Concepts¶
These pages explain how Numerisect thinks. They are worth reading before the feature guides, because they set the vocabulary the rest of the documentation uses.
-
Proven, probable, inconclusive
The four labels attached to every result, why the distinction is enforced, and what each costs. Start here.
-
Inconclusive results
Why an exhausted search is never reported as a negative answer, and how to act on one.
-
The engines
What each native engine is, what it is good at, and when Numerisect reaches for it.
-
Architecture
How a request becomes an engine invocation, and why Python and JavaScript compute nothing.
-
Security and privacy
Loopback binding, the session token, what never leaves the machine, and the two outbound paths that exist.
The one-paragraph version¶
Numerisect is a user interface over existing number-theory libraries. A computation uses a library routine first; an optimized C program with GMP or FLINT only when no library provides one; and never Python or JavaScript. Results are labelled by strength, and a search that hits a bound is inconclusive rather than negative. Everything runs locally unless you explicitly enable one of two outbound features.