what does lulu size do?

Prices a design you already set, and says what precision it buys. You fix the prompts, the repeats and the engines; it returns the bill and the precision that design actually buys.

how do you run it?

In the parser’s own words, what a design you set costs, and what precision it buys. Every flag this build accepts, on one line:

lulu size --prompts PROMPTS --runs RUNS [--engines ENGINES] [--brands BRANDS] [--confidence CONFIDENCE] [--rate RATE] [--per-brand] [--pilot PILOT] [--brand BRAND] [--ledger LEDGER] [--searches-per-call SEARCHES_PER_CALL]

lulu size --prompts 24 --runs 5 --engines perplexity,anthropic  ± two engines, priced apart

why price a design before it runs?

lulu plan answers "what do I need". This answers "what does what I set cost", which is the question that matters when the key is the buyer's: the spend lands on their account, so a figure they are asked to approve has to be answerable before the round rather than after it.

what will it not round off?

At one run per prompt it prints no interval at all, because the model's rerun noise and the prompt-to-prompt spread are then the same quantity and no split can be identified. A round of that shape returns a reading, and a reading is not a measurement.

And a second engine does not narrow the first, since engine answers are never pooled into one proportion. What more engines buy is coverage across systems, each carrying its own interval and its own bill.

which flags does it take?

--prompts PROMPTS ± how many tracked questions, required

--runs RUNS ± how many times each prompt is asked, required

--engines ENGINES ± comma separated, provider or provider:model, e.g. perplexity,anthropic, default perplexity

--brands BRANDS ± how many brands are read from the same answers, default 1

--confidence CONFIDENCE ± default 0.95

--rate RATE ± expected appearance rate; 0.5 is the worst case and the default, default 0.5

--per-brand ± size one brand at a time instead of holding every brand at once

--pilot PILOT ± a recorded round to measure the split from

--brand BRAND ± which brand the pilot is scored for

--ledger LEDGER ± where the pilot round lives, default ./ledger

--searches-per-call SEARCHES_PER_CALL ± searches one call may run, for engines billed per search rather than per call, default 3

where does it lead?

lulu plan ± lulu collect

the engines it can ask: perplexity sonar, claude.

the parts of the method it reaches: intervals, variance.

every command: setup, init, doctor, collect, draft, usage, ablate, lift, screened, publish, rivals, report, verify, plan, size. back to the docs