Online coding and Python classes in Margate

The learner takes the Project Gutenberg text of The Waste Land, keeps only the poem itself, and turns it into a stream of 3,071 lower-case words, 1,168 of them different. Words that repeat are the chances for a cache to help: at most 62.0% of the stream could ever be a hit, since each word's first appearance has to be a miss. Four rules decide what to forget when the cache is full. First in, first out (FIFO) drops the oldest word. Least recently used (LRU) drops the word unused for longest. A frequency rule drops the word seen least often so far. And the optimal rule, known as Belady's rule, drops the word that will next be needed furthest in the future, which is only possible here because the whole poem is known in advance.

LRU beats FIFO at every size here, a common reason it is taught first: remembering what was used recently is often a good guess about what will be used next. On this poem, though, the frequency rule beats LRU too, because a few small words dominate. "The" appears 205 times and "and" 107, so a rule that clings to common words keeps winning. The optimal column is not a practical rule, since no real program can see the future, but it shows how much room is left: at 32 words, the ceiling for any rule is 48.1%, against 29.0% for LRU.

The lesson is that no eviction rule wins everywhere. A different stream, a web server's requests or an AI agent's tool calls, could reverse the order of FIFO, LRU and frequency. Good engineers measure on their own workload and compare with the optimal bound before choosing.

A Margate teenager who understands caching can reason about speed and cost in any AI tool, and that is a strong reason to learn Python in 2026. The longer argument is in why teenagers should still learn to code in 2026.