Criminals are hacking into public Wi-Fi devices—at hotels, conference centers, and so on—around the world and changing their DNS settings. The goal is to redirect users to fake login pages and steal their credentials.

Nvidia is finding new ways for its customers to raise money, and it's expanding the risk of the AI buildout significantly.

Listen to this Update in your podcast player

On January 1, 1870, Jay Cooke, hailed as an American hero for his role in financing the Union effort in the Civil War, signed a contract that would, if you squint, lead to world war.

In 1864, Congress had created the Northern Pacific Railway Company with the goal of linking the Great Lakes and Puget Sound with tracks that would eventually run from Duluth to Tacoma; the charter included 40 million acres of land adjacent to the proposed line in exchange for accomplishing the build-out. For the ensuing six years, however, Northern Pacific struggled to secure financing, even as the Union Pacific and Central Pacific railroads built towards each other, driving the golden spike linking Sacramento and Omaha in May 1869.

Northern Pacific had approached Cooke about funding in 1866, but lacked the generous federal guarantees that undergirded Union Pacific and Central Pacific (which, it should be noted, led to an incredible amount of graft); Cooke, himself no stranger to the financial power of the federal government, wasn’t interested. Ultimately, however, Northern Pacific gave him an offer he couldn’t resist: a commission of 12 percent on every bond, and $200 of Northern Pacific stock for every $1,000 in bonds he sold.

Cooke soon found that his institutional peers agreed with his earlier refusal, and weren’t interested in his bonds, so he leaned on the same tactics he honed selling war bonds: appeals to patriotism, control of the media, and promises of railroad fortunes, backed by industrial-scale distribution. At the peak Cooke employed 1,500 salespeople and funded 1,300 newspapers (through a combination of advertising and direct payments) with a brand burnished by the Civil War. Retail investors could already buy railway bonds; Cooke made them his primary funding mechanism.

This was, to be certain, an incredible innovation. It used to be the case that if you couldn’t get loans from the government or from banks, you couldn’t get much money at all. The problem was that Northern Pacific’s capital needs were endless, and by September 1873, as credit tightened worldwide thanks to a crash on the Vienna stock exchange and the demonetization of silver, Cooke, who had been funding Northern Pacific from deposits in between bond issuances, could find no more buyers. The subsequent bankruptcy of Jay Cooke & Company triggered the Panic of 1873, culminating in endless railroad bankruptcies across the country, a multi-year depression, multi-decade deflation, and, one could argue, the financial conditions that made Europe, four decades later, into a tinder box.

Northern Pacific did eventually finish their line, by the way, with multiple bankruptcies along the way; ultimately, they were one of four railroads that were merged to form the Burlington Northern Railroad. Burlington Northern would eventually merge with the Atchison, Topeka and Santa Fe Railway to form BNSF Railway; Berkshire Hathaway would purchase the parent corporation in 2009.

Blowing Through Debt

If this story sounds vaguely familiar it might be because Cooke is — for obvious reasons — a central character in Liaquat Ahamed’s new book, 1873, released earlier this year. Ahamed is not shy about drawing a link between the collapse of the railroad buildout and the current AI moment; the book’s very first page — even before page 1 — is about translating sums of money, and concludes thusly:

In order to grasp the true significance of sums of money that relate to the economic situation of whole countries — such as the size of the indemnity imposed on France after the Franco-Prussian war — it is most useful not simply to make allowances for changes in the cost of living but instead to adjust for changes in the size of economies. To translate such figures into comparable 2026 magnitudes, multiply by a factor of 1,200. Thus the $500 million that went into U.S. railway bonds annually during the boom years of the early 1870s would today be the equivalent of $600 billion, roughly what is projected to be invested by major tech companies in 2026.

Microsoft CEO Satya Nadella is certainly aware of the connection: he cited 1873 as “the book to be read” on the company’s recent earnings call. Perhaps it’s not a coincidence, then, that Microsoft, alone amongst the hyperscalers, still boasts substantial free cash flow — $19.6 billion last quarter. Microsoft is the one hyperscaler still abiding by the dictum used to deny the existence of a bubble: its CapEx isn’t funded by debt.

This was, believe it or not, a defense that could be used for nearly all of Big Tech a year ago; then, between September and November, Oracle, Meta, Alphabet, and Amazon issued a combined $80 billion in debt for building out infrastructure. That was only the beginning: after raising a combined $108 billion in all of 2025, these four companies have, as of July 7, already raised $194 billion this year. Unsurprisingly, spreads are rising, and 86% of the bonds issued this year are already trading at higher yields than at issuance. Cover for recent issuance has fallen to less than 2x, from 5x in February.

The real shock, however, came at the beginning of June, when Google announced it would raise $85 billion in equity, including a special $10 billion issuance to the aforementioned Berkshire Hathaway. I wrote at the time in The Google Capital Company:

It is worth noting that $10 billion is a relatively small amount of money to both companies. To that end, perhaps the primary utility is as a signaling mechanism. On Google’s side, the signal is that the expected demand is actually far greater than anyone thinks, and that the company is ready and willing to fund supply using all means at its disposal, including equity; for them Berkshire Hathaway’s investment is an endorsement of this view and a validation of the wisdom of the investment. And, on the flip side, if the signal is correct, then Berkshire Hathaway is getting a deal and putting its cash flow machines to work building the future.

I concluded:

Implicit in this analysis was that there was enough compute capacity in the world to be bought; what happens, however, when and if there isn’t? What if the ultimate battle — the one that determines who gets compute — becomes a matter of who can bring the most cash to bear? And what if that advantage compounds, such that the company with the most cash capacity ends up with the most compute capacity (which we already know they will sell, in addition to using themselves) driving the ability to generate more cash? In that world, what company would be your best bet?

The implied answer, of course, was Google.

DeepMind Drama

Google right now is no one’s bet, at least in terms of the frontier. After the departure of DeepMind CEO Demis Hassabis (technically promoted to chairman, but no longer in charge of day-to-day operations) and Gemini co-lead and former Chief Scientist Jeff Dean, along with a host of other prominent researchers, SemiAnalysis declared that Gemini is Cooked:

For all intents and purposes, we believe DeepMind is no longer a frontier lab. We said as much a few months ago to our Tokenomics clients due to large numbers of departures from their reinforcement learning teams and poor compute allocation. Google will continue meandering on and releasing models, but their odds of reaching SOTA again have dropped to zero.

