[dvduval]

Absolutely for the average developer the token speed is just going to be too slow for it to be workable. I think we’re looking at 2028 when memory becomes cheaper again and they’ll be a lot more people using local models.