@CapitalObserver I think this is largely correct, but the models keep becoming more efficient as well. So AI spend might decrease at largest companies (but increase overall as more people use it), but AI impact on productivity will still increase.
@brandonjcarl The issue with this logic is that most AI models operate on a cost/intelligence curve. Frontier models (sorry, tractors) have very intelligent/expensive models, but they also have much cheaper models that are still better than non-OS models at the same cost. The threat is OS.
@emollick The visual analogy would be going from 1080p to 8k resolution. Significant technological jump, but really hard to tell with the naked eye for most people.
@RRiscio37389@emollick GPT5.5 Instant makes the same mistake.
But GPT5.5 Thinking (both standard and extended) get it right.
I think the issue is that all non-thinking models are garbage.
@ChrisTrapasso Well I've been told quite convincingly that Beane sucks at drafting (excluding Allen), so clearly your analysis is missing something ๐
How did you incorporate the "Beane Sucks" variable? It might need a higher weighting.
@ChrisTrapasso@BuffRumblings Very interesting. Is it relatively easy to recalculate / rank by ignoring all QB picks? I feel like that probably skews it massively.
@CapitalObserver AI likely helps a lot with top of funnel. But then you have more real interest and need more sales people to actually on-board / convert?
Not sure, but could make sense.