Furthermore, the biggest beneficiary of today’s news is neither Anthropic nor OpenAI—it’s Google Cloud. Whereas Gemini and GCP used to desperately fight for compute allocation, it’s now clear that Thomas Kurian won. We expect GCP revenue growth to meaningfully accelerate as a result.

From later in the post:

We’ve obviously been quite bearish on DeepMind thus far, and if we had to steelman the case for why they’ll still be able to train a true SOTA model in the future, it would go something like the following:

• The current setup clearly wasn’t working. With the existing leadership team, their odds of catching up to Anthropic/OpenAI looked extremely slim.
• Now that they’ve cleaned house, the new guys can start from a blank slate. Maybe they’ll even acqui-hire a neolab like SSI or Thinking Machines.
• With this new team, their odds of catching up to the frontier actually increase.

Perhaps there’s some world in which this happens, but we think the odds are basically zero. The issue with Google was not Jeff Dean nor Noam Shazeer, but rather their extremely bureaucratic, painfully slow, and strategically timid culture. Remember that DeepMind had an AI chatbot 1 year before ChatGPT but was not allowed to release it due to fears of disrupting their core business.

Actually, you could make the case the problem was also Hassabis and DeepMind. I explained in an Update after Google I/O how Hassabis’ vision of the frontier was fundamentally different from the other frontier labs because he believed in world models, not just text/code, and concluded:

What falls out of [Hassabis’ vision] are models with multimodality — in contrast to Claude, which outputs text only — and, it must be said, not nearly as impressive coding capabilities. This gets at the point of this entire digression: I think it’s possible that the reason Google is widely considered to be behind both Anthropic and OpenAI in terms of coding, particularly long-running agentic workflows that depend just as much on the harness as the model itself, simply comes down to their research team having other priorities. That’s why the coding parts of this keynote fell on the Antigravity team, not DeepMind, and why Hassabis was barely on stage.

From this perspective, last week’s events are less surprising, and were arguably foretold at I/O: Hassabis might be right about world models being the path to AGI, but Google has run out of patience in terms of letting him find out; Google co-founder Sergey Brin is reportedly deeply involved and closely allied with Koray Kavukcuoglu, the new DeepMind CEO, and I wouldn’t be surprised if the company is pivoting to Anthropic’s more text- (and thus code-) centered approach.

Google’s Infrastructure Bet

What is fascinating about Google’s position is that these machinations do not necessarily mean the Berkshire Hathaway bet was a bad one; indeed, it’s arguably good news. This is what the SemiAnalysis article was driving towards, and it’s a point I made last week about Google’s recent earnings:

The story seems to be very similar to last quarter, with even more Google Cloud growth: 82% year-over-year (compared to 63% last quarter, and 32% a year ago), with 36% margins (compared to 33% last quarter, and 21% a year ago). I wondered then how much of this growth was actually Anthropic, and while we didn’t get clear confirmation this quarter, I thought this answer from CEO Sundar Pichai on the earnings call about why Google needs to rent 3rd-party capacity was notable:

I think on the bridge deal, the main thing I would say is, look, there are — on the margin, there are very, very large customers of ours on Cloud who we are trying to support them through this extraordinary moment. And the incremental opportunities they are bringing to us, while a short‑term cost over a few months may be very high, in the lifetime of the deal, as we bring more capacity on, is highly ROI‑positive. So those are factors we are taking into account. So are you willing to take upfront a six‑month deal to be able to serve the customer in what is a multiyear opportunity where the margins and the returns are very, very attractive over that multiyear horizon? So hopefully that gives some color on how we’ve thought about those opportunities.

That customer is almost certainly Anthropic.

Again from SemiAnalysis:

More than 20% of total TPU shipments from 3Q26 to 4Q27 are being sold directly to Anthropic. This is excluding the hundreds of thousands of TPUs GCP already rents to Anthropic today, and the many hundreds of thousands more they’ve committed to rent to Anthropic and Meta over the next 6 quarters…

If you’ve ever listened to an interview of Google Cloud CEO Thomas Kurian, you know he is not AGI pilled. In one podcast, for example, he argued that it’s great for TPUs to become “general purpose infrastructure” that supports customers like Citadel, the Department of Energy, and generic high performance computing. And when asked why he was selling compute to Anthropic despite them competing with Gemini, he said this was the natural consequence of Google being a “platform company.”

Kurian said the same thing to me in a Stratechery Interview:

We sell different parts of our stack. One of the things people don’t realize is we monetize many different parts of the stack in different ways. Like Anthropic, there’s a lot of labs that use our stack — in fact, most of the large AI labs use our stack. So if somebody uses TPUs to either to train their model or to use it for inference, we’re monetizing that part of the stack, that gives us resources to then fund our R&D and other investments. Some of the labs use our TPU and our Gemini model, others may use our TPU and then buy our cybersecurity protection for their models. So as a platform player, we have to allow our technology to be monetized in as many ways as possible and we don’t see it as a zero sum.

We’ll see how zero sum compute actually is — there are reports Google’s researchers have been starved for compute — but the overall takeaway is that whether or not Google is competing for the frontier, they are absolutely competing to dominate AI infrastructure. And, in a world where intelligence is a commodity, TPUs in particular are a big deal.

Last month, in Who’s Afraid of Chinese Models?, I talked about commodity markets in the context of frontier labs versus everyone else; in commodity markets marginal costs are determinative of not just profitability but also viability, and I made the case that the frontier labs are well-positioned to have superior cost structures for any given unit of intelligence.

That cost structure, at least for now, includes the cost of renting compute, and it seems likely that TPUs are cheaper than Nvidia GPUs; Anthropic may have built for TPUs (and Amazon’s Trainium chips) because only Google and Amazon had the wherewithal to fund them, but at this point that ability may very well be a significant advantage. The fact that Anthropic is straight up buying TPUs for its own data centers (converting compute costs from marginal costs to capital costs) suggests that is the case.

