Thanks for your work. Moved my company from Slack to self-hosted Zulip after Salesforce unilaterally decided to transfer our account to Alibaba Cloud and the transition has been very smooth. We especially appreciate proper markdown support !
The slide is so bad it might be intentional. The intent would be to be reassuring enough that no one will point at Boeing and say "Boeing technical failure strands astronauts in space", while providing just enough information that in the event (at the time of writing the slide considered very unlikely) things go bad, they can say that nothing was withheld.
Yeah, the problem here isn't Powerpoint, it's what it said. This is a case of shoot the messenger. Powerpoint doesn't hinder communication, it lets people think they understand when their actual knowledge is superficial and can easily be a case of the primrose path.
From a quick look, generalized constexpr evaluation, but in practice it relies on parsing non-portable decorated function names from things like source location. An ugly, slow, but effective hack.
> do if the ions you’re using are collected from the atmosphere around you rather than coming from a tank on your vehicle
This [1] looks like a plasma propulsion engine [2]. They'd need to carry propellant. That said, they're pre-staging, so the traditional rocket equation doesn't apply.
I find it odd that twitch would rather shut down their service than simply stop offering a free tier. They already have an ad-free service called Twitch Turbo, surely enforcing usage in Korea wouldn't be prohibitively difficult.
This makes me think that they are trying to spark outrage.
I think that would decimate their user numbers, which is a bad deal for the streamers too since they make money on sponsorships/ads and based on audience size. Hard to grow an audience when folks have to pay for the service. No idea how much of an average size streamer's income is based on donations on stream compared to brand deals. I wouldn't expect the average user to both pay for a sub to twitch, and also donate to the streamers.
I won't pretend to know what Twitch knows, but watching for free is definitely integral to how Twitch acquires users.
That's not obvious. If the pay-to-watch version of Twitch isn't viable, it will have no streamers and no users, but it will cost more than the we-no-longer-exist version.
A fixed price tier cannot cover the full cost of that user's variable stream usage.
Twitch is unable to charge per hr streamed - nobody would accept that pricing, esp. if it was originally free.
Removing twitch from korea is the best move, since koreans are the ones losing out - which hopefully as a democracy, they elect someone to change this sort of lobbying from the ISP industry.
It might work short-term, but without a free tier, can they get enough new customers to replace subscriber loss over time? Otherwise they'll just slowly bleed subscribers until it's no longer profitable to run transit to and CDNs in Korea.
Twitch's market share in Korea is very small. Domestic streaming services like Afreeca are the dominant player. If I had to guess, Twitch doesn't have enough users in Korea for a paywalled system to turn things profitable.
They're not, and it's a very annoying requirement. Even 5V/3A was hard to provide reliably. On the other hand, rock pi supports 12V/2A, which modern chargers will provide without any issue.
For some, a few days of GDScript usage. For others, after they convert their bottlenecks into C++.
UE’s blueprints are much(!) slower than native code, but they’re still an incredible tool in the toolbox. GDScript is kind of the same; it’s covering a critical role of non-performance-sensitive glue.
That said, I’d really like to see Value Types added, as well as these sorts of heap allocations addressed in Godot 4 at some point.
It’s not about spending time, it’s about being ignorant of GDScript’s use cases. It’s like if people went into Python communities and shat all over it because C is faster. It’s true, but it misses the point that iterating in Python is leagues faster and is fast enough for a large section of projects.
If C bindings were slow because the underlying code was made to bridge with Python bindings, I would complain too, to be fair. Doesn't mean removing Python support is the perfect solution, but at some point that performance hurts everyone.
(not that that should be an issue due to python being able to directly output C code, but you understand the point).
I guess the argument is that GDScript has a different use case than Unity's C#? Seems like Unity devs know what a C# scripting engine is like and feel it covers the use case fine. To say they don't understand the use case would mean it must be different somehow. Can you go into that?
So are we fine keeping it that way and not addressing core performance issues because it's convinent for making simple 2D games?
To be honest, I'm surprised how dismissive parts of the gamedev community can be about improving performance of their game engine. That was a huge complaint from Unity and you'd think people would be interested in not repeating those mistakes with Godot.
Hardcore vim user with UHK here. I actually love that the caps lock key is remapped to esc in the firmware, so I always have the Esc where I expect it on any computer I plug it in (this is not unique to UHK of course).
In fact on the UHK, there is no "caps lock" key, it reads "mouse".
Despite being two keypresses, this is usually faster to type, or at the least to return to home position from, and those keys are probably more consistently placed than the ESC key between vendors / models.
This feels messier, partly because it's easier to fudge accidentally, but very much mostly because it relies on your configuration being propagated. I understand that may work in some shops, but generally it can't be assumed.
ctrl [ works everywhere vim is installed - no customisation required at all.
What practical benefit does that offer over existing (synonym-aware) keyword and phrase search approaches? The corpus of one’s mailbox is too small a dataset to draw conclusion from, surely?
The really difficult parts about keyword search are tokenization, normalization, and synonym selection. Especially keeping them up to date. The uses of search never know about any of this, but as the developer these things need to be top of mind.
Using embeddings basically lets the AI configure those things for you and auto updates when the AI updates.
You could also use the embeddings for far more advanced things like in LLMs, but the basic version that is just “better keyword search” is also valuable.
> Not to mention being far slower to query.
KNN on the embeddings is not obviously slower to query. In production using AWS ElasticSearch, for a very large search index, my team saw no meaningful change to latency when using embeddings instead.
If this is for my personal e-mail, then I'd only use it if it could run locally - which also means inobtrusively in the background, not slowing down my computer - is that feasible?
