Which of Zamenhof's mistakes was stupidest? A lot of his decisions were clearly the result of forgivable ignorance – he was after all working before linguistics as a science really existed. The inconsistent possessives; the phonemic inventory; the word patrino… there are plenty of mistakes to choose from. But when it comes to picking the stupidest, it's not a very hard decision. Two of the leading candidates as I see it are as follows:
Rigardu la vortfinaĵoN = “Look at the word ending”
Now, the various “design philosophies” for invented languages each have their advantages and disadvantages. The problem with trying to design a “simple” grammar is that when for instance you shrink a language's system of case‐marking endings, it becomes harder to tell who's doing things to who(m). Usually context makes it obvious, but otherwise some other part of the grammar has to do the work of distinguishing between agent and patient – some of the complexity has moved from the noun morphology to somewhere else, like a lump in the carpet.
Those defending the Esperanto case system take it as axiomatic that case‐endings are an effective way to free up word order, but as it turns out, statistical surveys of natural languages show a correlation going the other way! Noun‐case systems are a “dependent‐marking” trait, associated with relatively fixed subject/object ordering – the grammatical systems with the least restrictive order rules are strongly “head‐marking”. That is, they put agreement markers on the verb to show that (for instance) the subject is first person singular and the object is masculine plural. This means it doesn't matter what order the verb's arguments go in, and they can often be left out; but Zamenhof eliminated all traces of this mechanism from Esperanto.
And the lumpy‐carpet effect certainly needn't stop us improving the carpet's overall evenness. In the case of case, it isn't necessary for nouns' syntactic roles to be shown by their endings – a constructed language is free to follow the example of the languages that have no affixing at all. It would be perfectly workable to mark case with a system of prepositions instead – or not to; again, it's an optional extra. But word order isn't an optional extra, it's a universal; all sentences necessarily have a word order, and all natural languages make some use of reshuffles to distinguish possible meanings. Esperanto might as well own up to having a rule that by default the order is Subject–Verb–Object. And once that's established, who needs a compulsory ‐N ending?
Compulsory agreement on adjectives
La vortfinaĵoJ estas komplikaJ = “The word endings are complicated”
English‐speakers are of course always accused of native‐language bias when they complain about this feature, but it's an objective fact that Esperanto takes its ubiquitous adjective concord to an extreme uncommon in natural languages and almost unheard‐of in designed languages. It's all very well to allow speakers to use adjective‐agreement if that's how they're accustomed to keeping track of which adjective goes with which noun; but forcing the rule on everybody, even in the vast majority of contexts where there's no ambiguity to be resolved, puts an extra barrier in the path of billions of potential Esperantists. Zamenhof himself recognised this too late, describing adjective concord in 1894 as “superfluous ballast”.
However, since these two stupid mistakes interact, we have a single clear front runner:
Compulsory case‐marking agreement on adjectives
Rigardu la komplikajN vortfinaĵojn! = “Look at the complicated word endings!”
As usual I welcome feedback: any dissenters with alternative candidates for Zamenhof's Single Stupidest Mistake should check out my mailbox.