What is notable is how amenable Google is to share, even at the price of needing to issue equity. This, however, fits the Berkshire Hathaway model that I wrote about in The Google Capital Company:

One of the businesses Berkshire Hathaway used the See’s profits for was on the opposite end of the spectrum in terms of capital utilization: BNSF Railway. Railways require a lot of capital to operate; BNSF consumed $3.8 billion last year; they also make a lot of money: BNSF’s net income was $5.5 billion on revenue of $23.4 billion. To put that in perspective, the total amount that Berkshire Hathaway has made from See’s Candies is probably less than $3 billion (the last disclosure was “over $2 billion” in 2019), i.e. less than BNSF made last year…

In fact, you can make the case that Abel is actually just replaying Buffett’s strategy, only this time Berkshire Hathaway is See’s Candies, and Google is BNSF. At the end of last quarter Berkshire Hathaway had $373 billion in cash, and $25 billion in free cash flow in 2025. How many companies could actually employ that cash in a way that generated a high rate of return?

It’s hard to imagine a better option than Google. The company is not only investing in AI, but has optionality in terms of outcomes: its Services business benefits from the investment, it is in contention at the model layer with Gemini, and it can sell capacity to the frontier labs. Moreover, that capacity has a sustainable cost advantage because of TPUs, which means that in a world where compute becomes a commodity — as hard as that is to imagine right now — Google is the hyperscaler that is poised to make the most profit.

Notice that I didn’t say margin; if that were Google’s concern they would almost certainly be making different choices. Profit, however, is an absolute number, and Google is bringing everything to bear — first its cash flow, then its debt, and now its equity — on making money from the infrastructure build-out.

Nvidia’s Investable Asset Class

Today corporate executives and financial engineers don’t need to control newspapers; thanks to his new X account, Nvidia CEO Jensen Huang can go straight to the public. From an X Article posted last night:

NVIDIA AI Factory Compute Is Becoming an Investable Asset Class

Today, we announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent financing platforms designed to mobilize over $500 billion of third-party capital to support the buildout of AI infrastructure over time.

This is a major milestone for NVIDIA and the AI industry. We have moved from an era in which companies bought chips and built data centers project by project to one in which AI factories can be financed as productive infrastructure — with repeatable platforms, long-term institutional capital and a diverse customer base that uses compute to create revenue.

AI has reached an inflection point. It is moving from research into production. AI is creating real value, and the infrastructure behind it is becoming one of the world’s most productive assets. In AI, compute is revenue.

Huang argues that Nvidia-based AI factories are fungible, protecting residual value, and that CUDA makes AI factories better over time, extending their economic value; according to Huang:

These are the characteristics of an investable infrastructure asset: it produces revenue, serves a broad market, improves in performance over time and can be redeployed.

Thus the attempted formalization of a new investment structure:

The demand for AI infrastructure is extraordinary. But access to capital is uneven. Many great AI companies, enterprises and AI clouds have demand for compute but do not yet have access to financing at the scale or cost required to build quickly. That is why we are partnering with the world’s leading long-term capital providers.

Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR are also among the world’s leading infrastructure investors, with deep expertise in underwriting long-lived, productive assets. Together, we are creating repeatable financing platforms to help the AI ecosystem build the factories it needs.

What Apollo et al. are, are new sources of capital beyond the investment grade debt markets. In that sense this proposed structure is somewhat akin to Google’s equity issuance: a way to secure funding beyond bonds. The difference, however, is stark: whereas equity dilutes the upside for investors without adding risk to the company, this structure preserves Nvidia’s margins by finding new pools of capital willing to bear risk.

It’s not a total free ride for Nvidia: the company is backstopping opportunities with up to 25% residual-value based financing, suggesting that Huang believes his “investable asset class” pitch much more than the market does. That is, in a certain sense, a price cut, as the goal is to reduce the cost of capital for entities building data centers with Nvidia chips, by putting Nvidia’s profits on the line for uncertain investments. That guarantee is downstream from Google’s (and soon Amazon’s) aggressiveness: why build a data center with Nvidia chips if you can buy TPUs or Trainiums (Nvidia chips are likely better, but if the constraint on new data centers is capital, lower up-front prices may matter more than token efficiency).

Nvidia’s bigger problem is one that has been apparent for a long time; I wrote back in 2024:

In the before-times, i.e. before the release of ChatGPT, Nvidia was building quite the (free) software moat around its GPUs; the challenge is that it wasn’t entirely clear who was going to use all of that software. Today, meanwhile, the use cases for those GPUs is very clear, and those use cases are happening at a much higher level than CUDA frameworks (i.e. on top of models); that, combined with the massive incentives towards finding cheaper alternatives to Nvidia, means both the pressure to and the possibility of escaping CUDA is higher than it has ever been (even if it is still distant for lower level work, particularly when it comes to training).

The situation today, with Anthropic and OpenAI appearing to pull away, is even more problematic: Anthropic has not been dependent on CUDA for years, and OpenAI is moving in that direction, at least for inference. If those companies win then Nvidia’s profits will be squeezed — indeed, the implication of that backstop is they already are (this, needless to say, is why Huang’s first post was an open letter in defense of open models).

Risky Business

This might not cost Nvidia anything in the end: if AI revenues truly take off, then the debt markets will open back up, and ultimately companies will go back to funding infrastructure investment through free cash flows. Right now, however, is the danger zone, as hyperscalers blow through the debt markets and Google at least starts to tap equity. To the extent Nvidia competes through novel funding mechanisms that, at the end of the day, draw on things like insurance floats and pension funds and other long-run liabilities that are the bread and butter of the asset managers the company is partnering with, the risk — unmarked, unlike equity — is considerably higher.

That’s why I started with 1870 and Cooke’s ill-fated agreement with Northern Pacific. Yes, the upside the deal afforded Cooke was incredible, but it was incredible for a reason: it was very risky, and pioneering new funding mechanisms only served to spread the pain when it all blew up. It’s one thing to spend all of your free cash flow; it’s another thing to tap the debt markets. And, beyond that, it’s a completely new nerve-racking thing to bring safety-seeking assets to bear. AI better deliver before it’s too late.

Add to your podcast player: Stratechery | Sharp Tech | Dithering | Sharp China | GOAT | Asianometry

Following up on my post yesterday about modern iPhones and scratch resistance, and my personal habits of (a) almost never using an iPhone case, and (b) never setting the iPhone face down except on soft (cloth) surfaces.

