Cited paper with more context. arxiv.org/pdf/2512.14575

Yes there's no intelligence there (in the way we think of organic intelligence) and they are advanced pattern machines, but even pattern matching machines can be useful. Especially the kind that can find intersecting patterns between unrelated domains.

People want to capitalize on the hype - more semantic diffusion works well under that incentive structure.

Half-life 3 is confirmed.

I think this is the crux of the issue. It's much more useful to someone who has enough knowledge to identify mistakes and is diligent enough to do their own fact-checking than to the average layperson who may turn to it as an authoritative source of information.

www.aljazeera.com/economy/2025...

What I usually get is - I tried asking <LLM> because I didn't want to waste your time.

Appreciated, but I actually enjoyed answering the small questions. That said, more autonomy/independence is usually a net positive.

I'd be down to try it out if you're still looking for people

While prompting about text sampling strategies it randomly decided to include a diagram of the human skull

Then we'll ship your machine

One of my junior devs recently asked about what life was like before git. Explaining SVN and Mercurial felt like I was reading off of stone tablets.

Yep, that makes sense. TY for sharing. This is fascinating stuff.

"MOSAIC can flexibly predict galaxy properties from partial inputs: it can perform photometric redshift and stellar mass estimation, reconstruct SEDs and SFHs, and translate between photometric and spectroscopic measurements from different telescopes."

Example I found looking around arxiv.org/html/2510.07... is trained on the output of the simulations in order to make predictions on new data.

I don't know much about how HACC is used, but if it's reading internet data, then that is likely a general-purpose language model being prompted to generate code to run simulations.

Not commenting either way. I just thought op's question was a good one, and your response motivated me to learn more.

The lights, cameras, switches, cabling, routing infrastructure, and maintenance upkeep add up. A pilot in Toronto estimated CAN$100k (US$80k) and CAN$150,000 (US$120,000) per intersection in 2018.