@Teknium@waefrebeorn Thank you for the desktop app, it is SO MUCH EASIER AND SMOOTHER to use hermes agent rn
Faster to switch sessions, models, profiles, check on subagents, enable/disable skills and tools, add docs, pictures
Good UI and UX
Thank you
@Capetlevrai Les rust token killer ça fonctionne que sur les bash scripts, et ton LLM n'en run pas à chaque prompt
Donc ça marche mais c'est niche, pas placebo
Thats quite a stupid take actually
There's a sweetspot where saturated benchmark just means that in that area, the LLM is pro level of work ready and doesn't actually need a new disruptive upgrade
The new benchmark is actually good as it is very close to real coding, gpt 5.5 is the only model "saturating" the benchmark
Good to see chinese models where they really are, or at least how they felt to everyone despite having artificial analysis benchmark super close to closed models
@Polymarket Well its super ass and don't even come close to any chinese EV
Jony Ive was never a good designer, he rided that Jobs wave and impregnated himself with Jobs talent
When he died, Ive's "talent" died too