I should have anticipated this, because I’ve had this conversation with real-world normies repeatedly in recent years, but I got a bunch of questions from people who say they’re wary of ever putting their iPhone down on its back because they’re worried about scratching the camera lenses. That’s not a silly thing to worry about. It looks like those lenses are glass; glass scratches; and it sure seems like a scratched camera lens might forever ruin all the photos and videos you take with the camera.

There are a couple misconceptions here though. First, the glassy flat circles exposed on the back of your phone aren’t the camera lenses. Those are covers over the lenses, which are smaller, spherical (not flat), and recessed. You can look through the covers and see the actual spherical lenses inside. The exposed lens covers are made of sapphire, not glass, and are thus incredibly scratch-resistant. They’re by far the most scratch-resistant parts of the phone. I’ve never once found even a tiny scratch on any of my iPhone camera lens covers, and I’ve never taken any particular care to avoid scratching them. I mean, I don’t drag the lenses face-down on surfaces. I’m not trying to scratch them. But I set my phones down on hard surfaces lenses-down all the time and they never seem to pick up even fine scratches.

Zack “JerryRigEverything” Nelson is the guy on YouTube who makes videos where he scratches the hell out of products to see how durable they are. (And he bends them, burns them, and abuses them in other gruesomely creative ways.) In his iPhone 17 Pro video, starting around the 4:10 mark, he tries scratching the sapphire lens covers with a sharp razor blade. No effect. Sapphire is far more likely to shatter than scratch, because it’s so hard. (You may recall that a decade ago Apple pursued using sapphire for the displays of iPhones but it didn’t work out.) Don’t try scratching your lenses with a diamond hardness pick and you’ll be fine.

Also, believe it or not, a fine scratch on the lens cover will almost certainly not affect image quality at all. Try sticking a strand of hair (which is probably much thicker than a typical scratch) to the surface of your iPhone 1× main camera lens. Take a picture. Clean the hair off the lens. Retake the same picture. You almost certainly won’t see any difference at all. Here’s a great video from DPReview back in 2020 showing just how much dust or scratching you need to put on the outside of a lens to degrade image quality. It’s counterintuitive but the surface of the lens is not where light is focused — the sensor, inside the camera, is.

Dave Nanian:

The new version of SuperDuper is a ground-up rewrite, a tip-to-tail reinvention of what SuperDuper looks like, how it works, what it can do, and where I can take it in the future.

[…]

Documents have been retired and replaced with Copy Jobs, which are stored internally and organized by source. If you want to copy Macintosh HD, that’s where you start. Click, and you’re presented with one or more template copy operations, or you can start from scratch.

Configuration is done right inline by clicking bold, underlined link elements and selecting options from pop-ups. The effect of your choices is immediately reflected in What’s going to happen?, ensuring you know what to expect when you click Copy Now.

SuperDuper has been around for 22 years, and I’ve been using it since version 2 in 2006. Shirt Pocket had always been great about keeping the copy engine up-to-date with all the changes in macOS and its increasingly baroque file system structures. But the outward appearance of the app hadn’t changed much, aside from losing the brushed metal, and neither had the feature set. I know all too well how much work it can be as a solo developer just to keep an app working. Ironically, the better you are at adapting to the changing OS, hiding the complexity, and presenting a simple interface, the more it looks like you aren’t doing much. Actually, if you follow Nanian’s blog, you can see a fraction of all the stuff going on under the hood—tricky edge cases, macOS bugs, and undocumented behaviors that must be handled so that you can essentially just press a button and get a proper bootable backup.

Eventually, he found the time to rewrite the app, both the engine (which is faster and more robust) and the user interface (which is greatly expanded). My first reaction was: this is SwiftUI, isn’t it? And it is. You can just tell by the way the text reflows, all the different colored buttons and blobs, the hover effects, and how the design heavily emphasizes scrolling. (Like System Settings, it doesn’t remember scroll positions. Nor does it remember which disclosure triangles are expanded.) The keyboard navigation and focus behavior are off in places. For reasons I don’t understand, SwiftUI gets a lot of this wrong out of the box. In some cases, like the outline view for creating a copy rule, keyboard support was manually added and works as expected. But there are other places where arrow keys and type-selection don’t work. Sometimes I need extra clicks to get the proper focus. Sometimes the arrow keys scroll the window instead of changing the selection.

That said, this is one of the best SwiftUI interfaces I’ve seen. I think it successfully preserves the spirit of the app, making it really easy to configure things and see what’s going to happen, while streamlining the way documents/jobs are managed and expanding what you can do. It’s still essentially one window (plus a Settings with one checkbox), and you can still use it in the old way, but now you can also run multiple jobs at once and inspect their histories.

Click the Preview button and SuperDuper will run the copy without making changes and present you with a report that shows exactly what would have been updated, added, deleted, etc., had this been a real run.

This is great to have. If there are particular folders or files you want to check, you can browse the source tab of the preview to see what’s excluded or find them in the destination tab to see whether they would be added, updated, removed, or kept. The interface to browse this is a sort of columns view with only one column. It’s fast and easy enough to use, but keyboard navigation doesn’t work properly. I longed for it to be an outline view so that I could quickly expand an entire subtree and see at a glance which files would be changed. I also wished that it would show how much data would be copied.

SuperDuper 4 is much faster than before. And I’m not just talking 20% or 30% faster. Smart Updates are between 2x and, with the new Turbo feature, an order of magnitude faster.

It’s hard to benchmark such things, but my general impression is that this is true. The longstanding Smart Update feature copies only the files that have changed. This seems quite a bit faster than before and slightly faster than Carbon Copy Cloner. The new Turbo feature uses the FSEvents log to optimize finding such files. If macOS knows which folders might contain changes, SuperDuper can focus on those and not waste time scanning ones that are identical. When this works, it feels almost impossibly fast. SSD-to-SSD backups of my Mac’s boot drive will take 1–3 minutes instead of more like 20 minutes. This also seems to be faster than Carbon Copy Cloner’s similar Quick Update feature.

Unfortunately, for reasons seemingly out of the app’s control, sometimes Turbo isn’t possible. This can happen if you’ve modified the copy job or if there’s been a huge amount of file system activity since the last backup so that FSEvents doesn’t contain the necessary information. In such cases, SuperDuper will fall back on doing an old-school Smart Update. I found that excluding my destination drive from Spotlight made it much more likely that Turbo would work.

