That's a curious definition of number which Aristotle refers to as being held 'by some' when, in Metaphysics Z.13, he wishes to criticize the view that a substance could be compounded out of other actual substances.
If the substances that supposedly constitute a substance were fully actual, then, Aristotle remarks, what Democritus had claimed would be correct: you couldn't get a single thing out of two, or two things coming from one. (Two would remain two; and the putative 'one' thing would have had to be two from the start.) But the same would hold of any number (not simply two), because a number of things would, on this view, be no more than a 'grouping of units', as indeed some have alleged:
A substance cannot consist of substances present in it in complete reality; for things that are thus in complete reality two are never in complete reality one, though if they are potentially two, they can be one (e.g. the double line consists of two halves-potentially; for the complete realization of the halves divides them from one another); therefore if the substance is one, it will not consist of substances present in it and present in this way, which Democritus describes rightly; he says one thing cannot be made out of two nor two out of one; for he identifies substances with his indivisible magnitudes. It is clear therefore that the same will hold good of number, if number is a synthesis of units, as is said by some; for two is either not one, or there is no unit present in it in complete reality.
a)du/naton ga\r ou)si/an e)c ou)siw=n ei)=nai e)nuparxousw=n w(j e)ntelexei/a|: ta\ ga\r du/o ou(/twj e)ntelexei/a| ou)de/pote e(\n e)ntelexei/a|, a)ll' e)a\n duna/mei du/o h)=|, e)/stai e(/n (oi(=on h( diplasi/a e)k du/o h(mi/sewn duna/mei ge: h( ga\r e)ntele/xeia xwri/zei), w(/st' ei) h( ou)si/a e(/n, ou)k e)/stai e)c ou)siw=n e)nuparxousw=n kai\ kata\ tou=ton to\n tro/pon, o(\n le/gei Dhmo/kritoj o)rqw=j: a)du/naton ga\r ei)=nai/ fhsin e)k du/o e(\n h)\ e)c e(no\j du/o gene/sqai: ta\ ga\r mege/qh ta\ a)/toma ta\j ou)si/aj poiei=. o(moi/wj toi/nun dh=lon o(/ti kai\ e)p' a)riqmou= e(/cei, ei)/per e)sti\n o( a)riqmo\j su/nqesij mona/dwn, w(/sper le/getai u(po/ tinwn: h)\ ga\r ou)x e(\n h( dua\j h)\ ou)k e)/sti mona\j e)n au)th=| e)ntelexei/a|.
What is the source of this curious definition of number? Ross notes the attribution of a very similar formula to Thales:
This is practically the same as the earliest recorded Greek definition of number, mona/dwn su/sthma, which Thales is said to have borrowed from the Egyptians (Iambl. in Nicom. Ar. Introd. p. 10.8). Cf. D.1o20a13n.I gather that this attribution is not taken to have much weight. At least, "a number is a grouping of units" isn't included in the standard lists of the three things Thales is credited with having said.
But I wonder whether the Phaedo provides evidence that the definition originates at least with presocratic natural philosophy. What I have in mind is the similarity between Aristotle's discussion, and the passage, in the autobiographical part of the Phaedo, where Socrates says that he no longer accepts a view he had previously accepted, about the nature of the numbers one and two. It is true that Aristotle takes the definition to be inadequate on the grounds that constituents need to be, as it were, 'matter', if they are to be united into a true unity. But it would make sense for Plato to object to the definition on the rather different grounds that it treats a number as if constituted by a mechanical operation, viz. the placing of units in proximity. (Cornford remarked that the definition, mona/dwn su/sthma, is "crude and, so to say, materialistic".) Note that what Socrates objects to is a number regarded as generated by a pro/sqesij.
“By Zeus,” said he, “I am far from thinking that I know the cause of any of these things, I who do not even dare to say, when one is added to one, whether the one to which the addition was made has become two, or the one which was added, or the one which was added and
[97a]the one to which it was added became two by the addition of each to the other. I think it is wonderful that when each of them was separate from the other, each was one and they were not then two, and when they were brought near each other this juxtaposition was the cause of their becoming two. And I cannot yet believe that if one is divided, the division causes it to become two; for this is the opposite of [97b]the cause which produced two in the former case; for then two arose because one was brought near and added to another one, and now because one is removed and separated from other. And I no longer believe that I know by this method even how one is generated or, in a word, how anything is generated or is destroyed or exists, and I no longer admit this method, but have another confused way of my own."