I have used native dataclasses with pydantic starting with v1.5 and it’s only improved with the the newer 2.x releases.
Aside from basic inheritance and complex nested types, the pydantic ‘TypeAdapter’ is awesome for simply validating native dataclasses. It’s a little slow, but everything pydantic is =)
Maybe grassy wasn’t the right term. I don’t mean under developed, it’s kinda bitter and sharp without any sweetness. I’ve experienced this with pretty legit roasters like apollons gold. It seems like the lighter the roast, the more time it needs to degas before it hits the sweet spot.
> But maybe the CO2 removal stuff really is a big deal.
It’s definitely a thing. There are some super light roasts that if you try to brew them without enough resting time, you can sort of cheat by just grinding the coffee about 30mins before you brew for a hacky quick degas. It really helps get rid of that grassy vegetal taste.
16:1 water:coffee ratio is a pretty common starting point. Your brew is pretty intense!
I’ve found that darker roasts are much nicer around 195F, I mainly brew very light roast and almost all of them come out under extracted at temps below 200F
Where does the manifesto talk about a process of any sort? It’s a set of values. It doesn’t mention anything about iterations, breaking things down, etc
The “resting time” after roasting depends heavily on the bean and roast level.
Soft beans that are roasted dark don’t require much time since most of the oils have been brought close to the surface already.
Very dense beans roasted lightly take time to degas. In my experience 4-7 days is the sweet spot.
It’s not even a subtle difference. Drinking a very light roast right after roasting will be very sharp/sour/acidic, many times undrinkable. After a several days they will get sweet and fruity and delicious.
Regulating horizontal integration seems like it would be very difficult. How exactly would those horizons be clearly defined?
I’ve always thought of Apple having an emphasis on vertical integration(hardware+software=product). So maybe their foray into services would be the line in the sand?
This is a take on the Yamaha Tenori-on[0] style of sequencing which is legendary(but sadly no longer made).
I’m surprised there aren’t more software clones of this available. Even the official iOS app clone was discontinued a few years back. It’s such a fun and highly creative way to sequence, though better suited to melodic phrases than rhythmic patterns imo.
I use a Sonnet egpu box with an nvidia 1080ti for casual gaming on a framework 11th Gen running arch.
It mostly worked with minimal fuss. From what I remember, most of the headaches were getting sway and my multi monitor support configured to my liking.
> The future of desktop Linux is a VM under Windows.
Is there really any demand for this though? What Linux gui apps exist that windows users want/need? Enough that Microsoft sees an addressable market large enough to get roi?
The inverse seems to have way more practical use cases that could actually drive revenue - games and legacy business applications (as mentioned in the esr prose).
Aside from basic inheritance and complex nested types, the pydantic ‘TypeAdapter’ is awesome for simply validating native dataclasses. It’s a little slow, but everything pydantic is =)