jodleif

Also, the language model might not fully understand the code then add a comment, then the next iteration will treat assumptions in the comment as the truth.

Then you might be missing SWA. Gemma models are extremely memory hungry without

Its about the angle yes, but the main thing is basically for every square unit of solar radiation gets projected onto a larger surface because of the angle- meaning radiation per unit area is lower. The secondary effect is that sunlight has to travel through more atmosphere, and gets scattered somewhat.

Would be really nice. Seems like even android is getting more and more locked down

Yes, except for the odd anti-cheat that needs kernel level access

Threadripper is a good alternative. No point having a lot of dual channel ram for LLMs, too slow

In any functioning sports body testing is random and unannounced

As long as you tune the JVM right it can be faster. But its a big if with the tune, and you need to write performant code

Agreed. Also, it takes time to understand a domain properly- so the innate slowness of coding helps with letting things “simmer” in the back of the mind.

I found swift way more enjoyable than rust as a C++ alternative. It even has first class-ish interop now.

Deterministic uuids is a very standard usecase

The sad part is the “buddy hackathon” is kind of redundant now

For humans org-mode is good at this

I find this to be a bit “easy”. There is such a thing as bad tools. If it is difficult to determine if the tool is good or bad i’d say some of the blame has to be put on the tool.

I assume these needs to be indications on overall diet.

Edit: i.e a bacon eater consumes a higher than average caloric intake, hazelnut eaters have more greens/vegetables in their diet possibly

The huge difference with f1 is that each team more or less make their own _bespoke_ car (except engine). With indycar is much more of a spec series where Dallara(?) makes the chassies for each team.

A common pattern is to offload (most of) the expert layers to the CPU. This combination is still quite fast even with slow system ram, though obviously inferior to a pure VRAM loading

1. Have you seen the Qwen offerings? They have great multi-modality, some even SOTA.

Assuming your hardware premise is right (and lets be honest, nobody really wants to send their data to chinese providers) You can use a provider like Cerebras, Groq?

> Valuation is not based on what they have done but what they might do

Exactly what I’m thinking. Chinese models catching rapidly. Soon to be on-par with the big dogs.

I genuinely do not understand the evaluations of the US AI industry. The chinese models are so close and far cheaper

Did you mean list<a> -> b? To my mind a transducer is a tranform + reducer

This Carmack guy knows his stuff.

Ive always found the silversearcher defaults (or perhaps -Q) to just work better. Any alias/config tips?

This is really old now (I think 2000s) but even the jacks they lift the car with doesn’t drop with gravity alone - that’s too slow. They have some mechanism that actively drops the jack, so they can pull it away faster.

Recently they’ve done some shenanigans with extension support. it seems to me like a change only for the worse?

IMO it’s really hard to argue with the quality valve software is putting out. I can’t really name a game that isn’t considered a must play that they’ve made

Qwen have their own competitor to Claude Code.

I actually liked windows 7 quite a bit

He said he was very dedicated, but turned out he wasn’t (surprise). What else could he be lying about? Surely not his work ethic