Answers the question: What are AI agents actually providing value to us, other than WallE style laziness and running real value out of the internet food chain.
Commercial photography costs money. Hard to reconcile how public goods are all in private hands, but new jersey ain't launching satellites, they're also not going to buy daily satellite images.
If NJ's inspectors can't be bothered to visit a site this large for 9 months, then would it do any good for the photography to be magically free?
If NJ's inspectors are honest, competent, and numerous enough to actually do the job, then would cheap drones be good enough to provide quick "big picture" views?
What a weird argument. If one doesn't trust private enterprise to "do right" then the state is nannying and oppressing companies, and if one does and therefore only inspects once a year or so then the state is neglecting their duties...
Public inspectors can't do right.
Maybe it is also fair to ask, did the developer/operator try to get away with something there? Do and hope for forgiveness instead of asking permission?
(and the "fine" is laughable, especially since retrospective permit is rather likely; in fact the fine might not even be higher than the cost of any project delay from requesting the generators at planning approval stage...)
What we do need is the death penalty to corporations. Reap them, take their cash and tell shareholders to do their due diligence when investing in criminal activities.
Why we keep protecting corporations just because there's a piece of paper that claims to do XYZ is such a bizarre world.
for comparison, Qwen3.8-Flash-Next only requires 6B parameters for computation, but stores 125B, 51B of those can be comfortably offloaded as they're not actively used in decode but a single token look up.
The Quant iQ4 of this model loads, then, in ~60GB of vram, and on disk it's 85GB.
So if you could etch it, you'd need a ~25GB ssd chip and 60GB of vram.
The vram costs likely contributed to these things being out of reach of the current economic cycle.
I'm not smart enough to know why; I do know that 27B is greater for short/interactive on blackwell, but the intellgence leap of the MoE in Qwen3.8-Flash-Next is quite remarkable.
I'm pretty convinced the pathway to local models will be MoE, especially if they can find a way to keep tweasing out things like PLE into the slow bandwidth lanes.
The Taalas architecture makes loading weights free, but they basically has to pay the same silicon for every weight, whether it's used or not. MoE models are more efficient than dense models per weight you load, but less efficient per weight you have to store.
MoE models are the path to local models with traditional system architectures, but they are antithetical to what Taalas was doing. If you spent all the money to etch 125B weights into silicon, you'd want to activate them all for each token, instead of only touching 6B. You cannot match the 125B sparse model with a 27B dense one, but you might be able to match it with a 60B or so one.
it never respected the ICC in the first place. When you examine a lot of America's behavior internationally, it's much the same way billionaires and oligarchs operate. This is mostly mask off.
America has decided they'd prefer to be a 2nd world country. Literally, as defined at the time.
Answers the question: What are AI agents actually providing value to us, other than WallE style laziness and running real value out of the internet food chain.
reply