Not A Bug But A Feature

Source

Datafinnovation are out with more impressive research described in a blog post entitled Losing The Blacklisting Race and a paper entitled Enforcement Speeds In DeFi: Limits to a Race Against Time by Ben Charoenwong et al.

The blog post is more accessible, so that's what I will quote from. But for the details you need to read the paper. The blog post's TL;DR is:

First, we explore the empirical reality of OFAC’s blacklisting efforts. The results there are stark: nearly all addresses are completely empty by the time OFAC manages to blacklist them. This result is robust across time, attacker and type of attack. Second, we construct a simple model to explain why this outcome is inevitable with competent attackers. Evasion strategies which work in our model match those we find employed in the wild.

Below the fold, I start from Datafinnovation's work and explore its context.

2026 Optical Media Durability Update

Eight years ago I posted Optical Media Durability and discovered:

Surprisingly, I'm getting good data from CD-Rs more than 14 years old, and from DVD-Rs nearly 12 years old. Your mileage may vary.

Here are the subsequent annual updates:

• 2019
• 2020
• 2021
• 2022
• 2023
• 2024
• 2025

It is time once again for the mind-numbing process of feeding 45 disks through the readers to verify their checksums, and yet again this year every single MD5 was successfully verified. Below the fold, the details.

More On Robotaxis

Cybercab

Liam Denning reports on Tesla's quarterly results flop in Someone Call Tesla a Robotaxi, or SpaceX, Quick. He notes that:

The curated list of investor questions that gets teed up ahead of these affairs had indicated some brewing discomfort about the big issue: Tesla’s lack of progress on its long-promised mass rollout of robotaxis. Musk duly tried to finesse this by pointing out that the “constraint” is safety, with Tesla trying to scale as fast as possible “while trying to ensure that we do not harm anyone.”

It is good that Tesla is "trying to ensure that we do not harm anyone" but they need to try much harder. In last February's Tesla's Not-A-Robotaxi Service I quoted Fred Lambert:

By the company’s own numbers, its “Robotaxi” fleet crashes nearly 4 times more often than a normal driver, and every single one of those miles had a safety monitor who could hit the kill switch. That is not a rounding error or an early-program hiccup. It is a fundamental performance gap.

But I went on to point out that Lambert was making the wrong comparison:

However badly, Tesla is trying to operate a taxi service. So it is misleading to compare the crash rate with "normal drivers". The correct comparison is with taxi drivers. The New York Times reported that:

In a city where almost everyone has a story about zigzagging through traffic in a hair-raising, white-knuckled cab ride, a new traffic safety study may come as a surprise: It finds that taxis are pretty safe.

So are livery cars, according to the study, which is based on state motor vehicle records of accidents and injuries across the city. It concludes that taxi and livery-cab drivers have crash rates one-third lower than drivers of other vehicles.

A law firm has a persuasive list of reasons why this is so. So Tesla's "robotaxi" is actually 6 times less safe than a taxi.

In any rational jurisdiction, six times worse than the competition even with a safety driver would get the regulators to force Tesla back to the drawing board.

The New York Times article was the best I could find at the time but it was from 2006. We now have much better and more recent data. Follow me below the fold for one of the reasons why Tesla's numbers are bad, very probably even worse than six, and even some criticims of Waymo's numbers.

Microsoft's Project Silica

598

$68.6B

$0.115

1418

$28.0B

$0.020

59.2

$0.51B

$0.003

I summed up my big picture view of archival media eight years ago in Archival Media: Not a Good Business. Whatever your choice of technology, the economics are brutal. This is a version of the table in that post, updated to 2021 and again based upon IBM data. Note that archival media shiped 4% as many bytes as hard disk and generated 1.8% or the revenue. It remains a tiny market.

I have written several times about Microsoft Research's Project Silica, most recently in last year's Archival Storage:

I'm skeptical of "commoditizing the technology". Archival systems are a niche in the IT market, and one on which companies are loath to spend money. Realistically, there aren't going to be a vast number of Silica write heads. The only customers for systems like Silica are the large cloud providers, who will be reluctant to commit their archives to technology owned by a competitor. Unless a mass-market application for femtosecond lasers emerges, the scope for cost reduction is limited.

But the more I think about this technology, which is still in the lab, the more I think it probably has the best chance of impacting the market among all the rival archival storage technologies. Not great, but better than its competitors:

I followed this with a list of eight major reasons for my opinion.

Below the fold an update on the project and an assessment.

