giantrobot

> So what do you think it is that makes Reddit feel icky, versus the forums we've had in days of yore?

Votes and karma. The concept of karma isn't bad but Reddit's implementation is bad. HN has a similar problem. There's only up and down votes with no context.

Why was a post upvoted or downvoted? Did people find it funny or informative? Did it correct the original link or a parent post? No fucking clue. Just scalar value that goes up and down. Then you get bad proxies for context with vote ratios (controversial etc).

Slashdot's voting system isn't perfect but much better. Instead of some scalar value a post is moderated with some sort of context (Interesting, Informative, etc). Voting is also distributed among users and limited. If you get moderator points you're limited in how many you get and you can't post to a thread you've moderated. Meta moderation also helps to correct for overzealous/reactionary moderation.

The context is useful because you get an idea of why something was moderated as it was. The limited number and random distribution of moderation points helps inspire actual moderation. You either downvote obviously trolling/bad faith posts or upvote actually constructive posts. No one wastes their mod points on memes and "me too" posts. Drive by memes and short knee jerk replies just wallow in the Score: 1 area most users quickly learn to filter out. They're not punished but also not rewarded.

Giving everyone unlimited mod points devolves a site into gaming karma. Trolls enjoy seeing how low their karma can get while karma whores will post whatever gets upvotes rather than something constructive. Karma is a bad proxy for constructive contributions to the community.

There was also CD+Graphics which encoded graphical information in a subchannel not used by CD audio. It ended up providing about 28kbps of data for the graphics. It is used for karaoke but some commercial albums were released that displayed images of the bands and stuff when played on a compatible player.

> The problem, for me, is that self-running installers can create a mess that's hard to keep track of.

For reasons my machine with a beefy GPU is stuck on an older set of Nvidia drivers. I've got them pinned with apt. The ollama installer fucked everything up by updating stuff that apparently wasn't pinned. After I had fun cleaning up that fucking mess I found it overwrote my custom systemd service file so I had to go in and fix that.

Curl-to-shell is a bullshit antipattern. I have no interest in going back to the dark days of expanding tarballs to / and hoping for the best.

> so it makes you wonder what the fuss about ALPRs is and how it affects them.

Wow. Poe's Law in full effect here. I'm not sure if you're just oblivious or trolling. If you're oblivious do yourself a huge fucking favor and read up on what Flock is doing in addition to ALPR and how even that is abused by law enforcement.

You're right I was forgetting the Falcon Heavy. But it's certainly not cheap when the government is footing the bill. The Nancy Grace Roman launch cost NASA $255m, far above SpaceX's prices for even an expendable GTO launch.

There's a vast gulf between Android and a Linux desktop environment. Android is a Linux kernel with the Android environment on top and is not just some Linux distro like anyone uses on a PC. ChromeOS is very similar.

The war in Iran costs about $95m a day and is already over $43b cost to the taxpayers, just in military spending. That's nearly two NASA budgets in just over 200 days.

The "powers that be" chose to lose a completely optional and ill-conceived war on your dime. Instead of science or helping people they decided to spend money blowing up schoolgirls.

Some types of launches to some classes or orbits have gotten cheaper. Not all launches to all orbits. I don't know if any commercial launch services can even do planetary transfer orbits.

In space everything is an orbital change. To change orbits you need to speed up and slow down, all of that requires fuel. Even using what are known as "low energy" trajectories that use a bunch of gravitational slingshots need velocity changes to adjust trajectories to get the right angle of approach to some large body.

Most commercial launch services focus on LEO because the secondary stages don't need to carry much fuel and their engines only need to fire reliably once to get into a stable orbit.

A planetary transfer requires multiple engine firings. That's much harder than a single reliable firing. It's not impossible but the testing and engineering is expensive.

A science mission often needs to beg, borrow, and steal to get built. There's one shot to accomplish the mission objectives. That's more expensive engineering and testing. Some mission profiles require specific planetary alignments with very narrow launch windows to boot. So there's a lot of expensive work made more expensive by a tight deadline.

The actual marginal cost of the rocket, fuel, and launch is a very small part of the mission cost. Halving the launch price doesn't necessarily affect the cost of anything else.

Starlink satellites are effectively disposable. If a Falcon 9 explodes or fails to get the second stage to orbit they're out maybe a few tens of millions of dollars. If a science mission has some mission-killing failure there's no funding for a do-over.

> But Linux took over the world and it had a crappy UI.

Linux took over in all the places where a good GUI didn't matter. In ran most of the Unix server software on hardware orders of magnitude cheaper than what was needed to run a commercial Unix. Windows ran on that same cheap hardware but was much more expensive for the same capability.

Despite Sun's marketing push Linux was really the dot in most of the dotcom darlings.

AFAIK this was the case. I played a lot of Stars! (almost exclusively the shareware) and it had all the races and tech trees. It showed higher level tech trees but those were only available in the registered game. Map sizes were also limited in the shareware. Later versions might have required additional resources for the retail upgrade but I believe the version I had (2.5?) could become the full registered copy with just a key.

They'll change their name to Meta to get away from a scandal. I don't think that's ever happened before.

Which results in a plainly broken torrent. I've yet to find a torrent client that can effectively download IA's broken torrents. Of course they're not all broken. So you don't get to know it's broken until you're bouncing back and forth between 98 and 99% for a week.

Good thing that won't ever happen again!

They seem to be aggressively blocking IPv6 source IPs. I ran into this problem over the past month traveling. I got nothing but 429 errors until I switched on my VPN (which is IPv4 only) and magically the Wayback machine worked again. The lack of transparency on the part of IA is very frustrating.

