Guides
Three things worth knowing before you scale a prompt
No growth hacks and no model rankings: just the counting, the pricing and the break-even arithmetic that decide what a language-model feature costs.
How to estimate tokens before they hit your invoice
Why a token is not a character, where char/4 fails, the parts of a request people forget to count, and three ways to count before you spend.
LLM API prices in November 2026: input, output, cached and batch
The four prices behind every model row, cache economics with real numbers, when 50% batch pricing is the right trade, and the line items the price sheet leaves out.
Subscription or API? The break-even math
The formula that compares a $20 seat with metered tokens, a worked team example, margin math for products built on tokens, and three assumptions worth distrusting.