That's not my experience, the model is adding tons of unnecessary details, that make it much harder to find the important information for an expert that already knows all those details.

I don't want ELI5, I just want the model to go straight to the point, like a colleague human expert would do.

I don't know, that's exactly what Opus 5.5 feels like to me.

Why do you assume this order? I would assume the AI starts with lean (that's what it was RL'd on at least) then tried to translate the proof for us humans, which is quite hard.

I'm not sure about it but it seems plausible given the potential error.

Edit: they explicite say that your order is correct in the post

Why not using a cheap LLM with thinking completely disabled ? I don't think it will be much more expensive than jev.

I’ve tested this with some local LLMs and their accuracy is in general better than Jev/Laya, but they are super slow in comparison as well

For example, a typical/stock LLM can’t really play Doom in real time, but a Jev-like model can. Just because of latency

Of course, if you want the best Doom player, there are way better and faster adhoc models

LLM inference has two very different regimes of work: prefill & decode. You can think of the former roughly as processing a pre-specified prompt, and the latter as sequential processing (auto-regressive token generation) eg. "chain of thought". The latter is very important for LLMs and cannot be ignored; it deeply influences infra design, even necessitates copious amounts of high-bandwidth memory. Jev-like models can ignore the latter and therefore optimize much better for the former, consequently operating at both better cost and latency.

Strange they don't mention the Pink Floyd album in the article!

Yeah, I expected the photo of the cover https://en.wikipedia.org/wiki/Animals_(Pink_Floyd_album)

Hallucinations were defined in the context of text generation models so your question does not really make sense.

IMO your system can make mistakes that are similar in spirit to hallucination (i.e. answering with a false answer instead of abstaining to answer).

And furthermore, because the model is forced to answer in a boolean (if in boolean mode), if the user input is outside of the range of a boolean, it's forced to hallucinate. It can't abstain.

We didn't have the compute required (GPUs powerful enough to parallelize forward and backward pass). This compute is what allows us to train from human knowledge or distillation.

UUID as key is useful when you have a distributed system where multiple workers create items independently

Could you share your config?

It's fairly sprawling right now — a small init.lua that sources four other files, most over 100 LOC. What are you most interested in?

The heat pump will pull heat from inside the house? This sounds terrible for efficiency in winter, as you will need to reheat the room

There are systems (like the sanco2) that use an indoor/outdoor pump.

> This sounds terrible for efficiency in winter, as you will need to reheat the room

Sure, but lots of people have some point of the year they want cooling.

Even during the heating season it's only worse if you're heating the living space with something _worse_ than what you're using to heat the hot water. If you have a heat pump for room heat then you're moving heat from outside, to in the house, to in the water heater.

If you're heating the room with electric then in the winter it's no different than using an electric water heater (100% efficient).

Where does it day that? Would be great if it chose inside or outside based on target inside temp (e.g. cools indoor air in summer to heat water for washing and showers)

Looking at pictures it must pull heat from ambient air or from electricity. There is not enough tubes to have interior and external heat exchanging units.