The Internet Archive's torrents are a sick joke. I've yet to find one that actually manages to complete. They always get stuck at 90-something percent but that final blocks always fail verification and get retried, fail, and the process repeats forever. Because they're web seeds they're hitting IA infrastructure and not offloading to a real swarm. So their broken torrents are just screwing themselves.

But they're not compute starved. NVidia's kit is convenient for compute but not absolutely necessary. Chinese labs have been training with Huawei chips already. They also have a huge supply of existing Nvidia gear to use. Just because it's not the absolute latest generation isn't really important.

Also as we have seen, the Chinese labs will just buy their kit through cut outs or do training in locations where there's no embargo.

Do you not like your dogs or something? I wouldn't give Arby's to someone I hated.

> I have no idea why we can't do this now.

Power density and all of the problems that are fallout from high power density. A modern LiPoly battery contains a lot of energy. It's stored in a chemical that is highly reactive with oxygen, which you happen to be surrounded by at the moment.

Watch some videos on lithium battery fires. Not only do they have extremely hot flames but they spray out molten bits of battery and can't be extinguished with water.

To make a LiPoly battery that can be safely handled by normal people it needs a strong outer casing. A sealed battery in a phone uses the phone's case itself as the battery's protective casing. A user replaceable battery would need its own strong case. The phone then needs to be larger to accommodate the extra volume or have a smaller battery which stores less power than a sealed battery of equivalent volume.

The market has rewarded thinner phones with long lifetimes but also high performance. That triple point is challenging if not impossible to hit without sealed batteries.

The Switch is essentially that. Obviously not $60 but it's a several generations old Nvidia Tegra SoC. The AppleTV wasn't far off either before its massive price hikes.

The Pro, Max, and Ultra chips boil down to different combinations of their "Super", performance, and efficiency cores along with the number of GPU cores. The Max is two Pros bolted together. Thus far they've skipped generations of Ultra chips but Pro and Max versions of all the M-series chips have been released.

I can see them not releasing an M6 Ultra but no reason not to see Pro and Max versions of the M6.

Apple has not in the past three decades really courted the capital E Enterprise market. They'll definitely sell to Enterprise customers and have Enterprise sales teams for big customers. But they're not and never have been Dell or HP.

Enterprise sales sucks. There's infinite amounts of politicking and glad handing and buyers will get all sorts of sweet brib..."sales dinners" then go with the cheapest option. Margins on hardware sucks and the only money is in support contracts. Apple instead invests in consumer sales/support primarily and all the other channels are side businesses.

Stuff like the Xserve existed mostly for Apple internal purposes and ended up being sold externally to goose the scale enough to make them not a huge loss. At one point a large percentage of the offices on Bubb road were packed with Xserves running portions of the iTunes Music Store and the Apple online store. More offices were packed with Xserves doing media ingest and encoding for iTMS. Just about every building had racks of them as build and file servers.

> which in most cases favours the US

Used to favor the US. Trump's policies and behaviors are ruining the US' reputation as a stable trading partner by demonstrating what damage can be done by an idiot with unchecked power. The rest of the world is going to permanently veer away from US services because it's not impossible (and sadly likely) another feckless imbecile will get put in charge.

The M5 GPU added better matmul and dot product support in hardware (IIRC) that really boosted performance of those kernels over previous GPUs. Current transformer models rely heavily on matmuls. Prefill (processing the current context) is mostly limited by raw GPU processing power where token generation (predicting the next token from the current context) is mostly limited by memory bandwidth.

While an older M-series might run a matmul or dot product kernel just fine the M5 can run them much faster.

That sufficient supply is maybe on the distant horizon. In the meantime everyone is being fleeced because the scalpers have the only supply available. Even if there's a supply glut tomorrow the scalpers have the infrastructure in place to buy out retail channels keeping prices high.

Those Windows laptops always have a dealbreaker for me (besides Windows itself). If they have decent screens, sound, and battery life then the trackpad or keyboard will suck. If it's cheap it'll creak and groan like a Tesla interior...and have a terrible trackpad.

I for one can't wait. Current prices are absolutely insane.

However a problem exists with scalper bots. They're going to fight tooth and nail to keep prices artificially high even when supply increases. Same with the hyperscalers blowing "free money" at eating supply as a strategic position against smaller competitors.

This has been my bugbear for years. Even in places with good cell service on average there's a hundred individual places that have terrible service. Also a good signal to the handset doesn't necessarily mean good actual service. It doesn't even require traveling, just normal daily movements to get wildly variable network performance.

It's infuriating when it's obvious that the developer of an app only ever tested it in a simulator on their dev machine on their super fast WiFi. It never seems to connect with those people developing a mobile app (or web app) that the "mobile" part has a meaning more than just on a handheld device.

YaCy[0] is a P2P search engine. There's public instances but you can self-host it as well. It includes a spider to do your own crawling. It's challenging now between CloudFlare and other anti-bot CAPCHAS crawling effectively is really challenging. Bots (AI and otherwise) have ruined the web.

[0] https://yacy.net/index.html

Well we're rapidly approaching a world going back to 256GB (or smaller) SSDs because VC money is being used to buy up all semiconductor and HDD output by AI hyperscalers. Everything will end up cloud storage, you'll own nothing and like it.

Also the minimum Word document is far larger than 2KB. IIRC it's about 10x that size for an empty Word document (docx).

Parameter size and total number of parameters so ultimately the total size of the model in memory.

This leads to some interesting optimizations. You can quantize all the parameters (or certain layers) of a model and halve or quarter the memory requirement but maintain most of the model's intelligence. This increases the token rate inversely with the size reduction.

Popular quantizations for local models are 8-bit and 4-bit parameter sizes. The Blackwell series of nVidia chips now even support native FP4 math making 4-bit quantizations even faster.