quoting Serafín, 農, Canada:

Suggested stress pattern: Word-initial, or near the beginning of the word (maybe there's derivational prefixes that never take stress?).

quoting con quesa, Layperson, California:

For negation, why not copy English on this? A negative auxiliary verb analogous to "don't" transparently formed from an older negative particle, used in some TAM classes, with that same negative particle (let's say its ga to be concrete) used for negation in other TAM contexts?

Soŝna ɡás wek'né
Soŝna-NOM NEG.DO catch.fish
"Soshna doesn't catch fish"

Soŝna ga móch'í wek'né
Soŝna-NOM NEG want.to catch.fish
"Soshna doesn't want to catch fish"

Hmm... what is -s? Some sort of habitual-aspect marker for only auxiliary verbs?

What I had in mind was that -s (+ vowel length, but we've gotten rid of the length) is the fossilized remnant of a light verb, in the same way that the /d/ in "don't" is the fossilized remnant of the light verb "do". ga is a particle, not a verb in and of itself.

Some suggested morphological distinctions:
- determiners: number (singular, plural), case
- - probably with some syncretism

- nouns:
- - number: singular, plural (with an affix), supraplural (meaning "many many Xs", and let's make it easy using full reduplication of the plural: "x.PL x.PL", or maybe "x.PL from x.PL")
- - - - something similar to the supraplural is attested for some collective nouns in Arabic, where collectives can take plural morphology
- - cases: NOM, GEN, INST (instrumental)
- - - - this is inspired from Georgian, which doesn't have an accusative
- - - - for direct objects we could use NOM in positive verbs or sentences, and GEN in negative ones, cf. de in French j'ai un problème 'I have problems' vs. j'ai pas de problème 'I don't a(ny) problem'.
- - - - a paradigm of 2 numbers x 3 cases seems good and fairly easy to learn to me, even with morphophonological complications

- verbs: no idea yet, but maybe distinguish habitual, present, past, irrealis at least?

I like those TAM distinctions. Let's also mark subject and object person on the verb, with minimal syncretism. I like the case ideas, but want to add at least one more - I suggest a DATive case for indirect objects. Georgian of course has some interesting mappings from syntactic roles to case assignments based on TAM category, which I wouldn't mind trying to build something analogous to.