yeah, the break even is the power cost on mining. it tends to form a floor on the bottom of PoW prices, unless the protocol utterly fails like Verge where it can go to zero. the exact scheme of difficulty adjustment is every 2016 blocks (roughly 2 weeks) the timestamps of the previous 2016th block back from that point, is divided by the number of seconds of 2 weeks, and then the factor is applied to the current difficulty to raise or lower it, raise if it's too fast, lower if too slow. the difficulty value itself is encoded as 32 bits that are multiplied by a constant factor to become a value based on the same number expanded to 2^256. it's the absolute simplest possible thermostat-like adjustment that only adjusts when a clock ticks over, the block height, in this case. it works but it's prone to error and doesn't work nice if hashrate jumps up and down a lot, and more frequently than every 2 weeks. it's robust for typical expected rise in supply of hash power, but it is not robust for sharp drops. it can get stuck on too low or too high quite easily. too low corrects faster, too high corrects slower.