I have a history of asking incredibly basic questions about the assumptions that are going in to a debugging issue for exactly these kinds of reasons. Something changed somewhere; so have we verified our basic assumptions that things are what we think they are... or did something change under our feet without our knowing? Or maybe we actually have a different understanding of the state of the world, and one of us is actually wrong?
It's not fruitful every time, but the number of times you skip asking those questions and chase something around, only to realize that, actually, it _was_ the ridiculous/simple thing... well, it feels a lot better to realize it earlier :)
(personal footnote: do be mindful of the context you're asking in! I feel like in general those sorts of questions are _fairly_ well tolerated by technical people. Sometimes you feel a bit stupid asking the question, but many of the systems I've worked on are large and complicated enough that it's not that weird - and if there's downtime and a healthy team, people genuinely want to hunt every possibility down. People not used to that, though, will sometimes bristle a bit, thinking you're implying that they're stupid and don't understand the basics. There can be ways to be more tactful, but at the very least, it's worth being aware of that reaction so you can follow up with a "just making sure I understand!" or something like that.)
Given they’ve had several skirmishes with customs and law enforcement agencies around the world, this always struck me as similar to the “don’t talk about installing retail Switch games on the Switch modding Discord” type of deal - everyone knows you can do that, but allowing mentions in official channels opens us to liability and causes nothing but headaches for both us and for customers, so if you’re going to do that, you need to talk about it somewhere else. I freely admit that’s an assumption on my part, though, and I don’t know if there’s something uglier there…?
To be clear, while there is _specific_ language around DEI/gender identity, the actual language of parts of this goes far, far further, and essentially codifies giving political appointees explicit discretion over what gets funded and what conferences scientists with government funds can go to, and disallows any government money to go towards publishing research.
I just finished a video game called 1000xResist this afternoon. There's a lot going on, to say the least - a sci-fi story involving an alien virus, intergenerational trauma, what it's like to be an outsider. It's a hell of a ride, one where certain moments resonated in a way I'll think about for a long time to come.
One of the dynamics is that a character's parents, both of whom protested in Hong Kong during the Umbrella Movement, left for Canada in the wake of the crackdowns wanting to start a family. At a moment when both of the parents are tired and feeling regret, one of them asks why they left, why they bothered to protest, and if those actions had any meaning if the PRC wound up winning control anyway. The other says this:
> ...if we stayed silent? Didn't stand up for ourselves? They would say this is how it always was. They would say this is what the people wanted. But no. They can't say that. Because it has gone down in history that we resisted fiercely. That we fought for a different future until we couldn't.
I admit: ultimately, that statement doesn't mean anything quantifiable - in fact, it kind of states the exact opposite, which is not the most convincing on a site like this. Still, I think there is truth in it: even if the protests don't have a quantifiable number associated with them, people see them, and know that they happened.
Ultimately that may or may not matter; it may just be a sentiment lost in the wind, or papered over by the victors. But it's still _something_.
The reasoning tokens are really just there to extend the amount the LLM can "compute" the problem; put another way, the only way a given model can "think" more about a problem is to fill more of its context with predicted tokens, which has the effect of increasing the accuracy of each token. The reinforcement learning these models go through generally doesn't care what the chain of thought tokens look like (outside of preventing loops/gibberish/reward hacking), only how good the final answer is - so while it does look something like "reasoning" to us and has a rough correlation with the final answer, treating it as actually representative of what the final answer will be or an actual thought process is giving those tokens too much credit :)
I would buy this argument if Flash as a browser plugin had been proven to perform well on a mobile device of the time, but it never was, on Android or any other platform.
Even AIR apps - think Electron, an application shell for Flash apps - were on the edge of usable on desktop Macs of the era.
For the first year, Scott Forstall, the Senior Vice President in charge of the iPhone's software, very directly encouraged companies like Pandora[0] to jailbreak iPhones in order to get a head start on app development, protected that community from Steve Jobs' ire, and then used the existence and popularity of jailbreaking to convince Steve that a sandboxed app store would be a better idea than Apple writing every single app for the iPhone[1].
Once native APIs were available, that was true, but before it was even clear that the iPhone would have an app store, they very much did let it flourish.
My understanding is that the case was flimsy enough that no "hardcore" lawyers wanted to represent him. It's not just a matter of money; their record (and, therefore, future earnings) are on the line.
