There's a common trade-off between things done by humans and things done by algorithms: humans often have better judgment, but they're slow and thus expensive at scale; while algorithms are the opposite. This trade-off applies very much in recommendation: is it better to let a machine help you decide what to read (or watch, or listen to, or buy...), or should you seek manually curated suggestions?
One thing I like about The Sample is that it combines both approaches. We use an algorithm to help people discover newsletters, and many of those newsletters in turn include manually curated recommendations. This helps us get the best of both worlds. The algorithm is scalable; it can consider all the newsletters in its database whenever it picks a recommendation for you. It can introduce you to new writers even if no one you already follow has heard of them yet. And then when you subscribe to a newsletter you like, the algorithm passes the baton over to a human author. That author will have a far deeper understanding and awareness of their niche than our algorithm ever could.
So the bulk of your recommendations come from humans who you know and trust, and the algorithm tries to help you find curators and authors worth subscribing to. Even if some of the algorithm's recommendations are duds, that's OK as long as it finds you enough good newsletters to keep your inbox fresh.
An interesting implication is that The Sample isn't really about newsletters per se. By teaming up with human curators, it can help you discover anything.