>In my model, this root comes from a Neolithic IE paleo-variety (in Villar's terminology), reflected in Germanic *xanDu- 'hand' and Semitic *xam(i)S- '5'. Afterwards, it became fossilized in the IE numerals 10 ('2 hands'), 100 ('big hand') and 1000 ('fat handful'), as described in Mallory & Adams (2006).
> > > Suppose PIE had 'dwó do komt', 'trí do komt' vel sim. (cf. the Lat. -gint-, Gk. -kont- for decades), then by false division *dé-komt- "ten". Voilà!
> > >
> > Great, but did *komt- mean "bundle of fingers" or "bundle of hands" or something else? Why not just "hand", and then go along with Pokorny in making *dek^mt- a reduced form of *dwe/dwo k^mt (or *k^omt)? Maybe Gmc 'hand' was originally a consonant stem, and then became an u-stem because of the accusative endings -um and -uns, like Gothic <fo:tus>?
> That is an intriguing idea. However if *k^omt- or *k^ont- were the stem we could not get Gmc. *hanDu- because the accusative endings would not be accented for Verner's Law to operate, just as we get Goth. <fo:tus> not *fo:dus. If *hanDu- came from a C-stem we would need something like *kondH- or *k^ondH- for the unshifted stem.