Been running local ai inference for 2 weeks now on a strix halo using Hermes, opencode and openwebui. Hereโs my opinion so far:
- Qwen 3.6 35B MoE is over rated for tasks that arenโt coding
- Gemma 4 26B MoE is great as an IT helpdesk/System admin
@jmbollenbacher@LottoLabs Idk man, the amount of growth coming from open weights and how fast it accelerates the industry as a whole will have them or their competitors still dropping. Anyone that needs big frontier models wouldnt run them on local hardware so the markets still there even w/ open weights.
@jmbollenbacher@LottoLabs Didnโt the ceo state he is going to continue pushing open weights for the community when people were afraid 3.5 was the last drop?
@AbedSeif65371@MargraveJuro@muke10101 Where did I say it was eternal? Youโre fighting points you made up in your own head. I said that everyone would suffer if it collapsed. Itโs also not just SWIFT, 50% global trades invoiced via USD - 60% international debt.
@AbedSeif65371@MargraveJuro@muke10101 Thatโs not how it works, energy prices are literally pegged to the dollar. If you are dumb enough to think the entire world can all collectively agree to use and prop up a secondary currency in a timely matter idk what to tell you.
@fromage006@MargraveJuro@muke10101 I never said it wouldnt be rebuildable but there isnt an alternative to the dollar to handle global trades. The USD is currently 51% of the share for SWIFT, do you even realize how big of a deal that is?