I've recently run into and explored research explaining that natural sunlight stimulates energy production in mitochondria, resulting in an energizing and healing effect completely separate from vitamin D production.
I feel bad for posting this without references so here's one reference I could find right now:
> This study shows that longer wavelengths in sunlight are transmitted through the body. When these are presented via NIR LEDs in a laboratory-controlled environment at much lower energies, they again have the ability to be transmitted through the body and also are associated with improved visual function independent of ocular exposure. Hence, body penetration by longer wavelengths impacts systemically. Longer wavelengths improve mitochondrial membrane potential and ATP production and improve function in a species conserved pattern that can impact on mobility, visual function and cognition, particularly in ageing. In short lived animals they can extend average life span.
> Taken together, these data provide evidence of the importance of the full spectrum of sunlight for human health. They also highlight the potential dangers of the restricted spectra found in white LED lighting in the modern built environment that lacks longer wavelengths and whose output is generally restricted to around 400–650 nm. The absence of longer wavelengths from LED light sources may have implications for public health that should be addressed.
One interesting aspect of Vitamin D that is not regularly discussed is the photo-isomerization of Vitamin D into inactive isomers in the skin.
I think of it similarly to Beta-Carotene, in that consuming the "natural source" has no known toxic amount, while consuming processed vitamin A can be toxic in large doses.
Increasing sun exposure will not raise blood vitamin D metabolite concentration indefinitely, and we don't have good studies that accurately measure the amounts of Vitamin D and its photo-isomers and their concentration in relation to large doses of UV exposure.
Therefore, through sun exposure, people may be storing large doses of Vitamin D isomers in their skin, keeping their levels much higher than typically prescribed oral supplementation levels.
> My dream would be a local model that can do, say, 80% of the day to day tasks I need; "how does X Handler connect to Y storage?", "commit that feature, but leave out the bits that relate to billing" etc.
Qwen 3.6 27B can do that today, but setup properly and in a good quant, I run an autoround [0] with weights in int8 and attention heads in f16 on a single RTX 6000 Pro Blackwell Max-Q via vllm with mtp=2 and full context, --max-num-seqs 3, KV in f16, mamba f32.
>It would have 99% reliable tool calling
I managed to score 93/100 in tool-eval-bench [1]. For me this is very good already, at least in the pi coding harness I've never had an issue that wasn't auto-fixed in the next turn(s).
>the ability to go "this task is beyond my skills" and refer to a Big Boy Online Model in a gigantic datacenter somewhere
This is heavy on the harness engineering side I think, but also quite contrary to the nature of LLMs today. If you figure this out I'd love to know.
Hard to know when they don't give the price per token. Presumably it will be comparable to a low-mid range model in terms of price. But otherwise their 'Ideal Zone' is meaningless without factoring in the price per token. I don't how much tokens are being used, that's an implementation detail to me. I care about price / performance / latency.
Yeah the future is probably a number of highly specialised small models you can run on your own hardware rather than massive frontier models in the cloud.
That seems to be what Microsoft is betting on also based on what was shown at the BUILD keynote today + that new surface ultra and the surface mini PC with the new Nvidia chip. Nadella really played up local AI as the main use case they have in mind.
MOE basically work that way already, QWEN/etc with low active params (A-number in name) allows to inference big models locally (only active params have to fit into memory)
There will always be tasks that are withing reach of whatever the SOTA models are, but not of the cheaper, perhaps locally runnable ones. It seems that already people are finding Qwen 3.6 27B sufficient for many coding tasks (the llama.cpp author is now using it exclusively).
As models get better and smaller, I expect that we will rapidly (within a year?) get to the point where SOTA models are not needed for the vast majority of coding tasks, and even today it seems many people are just using them for the planning phase.
How many people drive Ferraris vs Fords? How many people driving a Ford would, on a utilitarian basis, be any better off driving a Ferrari?
So far there seems to be mainly two high volume use cases that have been found for LLMs - coding and business flow automation, and it seems neither of these need SOTA models. I wonder if there will continue to be enough market demand for massive expensive SOTA models to make them worthwhile developing?
