Back For Another Year

Welcome back to The Crux for 2024!

I hope you had a nice break and are ready to read about hilarious and sometimes depressing enterprise tech news.

Let’s dive right in…

Things to note

The "Triangulation" iPhone hack is a genuinely sophisticated one. This is the kind of benchmark we should compare claims of "sophisticated attack" against when companies put an unsecured FTP server full of customer data on the Internet.

Meanwhile the UK is making further moves to undermine encryption with a new Investigatory Powers (Amendment) Bill. The concerning bit is the power to “issue notices preventing them from making technical updates that might impede information-sharing with UK intelligence agencies” according to industry group TechUK. That’ll be the “don’t turn on end-to-end encryption” power. We know the Five Eyes intelligence agencies coordinate their efforts across nations to ratchet up their powers, so expect this sort of clause to appear in Australia shortly. It’s not like cyber security is already hard or anything.

This opinion piece in The Register on the idea of banning ransomware payments is worth a read.

Israel will subsidise Intel’s planned $25 billion chip plant in Kiryat Gat to the tune of $3.2 billion. That’s billion with a B. I was curious, so I looked it up: Kiryat Gat is about 25km from northern Gaza.

Bloomberg claims ASML cancelled a bunch of machine shipments to China at the USA’s request before the actual export bans came into effect.

It seems Japanese chip materials company JSR asked the state-backed Japan Investment Corporation to take it over after Germany’s Merck made an offer. Semiconductors are closely coupled with geopolitics at the moment and state intervention in markets is the name of the game.

Liquid cooling is back as chips get hotter. I remember when 10kW per rack was a challenge due to power constraints at the substation feeding a client’s datacentre, and now we have 40kW per rack expectations.

$162 million in CHIPS Act funding has been awarded to Microchip Technology to expand its Oregon and Colorado plants. This is second chunk of funding after BAE Systems got $35 million in December 2023.

Oklahoma’s largest not-for-profit health network, Integris Health, lost control of a bunch of patient data. Isn’t it great that we made all our really intimate and personal information easy to copy and then didn’t protect it?

23andMe is trying to blame its customers for its failure to keep their data secure. A bold strategy. You may recall that 23andMe originally claimed only 14,000 customers were affected when it was actually 6.9 million. I look forward to 23andMe going bankrupt and its executives going to jail. A man can dream.

Aussie design software company Canva is planning to let insiders sell off some stock, according to The Information. About $1 billion worth at a $26 billion valuation, apparently.

The New York Times has sued OpenAI and Microsoft for copyright infringement. I expect this to get settled before any kind of general legal precedent gets set that might threaten the business model of generative AI companies. The NYT will get a payout and an ongoing license fee, but everyone else who can’t afford a years-long legal battle will have their stuff enclosed and used without compensation.

Interesting improvements to the way Apple’s AirTags work that reduce stalker risk but also maintaining privacy. It’s worth reading the details on the technique used here.

Anil Dash has an optimistic piece in Rolling Stone about an imminent reshuffling of power on the Internet. It’s an interesting piece and worth a read. I think the regulatory changes are the ones to watch, personally. Especially the ones relating to privacy and security.

I don’t cover cryptocurrency usually but this scam is amazing enough to mention: The CEO of collapsed totally-not-a-pyramid-scheme HyperVerse apparently doesn’t exist. Alleged Australian regulator ASIC has done its usual nothing, demonstrating the purpose of this system.

Adrian Cockroft reckons it’s time to leave AWS and that the fast-growth phase is over. It’s consolidation time. He notes that a bunch of people he knows have already left. He also links to this blog post from someone who claims to be an AWS employee describing the internal vibe, and has been contacted by internal people supporting the post.

The head of the EU’s antitrust efforts, Margrethe Vestager, will meet with a bunch of big tech execs next week. Apple, Google, Broadcom, Nvidia, and OpenAI are on the list. Competition regulation is an area I’m keeping a close eye on.

Niklaus Wirth, the creator of Pascal, has died. He was 89 years old. His work was hugely influential, and I recommend reading his paper A Plea for Lean Software.

An interesting set of reflections on being a social scientist in tech.