As much as I like Turbo, I don’t end up using it except for the daily scheduled backup of my boot drive to an always-connected SSD. For my other backups, I have multiple destination drives for each source, and I find that the user interface doesn’t scale so well to configuring a separate job for each. That leads to lots of scrolling. So, instead, I use just one job per source and select the destination I want before starting each backup. This counts as editing the job, so it won’t be able to Turbo, but I find that backups to slow spinning hard drives take a long time either way and are dominated by the copying speed rather than the scanning speed.

Successful and failed runs are shown at the bottom of the copy job, with color-coded dots that can be clicked to display details.

This is my favorite bit of interface design in the new version. It very compactly shows you the recent history for that copy job using the same format that’s used for showing the completion status of the most recent run. I can quickly see all the information I usually care about, and you can click a button to see more detailed information if necessary. The downside is that you can’t see as much information at once—or across multiple jobs—as if there were a dedicated log window, and it only shows the most recent 7 runs. But I think this tradeoff is well in keeping with the app’s spirit of making things simple and clear rather than offering every possible option.

The most baffling part of the interface is that after each backup completes or fails it shows an “Apple Intelligence Summary”. Much space is devoted to labeling this and to assuring you that none of your data went into the cloud. I just don’t see what the point of this is. 99+% of the time, my backup either succeeded or failed for a very predictable reason like running out of space. Why do we need AI for this? I trust the AI summary less than I would a hard-coded string from the developer. Perhaps in an obscure situation that the developer didn’t anticipate Apple Intelligence would be able to offer a useful interpretation, but I have yet to see that.

At a technical level, SuperDuper has been split into a “helper” or “server”, which runs the copy, and an interface, which is the part you interact with. That means that, once registered, you can quit the SuperDuper application and your backups will still run. They’ll even run if you log out!

It doesn’t much matter to me that it can back up without launching the app—I like to see its progress—but I appreciate the new architecture for scheduled copies. Schedules are now much better integrated into the app proper rather than being driven through AppleScript.

You can run any number of copies at the same time. If you have four backups you need to do, just click Copy Now for each, and they’ll all run simultaneously.

This is the most important new feature for me. I used to do all my clone backups with SuperDuper. But, over years, my backup needs grew. I now have five main source drives to back up. They are much larger than before, and APFS on spinning hard drives is really slow. Thus, the backups, even with Smart Update, can take a long time. That was a lot to manage with old versions of SuperDuper. Carbon Copy Cloner has long supported multiple simultaneous backups so that I could connect a bunch of destination drives, tell it to start, and when I came back they’d all be done. I didn’t have to be there to wait for each backup to finish and then start the next.

It was just so much easier, so I started using Carbon Copy Cloner for most of my backups. But I still liked SuperDuper and trusted its copy engine, so I continued to use it to do some of the backups of my most important drive (the internal SSD). I think it’s good backup practice to have redundancies, not just in terms of multiple physical drives and storage locations but also for the software creating the backups. Both are quality apps, but it’s possible that there’s a bug or a misconfiguration on my part that might prevent the copying of a file that I need. Why put all my eggs in one basket? Now that SuperDuper also supports multiple simultaneous backups, I’m using it for half the backups of all my drives. On one backup day I’ll do all SuperDuper backups, then the next time I’ll do all Carbon Copy Cloner backups. (I continue to use Time Machine and Arq for additional redundancy.)

If you try this, I strongly recommend writing down which app you used for each destination drive and then doing all future backups to that drive using that same app. And it works better to do the first SuperDuper backup with Erase and Copy rather than trying to Smart Update a Carbon Copy Cloner backup.

There are a variety of reasons for this, but the main one is the different ways that the apps deal with APFS snapshots. Both use snapshots on the source to ensure they’re copying from a consistent state. And both mark completed copies with snapshots on the destination. But the retention of the destination snapshots is very different.

Carbon Copy Cloner can keep them around essentially as long as you have enough free space. It can also prune them out of order, so that you have daily snapshots for the last month and weekly ones going back further. This is managed by the app, though you can also manually delete snapshots yourself using Disk Utility.

SuperDuper’s snapshots are automatically managed by the macOS. I’m personally less fond of this design because in practice the system will choose to clean up old snapshots almost right away. I can’t count on being able to restore from any backup but the most recent. To me, this is a shame because APFS supports multiple snapshots, and there’s plenty of space on the drive to store them, but I can’t use them. It’s just not a design goal of the app.

Again, this ties back to how SuperDuper values simplicity. If it made longer-lived snapshots that weren’t cleaned up automatically, you might get into a situation where you (or another app) wanted to add files to a drive and couldn’t because it seemed to be full. You’d have to understand snapshots or have an app that could delete the SuperDuper snapshots to free up space. Third-party apps aren’t supposed to delete snapshots belonging to other apps, which is why you don’t want to mix SuperDuper and Carbon Copy Cloner on the same destination. If the former needs more space to complete a backup, it won’t be able to purge snapshots created by the latter.

The status information for the backups that are in progress is shown right in the Sources section of the main window. You can see at a glance which drives are being copied and when they’re all estimated to finish. I haven’t found the estimates very accurate, so I’d prefer to see the start time and the percent complete. Flipping between multiple active jobs to check their detailed progress is kind of a chore, but if all you need is the summary information you can close the app and monitor it from the new menu bar icon.

SuperDuper 4 now supports copying to and from any file system the Mac supports. Using a “Photos” drive with both Windows and Mac that’s formatted as exFAT? No problem: you can copy it.

It can also copy individual folders rather than just entire volumes, if you drag them into the sidebar.

I’ve added Shortcuts support, so you can create a shortcut that runs before or after a copy, on success, or on error. Want to send an email on success or failure? Beep when copies are done?

[…]

Not only does SuperDuper let you call a shortcut from a copy—you can create, run, and check on a copy job from a shortcut!

There’s a lot of potential here, and the completion shortcut receives a bunch of JSON describing what happened so that it can make decisions about what to do. On the other hand, I would rather see AppleScript support, as I find Shortcuts confusing and inconvenient. I wish that common completion actions like sending e-mails were built into the app so that they were easier to set up. (You can choose to eject the drive, post a notification, or sleep the Mac without having to write a shortcut.)

