In the early twenty-first century, computers do not have the direct ability to produce random numbers. The most common solution is through the utilization of a Pseudorandom Number Generators. These algorithms produce noisy, statistically well spaced distributions of numbers. But they are entirely deterministic: the exact same numbers will always show up in the exact same order.
To get around this limitation, Pseudorandom Number Generators make use of an input, known as a seed. The seed is used to skew the deterministic algorithm to produce different outputs.
As a simple example, imagine a list of a trillion numbers that have been set up to remove any obvious pattern. Whenever you need a new random number, you just read the next number in the list. If you ever got to the end of all trillion, you go back to the start. This will not produce random outputs, because it does repeat, and if you read from the start, you will always get the same numbers in the same order. But, rather than always reading from the start, you instead choose to read from the second number or the ten billionth number. Starting at these different numbers would make these lists look very different. So this starting number serves as the "seed" so the algorithm can produce different output.
Actual Pseudorandom Number Generators are more complicated. It is a problem if two similar seeds produce similar output. Starting at the second index is very close to starting at the third in the case above. But the principle of introducing variety into a deterministic algorithm is the same.
Choosing a seed for an algorithm is a simple problem to solve. Often, it is set to be based on the current time and date. If you measure when a program is run down to the millisecond, you will generate a seed that is hard to replicate. Sometimes programs use measurements from the real world, like atmospheric pressure, to generate effectively random seeds. And sometimes, especially when testing programs, the seed is chosen to just be a specific number.
Video games like Minecraft rely headily on pseudorandom number generators to add variety to the experience. The entire world is based on these algorithms making random decisions, and these algorithms, in turn, rely on seeds. As is common with many video games of this type, the seeds are available for view, can be copied, and can be manually inserted when starting the game. Two players with the exact same seed will see the exact same game world generated.
This has led to "seed hunting," an activity of trying to find a seed that will generate an ideal world. For example, many speed runners want to find a world that has all the conditions so they can beat the game as quickly as possible. Others might be looking for interesting notable features, such as the one described above.
Because Pseudorandom number generators are effective, and because similar seeds do not produce similar outputs (as part of the intentional design of the algorithms). It is generally very, very difficult to extrapolate what a game world looks like given a seed without actually running the game and checking. So seed hunting can be a very involved activity. But once, found, any player could use a seed to experience a very rare game world.
The concept of "seeds" would be understood by people who play video games that involve them prominently, by programmers, and by people who study computer science. Most people in the early twenty-first century would have no reason to know about this niche topic.
