Experimental discipline, like planning ahead (and preregistration) and blinding and good controls, serve to promote objectivity and help reduce bias.
I have no doubt that this experiment would produce "interesting things". I am very skeptical that, as presented, you would know why it produced interesting things.
Of course, I'm just a dog on the internet. Best of luck!
Oh agreed, it's definitely not an experiment by any research standard. In that sense we're using the word metaphorically. But there's a place for ad hoc trial-and-error, and it's common to call that 'experimental' in colloquial English. YC has done such things in the past, e.g. for a batch or two it was possible to apply with just an idea. Nothing interesting came of that—other than that nothing interesting came of it—so it got phased out. This is an 'experiment' in the way that was.
It's popular on HN to suggest making YC (or startup investment generally) more science-like via controlled experiments, but that's hard to do at all and extremely hard to do well, and we're not set up for it. Maybe one day. But this is not that—it's just trying something new to see what happens.
As for whether interestingness is a goal in itself, on HN it is for sure. That value is hard-coded in the DNA of this place. And there's a well-established if not easily reproducible path from "huh, interesting" to "undeniably useful". That's the nature of the creative process and it's a space HN (and YC, in a different way) particularly like to inhabit.
I think part of my reaction is because I'm not seeing a useful goal: "interesting things" is certainly a goal, but strikes me as being too vague from which to extract useful information. What are interesting things? If you don't know what they are, how will you know if they happened? Are you set up to measure interesting things? If so, can you quantify them, and what sort of sensitivity/specificity, dynamic range, etc. do you expect your instrumentation to provide? What is the duration/endpoints of the experiment? Does the experimental design support the objective, and with what power does the experiment have to inform?
I think that the other part of my reaction is that, as an experimental scientist, I am specifically trained against doing something so unstructured. Practical considerations like safety and budgetary waste aside, I am the person I trust least: it's very easy to fool oneself about the significance of a result with an ongoing experiment that one is conducting. Experimental discipline, like planning ahead (and preregistration) and blinding and good controls, serve to promote objectivity and help reduce bias.
I have no doubt that this experiment would produce "interesting things". I am very skeptical that, as presented, you would know why it produced interesting things.
Of course, I'm just a dog on the internet. Best of luck!