In all the years SuperDuper has been on the market, we’ve never charged for updates and never even raised the price. That’s 22 years of free updates.

The Apple Silicon version was a paid upgrade, but you could run SuperDuper 3 in Rosetta with an old license and still get all the features.

Version 4 costs $43.95, with various upgrade discounts based on the purchase date. There’s no discount if you purchased before July 2020, but us old timers got so many years of free updates it was one of the best deals ever in Mac software. Regardless, it’s not a high price for peace of mind. The new version is, I think, a massive improvement, maintaining the spirit of the original while modernizing and expanding it.

I’ve already said that I use both SuperDuper and Carbon Copy Cloner. I think both are excellent—great apps with great customer support. With SuperDuper 4 they’re more similar than in the past, but they remain very different. If you must pick one, I would say that Carbon Copy Cloner is better if you want lots of options and control; if you want persistent snapshots, verification, copying of cloud-only files, or detailed logging; or if you have a very large number of backups. (It does have a Simple Mode, but it’s arguably too simple and limiting.)

SuperDuper is better if you care about bootable backups or want something simple but not too simple. There’s documentation, but you probably won’t need it because so much is explained within the app’s interface. It’s just really easy to set up a backup, see what it’s going to do, see what it’s currently doing, and see what it did before. Pretty much everything happens in a single window, without tabs or separate modes. SuperDuper is all about striking a good balance for most people, doing the right thing with a minimum of fuss.

See also: Adam Engst and Accidental Tech Podcast.

Previously:

• Parachute Backup Acquired
• Backblaze No Longer Backs Up Dropbox
• Carbon Copy Cloner 7.1.3
• SuperDuper 4.0 Beta
• CrashPlan Consumer Is Back
• ChronoSync 10
• Arq 7
• SuperDuper 3.1 Supports APFS Snapshots for Both Source and Destination
• SuperDuper and APFS
• Synchronize Pro X 4.1.1
• My New Backup Strategy

Joseph Cox, writing for 404 Media (paywalled, sans gift links, alas):

Consulting giant Accenture is trying to figure out how to stop non-technical workers from blowing through companies’ AI token budget on trivial tasks like converting PDFs to presentation slides, according to leaked audio obtained by 404 Media. Across the industry Accenture is seeing “soaring token spend,” according to the audio. [...] It also undercuts the narrative that superpowered engineers generating mountains of code are behind the AI boom. In many cases it is non-technical staff burning through tokens for non-specialized tasks. [...]

At one point in the meeting, Kwak and Eduardo Salamanca de Diego, senior manager of product management at the company’s Center for Advanced AI, start presenting about what is described as “token ops.”

Kwak says he knows people aren’t using slides these days, but he has some. As he appears to be preparing to present, Stuart Henderson, Accenture’s client group lead, interrupts. He jokes he hopes Kwak didn’t just convert a PDF into images and then into Markdown files. “I’m learning that’s one of the big token chewers,” Henderson says. “Turning PDFs into Markdown: is that right?”

That’s when Kwak says that’s what Accenture’s own data shows.

My impression of consulting giants like Accenture (McKinsey, Bain, Deloitte, KPMG...) has always been that they are very good at looking smart but in fact very bad at actually being smart. The idea that they’re spending serious money on AI tokens to turn PDFs and PowerPoint decks into Markdown only makes me think I’ve overestimated the median intelligence of the people who work there. I really thought Markdown couldn’t get more successful but this takes the cake. I love it that my little baby is helping burn these companies’ cash. This is so stupid it hurts to think about.

Markdown is something you start with because it’s easy to write and universally readable. You write Markdown and use that to make slides — like, say, with iA Presenter — not the other way around.

(Via Simon Willison.)

Link: 404media.co/the-tokenpocalypse-is-here-companies-are…

The recent revelations that Jho Low has once again sought to negotiate a pardon from both Malaysia and the United States revives concerns about how this fugitive managed to escape to China with so much of his stolen Malaysian loot apparently still available for him to access.

When his name appeared on a notorious ‘‘250 pardons for 250 years’ list that the US president Donald Trump purportedly mulled signing on July 4th, there were even reports Jho had offered key intermediaries at least a billion dollars to arrange the favour.

Sarawak Report simultaneously learned the fugitive was separately negotiating a similar pardon from Malaysia as part of a settlement designed to draw a veil over China’s damning role in the criminal cover-up of 1MDB.

This is not the first but the third recorded episode of pardon seeking by Jho, who plainly desperately longs to escape from China The previous round in July 2022 was confirmed by the intermediary, ex-AG Apandi Ali, during which RM1.5bn ($370m) was reported to have been offered by Jho’s US lawyers to the Malaysian authorities to wipe his slate clean.

How the fugitive retained such enormous levels of ready cash and where it is being managed and protected is therefore a question of considerable interest to the people of Malaysia to whom the money actually belongs.

Did ‘Swiss Banking Secrecy’ Protect Jho’s Stash?

It is now a decade since the devastating press conference at the Department of Justice in July 2016 which initiated the US action against the Malaysian fraudster.

The then Attorney General, Loretta Lynch, and her colleagues described 1MDB as the world’s largest recorded theft, conducted on behalf of Jho’s political master, the now-jailed former PM Najib Razak, who is still facing a recovery action through the civil court.

Many of the stolen assets were subsequently confiscated and returned, but the evidence suggests  by no means all. Malaysia remains burdened by the crippling costs of the original $5.6 billion dollar theft and then the further theft of similar sums incurred during the later attempts to cover it all up [for the full details of this cover up refer to SR’s upcoming book The China Contract]

However, during the course of this subsequent decade the various investigations, law suits and official disclosures have bit by bit revealed more, not only about how the money was stolen but about how it was laundered and where it ended up.

In particular, details from regulatory actions taken against employees of Rothschild Bank and affiliated Rothschild trust in Switzerland, which came to light in 2023, raise a number of questions that Sarawak Report has now put to the Rothschild group about how much of the stolen wealth they managed on behalf of Jho and the Low family was surrendered to the authorities at the time they were revealed to be fraudsters, as opposed to being quietly moved elsewhere at a later date?

