Howdy, former Matasano pentester here.
FWIW, I would probably have done something similar to them before I'd worked in the security industry. It's an easy mistake to make, because it's one you make by default: intellectual curiosity doesn't absolve you from legal judgement, and people on the internet tend to flip out if you do something illegal and say anything but "You're right, I was mistaken. I've learned my lesson."
To the author: The reason you pattern-matched into the blackhat category instead of whitehat/grayhat (grayhat?) category is that in the security industry, whenever we discover a vuln, we PoC it and then write it up in the report and tell them immediately. The report typically includes background info, reproduction steps, and recommended actions. The whole thing is typically clinical and detached.
Most notably, the PoC is usually as simple as possible. alert(1) suffices to demonstrate XSS, for example, rather than implementing a fully-working cookie swipe. The latter is more fun, but the former is more impactful.
One interesting idea would've been to create a fake competitor -- e.g. "VirtualBagel: Just download your bagels and enjoy." Once it's ranking on Google, run this same experiment and see if you could rank higher.
That experiment would demonstrate two things: (1) the history vulnerability exists, and (2) it's possible for someone to clone a competitor and outrank them with this vulnerability, thereby raising it from sev:low to sev:hi.
So to be clear, the crux of the issue was running the exploit on a live site without their blessing.
But again, don't worry too much. I would have made similar errors without formal training. It's easy for everyone to say "Oh well it's obvious," but when you feel like you have good intent, it's not obvious at all.
I remind everyone that RTM once ran afoul of the law due to similar intellectual curiosity. (In fairness, his experiment exploded half the internet, but still.)