Linked List: January 27, 2025

Monday, 27 January 2025

Shakeup in the App Store Top Downloads List ★

MG Siegler, writing at Spyglass:

Anyway, this matters because it doesn’t mean that DeepSeek, an app based off of the Chinese-developed AI model of the same name, is the most popular app ever despite its current place atop the charts. Your mom probably isn’t downloading it. Just as last week, your mom probably wasn’t downloading Xiaohongshu (‘Little Red Book’), incidentally another Chinese app that rose to the top of the charts. But it is still interesting because again, the mainstays have in recent years dominated these charts. Sure, new entrants would rise (and fall) from time-to-time but it was almost always some order of: Facebook, Instagram, WhatsApp, Threads, TikTok, CapCut, YouTube, Gmail, Google Maps, etc. Right now, there is only a single app from Meta (Threads) and one from Google (Google) in the top 10.

Instead, we have the aforementioned DeepSeek, ChatGPT, Paramount+, Temu, Fox Sports, Bluesky, ReelShort, and VPN rounding out the top 10. It’s pretty easy to see why each of these has risen: interest in AI, interest in sports around the NFL playoffs, interest in Twitter-like social networks given the Xitter situation, and yes, the TikTok ban.

The specific reasons are varied, but altogether it really is true that, for the moment at least, there’s a serious shakeup atop the App Store top free downloads list. That’s interesting for several reasons.

DeepSeek is at #1 and ChatGPT #2. That’s great for “AI” but a bad look for OpenAI at the moment. I can’t help but think that DeepSeek having rocketed to the top spot, seemingly (but not actually) out of nowhere helped drive today’s stock market reaction. (Google Gemini is down at #14.)

Secondarily, and perhaps counterintuitively, it showcases Apple’s strength in AI. Sure, Apple’s own Apple Intelligence is years behind and pretty embarrassing right now, even with its much ballyhooed partnership with ChatGPT. But the iPhone is where people actually use AI and the App Store is how they get the apps they use. To borrow Ben Thompson’s framing, the hype over DeepSeek taking the top spot in the App Store reinforces Apple’s role as an aggregator of AI. The measuring stick for consumer AI products and public social media networks is where they’re listed on the App Store. Not app stores, lowercase. The App Store. Apple’s. That’s The Show, the Major Leagues. So sure, it’d be better for Apple users and Apple shareholders, and thus better for Apple itself, if Apple’s own AI was at least sort of kind of competitive, rather than the butt of jokes. Aside from the easily forgotten iTunes Ping 15 years ago, Apple hasn’t even tried to offer its own public broadcast-style social media platform. But the iPhone is the place where social media networks are used and ranked. The App Store today is like the cable company of yore. It didn’t matter if Comcast’s own channels were the most popular — so long as everyone was watching channels through TVs connected to Comcast TV service, Comcast was getting their cut.

Lastly, when it comes to Meta: Threads is #3 (the people arguing that Threads is a bust sure are quiet lately), Instagram is down to #16 (not great if they were hoping for a bounce from TikTok’s questionable future), WhatsApp #17 (not bad on iOS in the US, where iMessage is strongest) and Facebook is ... scroll, scroll, scroll ... #32, just behind Walmart and Microsoft Teams. 32 isn’t that far ahead of X, which clocks in at #41. Facebook might never die, but we seem to be watching it fade away.

CNBC: ‘Nvidia Drops 17 Percent as China’s Cheaper AI Model DeepSeek Sparks Global Tech Sell-Off’ ★

Jenni Reid and Alex Harring, reporting for CNBC:

Nvidia and other U.S. technology firms plunged on Monday, part of a global sell-off as Chinese startup DeepSeek sparked concerns over competitiveness in artificial intelligence and America’s leadership in the sector.

Nvidia, the chip designer who has been a major beneficiary of the AI hype, last slid 17.3%. With that, the megacap tech stock was on track to notch its worst day since March 2020.

To make a long story short, it’s been broadly assumed that leading-edge model training and inference requires the highest of high-end hardware — which hardware, especially for training, pretty much exclusively comes from Nvidia. The assumption was that any company trying to compete in this fiercely competitive field — a field with massive implications for industry, culture, and national security — must get in line to buy (or rent) enormous amounts of computational power exclusively available on the very best systems from Nvidia. That hardware isn’t available (at least legally) to Chinese companies, because of the Biden administration’s export bans. DeepSeek engineers found and implemented multiple massive efficiency improvements that allowed the company to train its latest models at far lower prices using far-less-capable hardware, and perform inference under heretofore unprecedented memory constraints. DeepSeek’s achievements render obsolete several prior planks of conventional wisdom regarding the state of AI. My link a few sentences prior in this paragraph is to a WSJ report from 14 days ago that now feels like it dates from at least 14 months ago.