Since you don’t really want to be back at work yet, enjoy playing Paku Paku: 2D Pac Man

Longer reads

TechCrunch has a long piece on Australia’s tech sector.

It’s a bit of a puff piece, and there are the obligatory mentions of the only Australian tech companies people know: Atlassian and Canva. It provides a broad overview of the situation here in the country where I live that’s worth reading, but don’t regard it as definitive.

I had to chuckle at the assertion that “Australia’s tech ecosystem has hit a maturation phase”. I would assert that it’s still a fairly immature tech market by global standards.

There is a glaring omission, though: the whole cybersecurity side of the market. Globally it’s a huge market, and Bugcrowd was founded here; founder Casey Ellis is Australian.

The article talks about government incentives, but completely misses the plan to spend a huge amount on cybersecurity in coming years after some big Australian companies (Optus, Medibank) got thoroughly and embarrassingly pwned. That kind of large cash injection, and the prospect of attaching yourself limpet-style to an ongoing government money pump, is surely worth a mention.

It’s a weird oversight, but also an opportunity for arbitrage, I guess.

Bonus Longreads

A retrospective on the Go programming language after ten years.

The Swiss used to all have goiters. This is the tale of why, and how the Swiss eventually stopped it. It’s a very entertaining read.

Microsoft’s Copilot Obsession

Microsoft has gone really hard on AI with its Copilot brand. Alarmingly so.

The company recently put out Copilot apps that support GPT-4 and DALL-E 3 for Android, iPhone, and iPad.

The latest angle is forcing partners to put a Copilot key on keyboards if they want to ship laptops and PCs with Windows. It’s unclear why a Clippy button should exist, but I have a scroll lock key I don’t use so why not? Let’s relive the glorious days of LISP and emacs and CTRL-HYPER-META-TAB chord combinations.

More seriously, it’s yet another move from Microsoft that is forcing its Copilot branded products on customers, whether they want them or not. Every even vaguely AI-related thing Microsoft has is getting branded as Copilot and shoehorned into existing products with all the subtlety of a cow exploding in your loungeroom.

Which prompts the question: Why?

66% of AI venture investment was from Microsoft, Google and Amazon, according to private research from PitchBook. This excludes Facebook/Meta, which we know from past coverage in The Crux bought a huge chunk of Nvidia’s AI-targeted inventory last year.

Making back that money with a risk-weighted return that justifies the investment, particularly in a world with positive interest rates again, is going to require a lot of revenue. How will this revenue be generated?

Generative AI stuff requires a lot of compute and a lot of data to train up the fancy models, and the big cloud players have a lot of compute and storage to sell. If they can convince enough customers that renting computers from them is the future, well, they’ll make a lot of money, right? If you can connect everything else customers already have to the cloud/AI and make it impossible to escape, you can sit back and enjoy the equivalent of ground rents forever.

Microsoft has two key advantages over competitors here: It has both a cloud in Azure and the Windows ecosystem that runs lots of enterprises. Microsoft’s ecosystem is wider and has more hooks into things that it can tie into a rent-generating AI system. Google has Chrome and Chromebooks, but they’re less widely used, though keep an eye on the education market. Facebook has an app that billions of people use constantly, but it’s just an app and it isn’t baked into critical infrastructure the way Windows Server or Excel or Exchange/email is. AWS has its cloud but has struggled with anything more than infrastructure-as-a-service building blocks.

This larger surface area gives Microsoft a bigger opportunity to place hard-to-avoid rent generating turnstiles to everything. A bit like how touching any part of the Internet virtually guarantees sending some money to either Google or Facebook via the sticky strands of adtech that make up the world wide web. Microsoft would very much like to be the spider at the centre of the Internet of AI. So would Facebook, Google, and AWS, but Microsoft started spinning its web before any of these competitors existed.

An alternate explanation is that Microsoft is branding everything Copilot so that if its over-rotation into AI doesn’t pan out it can sweep the losses under so many carpets that it won’t be as obvious. Plausible deniability might keep the shareholder lawsuits at bay if/when another AI winter arrives.

This alternative requires a lot less competence and is also funnier, which is why I prefer it.