I‑13

one letter · twelve forms · thirteen opcodes LIT · RUNS

I · DESIGNATE

I x NAME 184342 "s" CONSTANT 1843x.f ATTRIBUTE 1966

II · BIND — the write port

x <- e ASSIGN 1843def f(I a) ARG 1879-> e RETURN 1949

III · DECIDE

e ; EXPR 1957if c { } IF 1843a < b COMPARE 1843

IV · TRANSFORM

f(a) CALL 1936def f() { } FUNCTIONDEF 1936a + b BINOP 1843

SOURCE

IVM‑13 BYTECODE

press compile

RESULT

press run

OPCODE HISTOGRAM — the stage-13 corpus

What this is. A language whose entire form set was derived by counting: 649,634 AST nodes across the Python standard library, twelve constructs carrying 83.27% of them, six contributed by Lovelace in 1843 and six by five men between 1879 and 1966. I is the thirteenth symbol and the only one that is not a verb.

There is no loop. For and While are Dijkstra's, and his contribution is a removal — 0.70% by node count, invisible to a counter. So iteration here is recursion, and Note G's return to Op. 4 compiles to a call. The default program computes Bernoulli numbers, which is what Note G was written to compute, using the language reverse-engineered from what it contained.

The machine has thirteen opcodes, one per form. ASK and ANSWER are the two heads on I: in the Python stdlib they run 5.09 to 1 in favour of asking, and BIND holds 88% of every write in the language. Run the default program and the histogram comes out near 3.3 to 1 — recursion answers more often than iteration, so the ratio is a property of style, not of the language.

Stage 13, not yet built. That histogram is the point. A thirteen-opcode instruction stream should be the most homogeneous, highest-recurrence corpus obtainable — by construction rather than by luck — and every corpus screened this session failed on exactly those two properties. Whether a 13-form corpus is measurably more learnable than a 100-form one is the open question this was built to answer.