The most shocking result has been a 17 percent hit to Nvidia’s stock price today, knocking somewhere around $350 billion off their market cap. Google is down about 4 percent, and Microsoft down 2 percent — but Meta is up 1 percent and Apple is up a little over 3 percent. I think the market is reacting pretty sensibly:

Nvidia has taken an unprecedented beating. $350 billion is more than the combined market caps of Verizon and AT&T (each worth roughly $170 billion). But Nvidia has been — and even with these breakthroughs from DeepSeek, remains — in an unprecedented position. Nvidia’s high-flying valuation reflects its unprecedented position as the essential hardware company for AI. Nvidia’s position, post-R1, remains essential — just not as essential as everyone thought. Hence the haircut. But Nvidia remains the third-most-valuable company in the world — it took a massive plunge today but in no way collapsed.
TSMC and Broadcom both took big hits today too (-13% and -17%, respectively). Being a “chip company” doesn’t look as sweet today as it did last week.
If OpenAI were publicly traded, I suspect it might have collapsed — the sell-off it would have faced today might have triggered emergency measures that halted trading.
Microsoft has been distancing itself from OpenAI, but is still intertwined with them like no other company. (Recall that Apple walked away from a big investment in OpenAI just four months ago.) A -2% hit feels about right.
The broadest implication of DeepSeek’s achievements is that really good AI is going to be even cheaper and more openly available than expected — sooner than expected. That’s bad news for Google, whose entire enterprise value is based around their having an unassailable and sustainable lead in these areas. Google doesn’t have to worry about a competitor coming along to rival them in search. They have to worry that the entire field of “search” is on the cusp of being commoditized. Thus the -4% hit.
Meta, the thinking goes, benefits as generative AI becomes cheaper. That’s why Meta’s own AI efforts are sorta-kinda open source. Meta has nothing to do with DeepSeek, but DeepSeek’s achievements seem perfectly aligned with Meta’s own interests in AI.
Apple benefits, indirectly, in at least two ways from DeepSeek’s breakthroughs. First, a vast increase in the inference capabilities of RAM-constrained hardware strongly suggests that Apple’s consumer devices will soon be able to perform far more AI computation locally. Apple Silicon’s shared memory architecture is perfect for this — and still un-replicated by anyone else in the industry. Second, Apple’s decades-long standoffish relationship with Nvidia suddenly looks like less of a problem.

Ben Thompson’s DeepSeek FAQ ★

Chinese AI lab DeepSeek made waves last week when they dropped their new “open reasoning” LLM named R1. But over the weekend the full repercussions of their achievements (plural) began to sink in across the industry. My partner-in-Dithering Ben Thompson has written an extraordinarily helpful FAQ-style explanation at Stratechery today. If you, like me, were looking at the news today and thinking “Jeebus what the hell is going on with this DeepSeek thing?”, read Thompson’s piece first. Two choice excerpts, first regarding OpenAI:

R1 is notable, however, because o1 stood alone as the only reasoning model on the market, and the clearest sign that OpenAI was the market leader.

R1 undoes the o1 mythology in a couple of important ways. First, there is the fact that it exists. OpenAI does not have some sort of special sauce that can’t be replicated. Second, R1 — like all of DeepSeek’s models — has open weights (the problem with saying “open source” is that we don’t have the data that went into creating it). This means that instead of paying OpenAI to get reasoning, you can run R1 on the server of your choice, or even locally, at dramatically lower cost.

Second, regarding DeepSeek’s use of distillation (using existing LLMs to train new smaller ones):

Here again it seems plausible that DeepSeek benefited from distillation, particularly in terms of training R1. That, though, is itself an important takeaway: we have a situation where AI models are teaching AI models, and where AI models are teaching themselves. We are watching the assembly of an AI takeoff scenario in realtime.

Apple Removes ‘2024’ Timeframe From Next-Generation CarPlay Page ★

Joe Rossignol, writing at MacRumors on Thursday:

Apple’s website said the first vehicle models with support for next-generation CarPlay would “arrive in 2024,” but that did not happen. A little more than three weeks into 2025, Apple has now updated its website in the U.S. to remove that 2024 timeframe from the next-generation CarPlay section of its overall CarPlay page.

It’s good that they updated this page last week, because it really was starting to look unlikely they’d hit their 2024 ship date.

Meh.com ★

10 years ago I ran a sponsorship from Meh, a then-new daily deals site from the founders of Woot (the OG daily deals site). In my thank-you post to Meh a decade ago, “I’ve been selling weekly DF RSS feed sponsorships since 2007 — just a hair under 400 consecutive weeks. I’ve never had one quite like this week’s.” I’m now up to 17 years of weekly DF sponsors, 887 weeks and counting, and while Meh has sponsored multiple times in the ensuing years, it remains true that there’s never been another DF sponsor quite like (or maybe even at all like) Meh.

For their sponsored entry in last week’s feed, Meh co-founder Dave Rutledge decided to re-run the exact same post they ran the first time 10 years ago. Here it is:

Fucking Amazon

I sold Woot to Amazon and they made it shitty. So I quit. Then I got bored.

I started A Mediocre Corporation with a few others from Woot. We just launched a classic daily deal site — only one thing for sale each day. Meh.

Oh, and since you seem to be into RSS, we put one together just for you, at meh.com/deals.rss. Of course you can also just go to meh.com.

When this ran 10 years ago, the headline — “Fucking Amazon” was so unusual that a bunch of readers were concerned that my site had been hacked. Nope, not hacked, just sponsored by Meh. I thank Meh for sponsoring DF last week, thank them for a decade of ongoing support, and I thank them for making me laugh.

Also: the daily deals at Meh really are fun. I follow on RSS — you’ll be amused every morning, and you’ll find a surprising number of things you want to buy.