An additive power of exponent is a word over an integer alphabet
that is the concatenation of
nonempty blocks having equal lengths and equal sums of their
letters (Pirillo and Varricchio 1994). Exponents 2 and 3 give additive squares and
additive cubes, respectively. For example, 1423 is an additive square, since 14 and
23 have the same length and sum, although it is not an Abelian
square. A finite or infinite word avoids additive
th powers if none of its finite consecutive
blocks is such a power.
Shallit (2026) constructed an infinite word avoiding additive squares on an alphabet of six positive integers. This refutes the conjecture that every bounded sequence of positive integers contains an additive square (Brown and Freedman 1987). The alphabet can be taken as
where ,
,
, and
. The minimum possible alphabet
size is therefore 4, 5, or 6. Shallit (2026) credits GPT-6.2 Astra with the accompanying
Lean formalization.
For additive cubes, Cassaigne et al. (2014) constructed an infinite avoiding word over . It is the fixed
point beginning with 0 of the word morphism
given by
,
,
, and
. Jamet and Rosenfeld
(2026) report that applying the word morphism
with
,
,
, and
gives an additive-cube-free word
over
. Their computer-assisted argument reduces avoidance
to finite graph reachability using numerical spectral
bounds. They also obtain avoidance over the three-letter alphabets
and
. Jamet and Rosenfeld (2026)
credit ChatGPT with suggesting the first coding word
morphism, which they then analyzed.