That soup is already cooked. And obtaining consent is prohibitive, so your data set will be tiny.

I don’t crave it, but lite beer definitely helps to dull hunger. Less calories than eating solid food.

Yeah, Nature has certainly declined in quality over the last decades.

I would expect more from multijunction cells. Wikipedia has a very nice article https://en.wikipedia.org/wiki/Solar_cell_efficiency

https://www.nature.com/articles/s41586-026-11223-x

Abstract

Kirchhoff demonstrated the reciprocity between absorptivity and emissivity for blackbody radiation in 1860 (ref. 1), thereby establishing the general principle that a good absorber must also be a good emitter. This principle underpins the generalized Planck’s law2 and, when applied to semiconductors—in the form of the van Roosbroeck–Shockley equation3—results in substantial limitations to solar energy conversion. Here we demonstrate how this reciprocity can be circumvented in semiconductors, such as bismuth sulfide, showing a pseudo-direct bandgap character—that is, featuring a lower-energy indirect transition paired with a direct transition placed a few kBT above. After determining the pseudo-direct bandgap nature of bismuth sulfide, we reveal how this feature substantially retards bimolecular electron–hole recombination in this material with respect to classic direct semiconductors4. The resulting combination of a high absorption coefficient and delayed recombination offers the unprecedented possibility of overcoming limitations associated with Planck’s law, with far-reaching potential for solar energy conversion.

Are we talking about the same book https://en.wikipedia.org/wiki/The_Selfish_Gene ? What on Earth is a “darwinian solution”?

The Nature article is about plain darwinian evolution though.

That is actually an advantage. Our massive advantage though is modern analytics and modern industry.

Fahrenheit? Really?

Abstract

The commensurate charge density wave (CCDW) phase in 1T-TaS2 is a prototypical correlated state tied to diverse emergent quantum phenomena. However, its stability is intrinsically limited to cryogenic temperatures. Tailoring correlated electronic states and stabilizing the CCDW phase at room temperature offer new opportunities to investigate and control emergent phenomena under ambient conditions. Here we report a molecular engineering strategy to modulate the charge density wave (CDW) phases in 1T-TaS2/arginine hybrid superlattices. We show that the intercalated arginine molecules exhibit strong non-covalent interactions with the 1T-TaS2 atomic layers, effectively modulating the correlated electronic phase and increasing the CCDW stabilization temperature to 330 K. In contrast, other weakly interacting amino acids produce a less pronounced modulation of the CDW phase transitions. Computational modelling reveals that arginine molecules reshape the CDW energy landscape, enhancing both the thermodynamic stability and kinetic robustness against thermal fluctuations. These findings establish a versatile molecular engineering strategy for tailoring correlated states and electronic orders in van der Waals materials.

The last time SciAm was good was in the 1970s.

It’s a standing on the shoulders of giants thing. That’s how humanity have been making progress historically.

People running local LLMs with open weights abliterated models, especially with solar PV power (perhaps not as rare as you think, I’m upgrading my solar PV system in the next couple weeks for that explicit purpose), is about freedom and privacy. At some point there should be enough of them to link up into a grid over the Internet, collaborate or donate resources to distributed community projects. This is a way to allow for the “pillaged” content to pay back for itself.

Do you have qualms with public pirate libraries, like SciHub and LibGen, too? These have also been used for training LLMs.

Whoah. Have you ever thought about making some videos demoeing and explaining your setup?

Yeah, a useful Ryzen AI Max is also some 6 k$ which is also wildly out of budget for a tinkerbox.

A fanless thin client with Opnsense and a four-port Intel NIC.

If I fire up everything in the rack it’s several kilowatts. Probably more than a 230 V wall socket can handle.

Usually it’s way less, though right now it’s just a switch and the router/firewall, mere 57 W.

I make most of my own net power with solar PV, soon 4 kWp in total.

You still need to log in. Hard pass.

If you need to move the document bulk, you can either rsync the docs and repin them, or use IPFS native means of replication to remote – haven’t used those.

LibGen on IPFS used to work, try e.g. https://bafybeifaxsen4np54stc43phpnn3rh2crxie34d74yki43dultrh4vglqy/ – but I have no idea about how much of the library corpus is pinned currently.

I find LLMs very useful in my professional (senior IT consultant) and hobby IT life.

Which is not a problem for tens to hundres of terabytes of library documents, which are all static. And there’s https://docs.ipfs.tech/concepts/ipns/ and https://docs.ipfs.tech/concepts/dnslink/ but latter suffers the same problems as DNS.

IPFS is tricky, but can be made to work, at least at Libgen scale. Content discoverability is always a problem, so depends on the number of pins on the network.