Top Customer 

It first came to light that Jho Low was a client of the Swiss based Rothschild Bank SA (and of its subsidiary Rothschild Trust SA) in 2017, after FBI investigators traced the ownership of Jho Low’s assets in the US to a series of BVI companies placed under a separate New Zealand based trusts which had been incorporated and were being managed by the Rothschild group in Switzerland.

These included his Beverley Hills mansion, New York apartments, his private Bombardier jet, flats and buildings in London and Paris, a super-yacht, company shares in EMI, shares in the New York Park Lane and L’Hermitage hotel groups, incorporated under 23 named companies each held by these separate trusts.

Rothschild dutifully surrendered its control of these contested assets to the US Department of Justice (DOJ) and a major court battle ensued whereby the Low family, together with Jho, sought to remove the bank from its formal control as trustees in order to appoint another management outfit called FFP in the Cayman Islands that was willing to resist the US seizure orders.

In early 2017 the Low family won an initial victory in the New Zealand court which agreed they could appoint the new managers, thereby side-lining Rothschild. Reporting on the litigation Reuters and others put a value on the trusts at stake of $260 million.  These were transferred to the management of FFP which then continued to support the family in their legal battle with the DOJ.

Eventually, in October 2019, the Low family reached a ‘global settlement’ with the DOJ surrendering those assets, including ownership of Jho’s share of the Park Lane Hotel (earlier off-loaded to Greenland Properties in Hong Kong), which the US press release of the time claimed now represented a total value of $700 million.

The US court’s relevant confiscation order named the 23 trusts that were involved in that settlement and the asset each represented. At which point it was widely assumed that Rothschild Bank’s relationship with Jho Low, which had by then been denounced by the Swiss regulator FINMA as a ‘serious breach’ of compliance rules, was over.

Indeed, FINMA had stated at the time that this admonishment of Rothschild, which took place earlier in July 2018, represented the ‘conclusion of its final 1MDB proceedings‘, leading observers to logically conclude that there was no more stolen Malaysia money in the Rothschild banking group – nor indeed any Swiss bank subject to FINMA regulation.

No financial or other penalties were exerted against the bank itself over these blatant institutional failings on the basis that a new monitoring process had been agreed and implemented by FINMA to supervise future activities by Rothschild Swiss and its subsidiary, Rothschild Trust.

The CEO of the group, the British banker Nigel Higgins (who had spent 36 years at the bank) departed shortly after to take up a new position as the Chairman of Barclays in the UK, where he remains in situ.

However, a full four years later and barely remarked by wider media, further proceedings prosecuted by Switzerland’s Federal Department of Finance against a relatively lowly compliance officer and the former CEO of the Rothschild trust management subsidiary revealed some startling new information, which slipped out into the financial press in 2023.

Those proceedings took place in German and were covered primarily only by specialist finance platforms, such as the French language site Gotham City which memorably described the charges laid against the former chief compliance officer as reading “like a horror novel when it comes to banking supervision”.

The officer pled guilty to failings in his reporting obligations with respect to blatantly suspicious dealings and money-laundering, and was fined CHF 40,000, even though it was acknowledged that he was only acting according to the views of his superiors at the bank.

Indeed, the prosecution papers clearly show that the senior personnel of the back had been regularly consulted on the thorny issues surrounding their most lucrative client and consistently decided over the course of more than five years not to report him.

Of these, only the Trust company CEO was prosecuted. He was found guilty and fined CHF 185,000, but has appealed.

Thus it is established that the hierarchy at Rothschild Swiss and its wider group successfully decided to ignore red flags for over five years, despite evidence presented that shows Jho Low failed numerous due diligence investigations and reports by independent consultancies who concluded they were ‘unable to verify many of the business activities described in [Jho’s company’s] annual report’ or indeed ‘any significant professional activity since he left university’.

The same reports also pointed to the dire news coverage that followed the young man’s mysterious ostentation, but Rothschild didn’t flinch. The reason is clear from an email sent by the CEO of the trust company to board members of Rothschild Global in January 2015:

 ‘As you know, our largest client relationship in terms of revenues is a Malaysian family. We manage and oversee assets worth $500m and bill around CHF 3m a year (the relationship has been with us for over five years now). The family runs a business in third generation and is estimated between $2 and 3 bn. The family are extraordinarily connected across the globe and we belief [sic] there is more scope for us to offer Rothschild services if we get things right’.

Indeed, insiders have confided to Sarawak Report there was a whole department at the trust management division dedicated to Jho Low’s affairs.

Thus it was that the bank had chosen not to report blatantly suspicious transactions till as late as November 2015, months after the exposure of the role of Jho Low’s Good Star Limited account at Coutts Bank in Zurich during the initial heist from 1MDB. Good Star was the source of hundreds of millions of dollars funnelled into Rothschild, according to the complaint by the regulators at the Swiss Federal Finance Department (FFD), yet Rosthschild still said nothing.

Jho Low’s 130 Rothschild Trusts

The detail to emerge from the FFD submissions regarding the “Over $500 million” Rothschild managed by 2015 is even more pertinent to the questions surrounding Jho Low’s evident remaining wealth.

The regulator reveals that Rothschild’s mystery client had been introduced to the bank by none other than a Swiss wealth manager at Goldman Sachs, the bank with whom Jho Low and Najib had connected with in 2008 to advise on the Terengganu Investment Authority and which later colluded in raising of 6.5 billion dollars in bonds for 1MDB, much of which was stolen.

That introduction took place as early as October 2009, days after Jho Low had made his first $700 million windfall from the bogus 1MDB joint venture with PetroSaudi (money that was paid into Good Star Limited in Zurich).

The wealth manager had tipped off Rothschild that even Goldman Sachs had decided Jho Low was too hot to touch in terms of a personal account.  He nonetheless performed an introduction for Jho Low, accompanied by his brother and father, to Rothschild bankers at the Goldman Sachs offices in Zurich in full knowledge of those concerns.

Rothschild had proceeded to onboard Jho and his family, and just over a year later the Goldman banker who had made the original introduction moved over to become the CEO of the Rothschild Trust subsidiary.

These are not the only facts that stick out from the belated indictments. For the first time, the case revealed that there had, in fact, been no less than 130 trusts with 130 matching bank accounts held on behalf of the Low family by this one bank by the time the matter was brought to the attention of the Swiss authorities.

