@N_Rikhil@1337hero I haven't had anything useful, I've seen some people running harnesses with qwen 27b as executor, but I have not had good results myself. Local stuff isn't there imo, and let's not mention running multiple agents/models on local hardware
@transkatgirl@cha0s10g1c If the market prices of models reflect their intelligence and time to completion, I view that trying to build systems on subpar models requires more time investment to chase the hallucinations and errors, the price of the model correlates with prowess. Pay with money or time
@sambam_at@brubrushanghai@mrbcyber a very miniscule portion of the collective knowledge of the US defense industry, it's similar to how Chinese LLM models steal training data from opus, stealing information from another country only gets you so far, supply chain is much more than raw materials
@sambam_at@brubrushanghai@mrbcyber raw materials is just the starting point, it's the millions of combined man hours of research and classified information that's the moat