The post-Watergate dream of private electronic cash is still alive. We have both private cryptocurrencies like Zcash and (mostly) private stablecoins. Funnily enough, usable wallets are one of the harder problems here.

The AI cyber doom and gloom just isn't creative enough. Imagine a worm that updates all PHP websites, adding user-demanded features, but then makes the sites brutalist. Or there are competing swarm ideologies and you're like: the DMV website hasn't been the same since the anarcho-syndicalist hive mind patched it.

Relevant XKCD strikes again, this time for real. A certificate authority actually generated private keys where 4 was part of the key.

"Computers now constitute or will soon constitute adangerous threat to individual privacy " Some of the first notions of cryptography, privacy , and computers came from someone at .... fucking IBM. Yes, this the Feistel of Feistel network fame. These guys were not just in it for the math. (from Steven Levy's Crypto, written in january 2001 when Crypto meant Cryptography)

Occasionally your cyberpunk future is a tragic comedy. But its never a utopia.

Dissent is essential to democracy. And when it's needed most, you need either courage or anonymity. We've repackaged anonymity as privacy, dressing it up in friendly clothes. But no one thinks they have privacy today, and it's hard to articulate why that matters. Anonymity and privacy aren't the same thing. Even if they know everything about you, you can use AI to disguise your writing style, use Tor or Apple Private Relay, make a burner account, and still post anonymously. For now.

Bitcoin isn't private, but against hostile intelligence services (whether that's the CIA or the SVR), the traditional financial system isn't private either. We need better. We finally have the cryptography for the 1980s dream of both encrypted money and encrypted messaging.

(this comment is exaggerated for comedic effect but it’s not wrong)

Why buy a quantum computer if AI keeps improving classical computing? If we can build a big enough quantum computer, we get asymptotically faster algorithms for chemistry, optimization, and finance. But the funny thing is, classical computing keeps catching up on the commercially interesting problems. So again, if classical algorithms or their AI-optimized variants are good enough for the commercially viable instances, why buy a computer to run quantum algorithms? Why invest in building one? A number of “hard” problems in theoretical computer science aren’t hard for the average commercial version of the problem. For example, checking for a low-cost route that visits all n cities is the traveling salesman problem, which is NP-hard, meaning the worst-case version is very, very hard. But real-world versions of the problem aren’t, and in fact you can solve many of them with some clever algorithms and a smartphone. A few caveats. For optimization and finance, it’s not even clear you need AI improvements, since the quantum advantage there was smaller to begin with. There's an argument for quantum in scientific computing, specifically simulating quantum physics, but that’s science, not much of a business. And AI may well improve quantum computing itself, e.g. with better error correction, but that helps the supply side, not demand.

Student: the demo characters in our anonymous group chat are Alice, Bob, Master Yoda and Darth Vader. Me, if you are going to do Star Wars references for real world cryptography , you really ought to do: • Cassian Andor • Luthen Rael • Kleya Marki * Lonni Jung

Zeroday malware delivered by targeted labs. Interesting to get confirmation of this and evidence of just how targeted ads can be.

Your cyberpunk future was never going to be a utopia, Snow Crash edition. Atlanta apartments now have robot dog guards.

Your cyberpunk future was never going to be a utopia, stop building things as if it was.

I've long wanted to build a secure wallet as software. Repurpose an old iphone in lockdown mode, put it in a faraday bag, and only communicate via QR codes on the front facing camera. Originally the point was something more accessible than a Ledger which was both secure and made folks feel safe. Would basically eliminate the hardware supply chain risk, though increase the TCB and of course depend on Apple's software security. (The QR code part of this already exists in a few wallets and standards)

I've heard a lot of bad "why didn't Ledger do x" takes. To add my own: why not fill the device with epoxy? It's not foolproof at all. I've heard of folks at the NSA breaking secure epoxied hardware as a hobby, but I imagine it would decrease the attack yield.

An interesting argument for AI hacking gets worse, even if AI makes software more secure. Most hacks aren't 0day exploits. And, as someone points out down thread, most attacks start with phishing. Of course, AI defenses might be able to make all those attacks harder too, but its not by bug finding. Maybe an AI red team, if you can trust it not to go rogue.