what does lulu rivals do?
Prints every name a recorded round reached for, and where it reached for it. The questions in a screening round carry no company in them, so the names in the answers are the model's own.
how do you run it?
In the parser’s own words, every name a recorded round reached for, and in which questions. Every flag this build accepts, on one line:
lulu rivals --snapshot SNAPSHOT [--ledger LEDGER] [--least N]
lulu rivals --snapshot ROUND ± every name, in every questionlulu rivals --snapshot ROUND --least 2 ± names that turned up in two questions or more
why per question rather than one total?
Each name is reported with the questions it turned up in and how many draws of each named it. A name in every draw of one question and a name in one draw of every question are opposite findings, and a single total would print them the same.
--least is a floor on how many questions a name has to turn up in, and it defaults to letting everything through. A name in one question is a real observation about that question, and this command does not get to decide which observations a reader is allowed to see.
which names get printed?
Which of those names is a competitor is the one judgement it does not make. A name is printed when the round wrote it inside a sentence at least once, or when it is spelled the way no ordinary word is, and never when the same round also wrote it in lower case. Every name printed appears in a recorded answer character for character, which is the rule a quote is held to everywhere else here.
what does it still get wrong?
It prints type words above real names. API, AI and ML appear where a company should, because the rule is positional plus lowercase evidence inside the corpus, and those three pass both gates. A person reading the table strikes them out in a second; the table does not.
And it drops a name nothing spells. A name that never appears inside a sentence, and leaves no trace in the surrounding spelling, is not counted.
which flags does it take?
where does it lead?
lulu screened ± lulu publish ± lulu draft
the parts of the method it reaches: names.
every command: setup, init, doctor, collect, draft, usage, ablate, lift, screened, publish, rivals, report, verify, plan, size. back to the docs