At least as of when I left the company, GitHub was being deployed to fairly close to once every 60-90 minutes (the frequency of a deploy train/merge queue batch going out) 24 hours a day, at least during weekdays… there are a fair number of international engineers and deploy trains get crowded during main US business hours, so while there are fewer PRs going out at odd hours US time, there were typically still some. There aren’t dedicated releases as such for GitHub-hosted instances - everything you release needs to be gated behind a feature flag or other mechanism if it’s not going live immediately, and your code either needs to handle the database in both its pre- and post-migrated state, or you need to run the migration in advance of your code shipping out.
Fun fact: it used to be the case that GitHub was actually _less_ reliable if nobody deployed to it… there used to be various resource leaks that we didn’t see when people were deploying all day, since then the app wasn’t getting restarted constantly. After GitHub went down during a holiday break we had volunteers to deploy GitHub once a day during holiday breaks, until the underlying issues were eventually fixed.
Oh, I completely agree. But I admit I'm a bit of a retro enthusiast and preservationist at heart, so I was curious what "mattered" to a rando HN commenter :P
Wow - I remember that from the System 7 days. I thought that was long gone - how did I not notice that checkbox was still there in Get Info this whole time?
EDIT: For anyone unfamiliar, the MiSTer is a homebrew FPGA project originally built around a Terasic DE-10 Nano that can emulate in hardware a wide variety of consoles and computers, leading to extremely low latency and (often) higher accuracy than most software emulators due to it being easier/more efficient to recreate cycle-accurate effects in hardware. It’s extremely flexible, allowing for both HDMI and analog output (with scaler effects, if desired), as well as both modern USB/Bluetooth HIDs as well as adapters for original controllers. It’s a very cool project and worth checking out if you’re enthusiastic about such things - retroremake.co has had some well-liked clone/re-engineerings of the MiSTer hardware but they’re going through a big shipping backlog so I don’t know when they’re in stock; there were some decently regarded Aliexpress clones as well, but I don’t know what the status of those is. An authentic DE-10 Nano is an option too, it’ll just be more expensive and you’ll still need to get an SDRAM board to run most cores.
Exactly the first thought I had too. I know extremely little about FPGA development, but three things I noticed that came to mind re: difficulty:
- Alex used a Xilinx FPGA, the MiSTer uses an Altera Cyclone - dunno how portable code (if that’s even the right term for e.g. VHDL) is from one to the other. I know the MiSTer has a light framework for cores to plug into to get input handling, scalers, etc.; so maybe it’s more a matter of porting to the framework…?
- Alex mentioned the SCC didn’t have a pre-made FPGA core so they used a real one. I don’t think serial handling would be critical but I do suspect you’d at least need a dummy to get the OS to pass self-tests and boot properly. Possible that maybe the Mac core has already handled this, though.
- What little I know of RAM and the MiSTer would lead me to think the SDRAM card a MiSTer setup typically needs wouldn’t be a problem over the SRAM Alex used, and that either the framework or the wiring of the RAM card handles the details for you - but I definitely don’t know that.
On the plus side I suspect/hope maybe a bunch of stuff from the classic/original Mac core could be borrowed to get it up and running.
There’s definitely plenty of cores that haven’t yet been developed on the MiSTer… for instance there isn’t a color 68K Mac core, only recently have people started on 3D0 and CD-i and Apple IIgs cores, the Saturn core was pretty shaky until a recent overhaul, etc. I think what’s there is just a function of what was either already developed for an FPGA or what had the biggest demand from their respective communities.
I just watched the Vertasium video[1] on ASML's EUV lithgoraphy machines over the weekend, and I think the qualifier they used was "most complex machine _you can purchase commercially_".
I can't remember if it was an ASML representative that said that, or if it was an overlaid asterisk that popped up on the screen at some point - but I definitely remember thinking about the space shuttle and Saturn V/Apollo and those sorts of things before I saw the qualifier.
It's a function of the shape. On a capsule-sized spacecraft, the ionized plasma completely surrounds the craft, so no radio communications can get in or out. For an oblong-shaped spacecraft, like the Space Shuttle or Starship, the descent tends to be angled such that you have a "hole" in the plasma you can get a signal through.
You're getting lucky with the games you're playing, then; there are absolutely PC games that have had 20-30 minute long shader compilation times _on high-end gaming hardware_. (I think some of Sony's ports were known for this; Googling tells me Borderlands 4, Stalker 2, and Starfield also had notably long shader times.) Typically those occur within the game's UI after launch but before the game starts playing, though, which makes me wonder if Valve might still be caching a non-GPU-specific intermediate of the DX12 to Vulkan conversion, and _that's_ what Linux Steam clients are compiling pre-launch and/or sharing with other clients. That's pure speculation on my part though, as I haven't played any of the worst-case-scenario games on my Deck, nor have I done anything that would cause the shader downloading to not operate.