I think I should elaborate a bit on the "concise" part. I don't mean that the design of the core is smaller than that of the Scheme. Although Janet was designed to be an embedded extension language as one of its use cases, I don't think its core is much smaller than Scheme, I would say it is comparable and roughly on a similar level.
But the writing is more concise in my subjective opinion. For example, you don't need extra parens for each `cond` condition, it just counts how many there are and every odd sexp is a condition, every even sexp is a corresponding action. Hash tables, arrays - very short notation for access and write. And records/objects - in Scheme it's like a tea ceremony with million rules and extensions, in Janet it's just a hash table. Many things are borrowed from Common Lisp as well (the good ones).