AI and programming classes in Basingstoke

The learner downloads the Project Gutenberg text, keeps everything from "It is a truth universally acknowledged" to the end of the novel, and removes the illustration notes. That leaves 122,978 words by our count. The first program splits after every full stop, question mark or exclamation mark that is followed by a space. It reports 5,767 sentences, averaging 21.32 words. That looks reasonable, and a less curious programmer would stop there.

Then the learner prints the shortest "sentences" and finds fragments such as "My dear Mr." and "for Mrs.". Every Mr. and Mrs. has been treated as the end of a sentence: 781 plus 343, and 7 more for St., 1,131 false breaks. Protecting those abbreviations fixes that, yet the count drops to 4,636, far lower than before. Why? Because in dialogue a sentence often ends with a full stop followed by a closing quotation mark, and the rule only looked for a full stop followed by a space. Adding that case gives 5,776.

Here is the unsettling part. The naive count of 5,767 was within ten of the careful count of 5,776, not because it was right, but because 1,131 false splits almost exactly cancelled about 1,140 missed ones. A program can be wrong in two ways and still print a believable number. The learner writes tests for each case separately, an abbreviation, a quoted ending, a question mark inside quotes, so that each rule is checked on its own rather than trusting a total that happens to look fine.

Testing the parts instead of admiring the total is a strong reason for Basingstoke teenagers to keep learning to code in 2026. The longer argument is in why teenagers should still learn to code in 2026.