According to the 97 page FFD ‘criminal order’, the amount of money that had entered this network of cash and assets totalled $613 million by the time of the DOJ crackdown in 2015. As of August 2015, there remained $80 million in cash plus the 130 trusts curating assets purchased with that inflow. $313 million had entered from Good Star Limited 2009-2011, after which just under $300 million was later added from other sources connected to Jho Low.

Those figures, which only became available in 2023, are in considerable excess of what was published at the time of the DOJ settlement in 2019, which named 23 trusts representing a reported investment of $260 million (according to the media reports on the court case).

Neither did the 2019 DOJ settlement make mention of any surrender of the $80 million in cash now known to have been held by Rothschild on behalf of Jho Low.

It appears that $350 million worth of cash and assets may have therefore remained, undeclared, in the Rothschild accounts – assets that could be now worth many times more with the benefit of expert investment.

Hong Kong Hideout?

Sarawak Report repeatedly asked Rothschild Bank to confirm whether all of Jho Low’s accounts were surrendered to the authorities at the time of the US asset seizure in 2016.  If not, we asked the bank to explain what happened to the $350 million that appears not to have been accounted for?

Rothschild responded that “Rothschild & Co has fully cooperated in the 1MDB matter with the relevant law enforcement authorities since 2016 and provided full transparency …. Rothschild & Co has no remaining role in the management of Jho Low’s wealth“.

When asked for clarification as to whether the remaining hundred plus trusts and accounts that apparently were not surrendered to the DOJ in 2016 contained any assets, the spokesman declined to comment further, saying “The previous statement provides our full response on this matter“.

It was not till over one year after that surrender that FINMA announced its conclusion, in July 2018, that Rothschild had been in “serious breach” of suspicious transactions and money laundering reporting requirements with regard to Jho Low, placing the bank and its trust management subsidiary under special monitoring. The announcement did not state whether any assets remained within the bank belonging to Jho Low or his family at that time.

However, enquiries by Sarawak Report have confirmed that US investigators faced considerable hurdles during their original investigations relating to Swiss banking secrecy. The burden remained on the FBI to present its own findings to the bank in order to gain the cooperation and surrender of the trusts they had identified as purchased by the Low family through the money laundering of stolen assets.

There was no commitment made by the bank to reveal further details about their client and at no point did Rothschild offer to open up the Low family portfolio to reveal what other assets were being curated by the bank.

This, despite the fact that Jho Low had been unmasked as a fugitive from justice who had stolen billions from Malaysia and had no legitimate independent wealth. Sarawak Report has written to FINMA to enquire if it investigated whether any of Jho’s wealth remained at the bank or had been transferred elsewhere; so far, we have received no response.

In November 2018, just months after the reprimand from FINMA, Rothschild SA facilitated a ‘management buy-out’ of its trust subsidiary, enabled by a loan provided by the bank to its former staff. This now ‘independent’ operation was named Sequent and appears as a result of the sell off to have no longer been subject to the enhanced monitoring put in place against Rothschild as a penalty by FINMA.

Responding to questions on this matter Rothschild has replied “The sale of the trust business in 2018 was unrelated and a consequence of a strategic review of the entire Wealth Management business of Rothschild & Co.”

It was more than another two years before the Swiss FFD completed its investigations into the Low family funds that entered Rothschild and brought its cases against the two officers in 2022.

Meanwhile, shortly after the Low family’s October 2019 settlement with the US Department of Justice, another development took place in Hong Kong (from where the Low family based their businesses and had worked with local Rothschild investment managers prior to the scandal breaking out). In early 2020, a senior and well-connected former member of that Rothschild team announced the setting up of a private investment fund valued at well over half a billion US dollars.

Sarawak has received information that the fund is believed to represent wealth beneficially owned by Jho Low and his family members: Rothschild itself is a substantial minority shareholder in the relevant fund management company.

The fund management company has failed to respond to questions. Rothschild’s spokesman has told Sarawak Report that, as a minority shareholder, the bank does not have direct access to the investors’ ownership details, however the bank has “received assurance in writing … that none of its investors , nor any of the ultimate beneficial owners of its investors, are connected to Jho Low.”

The fund has advertised minimal activity, supporting just three named investments during the past six years. One investment in a European tech company was originally purchased by a Rothschild trust vehicle for $200 million in 2017 then divested for an extraordinary mark up of $1.3 billion in 2022 to a separate Rothschild Trust vehicle, investing alongside the Hong Kong fund and another private equity partner.

A second investment is likewise linked to a Rothschild-related earlier investment and the third was soon divested. In response to questions as to whether Jho Low or his family did at any point retain a beneficial interest in these investments Rothschild have answered “we have checked our investor records, including the anti-money laundering and Know-your-customer records of … the indicated transaction. According to these records, that are kept in line with our regulatory obligations, we did not identify Jho Low as one of the investors“.

Sarawak Report has asked for clarity on which of the string of transactions this refers to.  Meanwhile, is seem clear that barring further investigation or the release of further information from the bank, a sum of over $350 million identified as having been invested in some 100 of 130 trusts prior to 2015 remain inadequately accounted for following the surrender of Jho Low’s identified US assets to the DOJ.

There appears to have been little oversight of those trusts or the bank over much of the relevant period, during which the later heavily criticised institution was free to manage the accounts and assets as it chose.

If this wealth were to have been transferred to Hong Kong, by 2020 it would have come under the aegis of an increasingly opaque financial centre by then controlled by China, a country which chooses to harbour and protect this global fugitive who has carried out numerous missions on behalf of the Chinese state in Malaysia, the Gulf and United States.

Most recently, in July, Sarawak Report reported the fraudster’s key engagement in the negotiations to settle outstanding 1MDB related debts still owed by Malaysia to China.

Yet, as all Malaysians know, any assets remaining under Low family beneficial ownership are theirs: assisting in their transfer post 2016 would have amounted to aiding and abetting a known criminal.

As for those pardon offers, given the level of profit realised from other investments made by the 1MDB looters (Tarek Obaid’s Palantir shares, for example, were bought for $2 million and are now worth $468 million) the $353 million apparently unaccounted for in 2016 could very well represent billions now – certainly enough to fund the pardon offers Jho Low is reputed to have made.