Distilling The Moat

Whisky Still

The original function of a Web server was to respond to queries by revealing the appropriate part of their internal data. This necessarily meant that repeated queries, for example from a search engine's or an internet archive's web crawler, could extract the server's entire internal data. Since the extracted data had been published on the Web, it was not trade secret. It was protected by the publisher's copyright. This has led to many lawsuits, for example against the Internet Archive, Google and others. It is the reason search engines only display "snippets" of the content they collect.

AI companies' intellectual property is their models. They spend vast sums funding the technical and human resources to "train" these models, the racks of GPUs in the data centers, and the hordes of workers labeling images, and having "genuine human conversations" with the nascent model. These expenditures are thought to create a "moat" around the value thus generated, because it would be equally expensive for a competitor to create an equivalent model. It is this moat that supports their extraordinary valuations, despite their lack of earnings.

Below the fold I explain why their moat is very shallow.

Portents Of Doom

Elon Musk is the world champion of totally implausible projections, and Kim Khan reported on a personal best in SpaceX sees total addressable market rivaling size of the U.S. economy:

The $28.5T forecast compares to U.S. Q1 2026 nominal GDP of nearly $32T, with the estimate for the market of AI enterprise applications of $22.7T about 70% of total U.S. economic output.

Sam Altman and Dario Amodei just aren't this good, but their projections of their Total Available Market (TAM) are still turning out to be vastly optimistic. In AI's Affordability Crisis I showed evidence that the AI platforms could no longer afford the massive subsidies they were using to artifically inflate demand for their product, and that reducing the subsidies had made their enterprise customers reconsider their enthusiasm for deploying them. This is leading to investors belatedly realizing that AI platforms' projections of their TAM and thus their valuations are totally implausible.

This re-calibration is just one of the many signs that the AI bubble is about to deflate. Below the fold I present a necessarily incomplete list of them, which I will try to update as more appear.

The Kelvin Limit

I have been a small part of the chorus of voices critiquing the AI bubble that I described in Portents Of Doom. But what if we're wrong? What if the overwhelming demand for AI sin't the result of the AI platforms massively subsidizing their products, but because the world needs more and more non-consensual sex images, slop web pages, agentic ransomware attacks, students cheating on exams, hallucinated lawsuits, and all the other benefits of this transformative technology?

Please suspend disbelief and follow me below the fold as I look into a fascinating examination of the implications of the exponential growth in the data centers needed to provide these benefits.

My Introduction To Computer Graphics

Boeing's PDP-7/340

Most of my career has been involved in various ways with computer graphics. Below the fold I recount the story of how I got started in the field just as it was getting started. To give you some idea of just how early my introduction was the Mother of all Demos had been the year before. The displays I got to work with drew lines in monochrome, not rasters in color. You created the image by writing a loop of instructions in the "display processor" instruction set. These told it the lines to draw at each refresh cycle. There was no mouse.

Coprophagia Is Bad For You

Divine, Pink Flamingos

Wikipedia defines Coprophagia as "the consumption of feces".

Since "brevity is the soul of wit", my favorite science fiction includes the 254 words of Fredric Brown's Answer from 1954. It describes a galactic civilization holding a ceremony to mark the final connection of all their computers. What happened was:

Dwar Ev threw the switch. There was a mighty hum, the surge of power from ninety-six billion planets. Lights flashed and quieted along the miles-long panel.

Dwar Ev stepped back and drew a deep breath. “The honor of asking the first question is yours, Dwar Reyn.”

“Thank you,” said Dwar Reyn. “It shall be a question that no single cybernetics machine has been able to answer.”

He turned to face the machine. “Is there a God?”

The mighty voice answered without hesitation, without the clicking of single relay.

“Yes, now there is a God.”

Sudden fear flashed on the face of Dwar Ev. He leaped to grab the switch.

A bolt of lightning from the cloudless sky struck him down and fused the switch shut.

This may have inspired Douglas Adams' similar but much longer scenario in which the answer turned out to be 42.

Below the fold I trace the connection between these two ideas.

AI's Affordability Crisis

A year ago in The Back Of The AI Envelope I pointed out that the AI platforms were running the drug-dealer's algorithm, "the first one's free". By massively subsidizing the use of their products, they were generating overwhelming demand for them. They used this demand to justify massive investments, in the hope that, by the time they had to show a return on these invetment, the users would be so addicted that they would pay the vastly higher prices needed to generate a return.

David Cahn, Sept '23

I have to confess that I was late to the party. The earliest skepticism I've been able to find was from Sequoia Capital's David Cahn in September 2023, entitled AI’s $200B Question. Only nine months later Cahn re-ran the same analysis in AI’s $600B Question. His estimate of the revenue gap had tripled. Cahn wasn't alone. Independent journalists such as Ed Zitron were flagging this problem long before I was.

I started to write this post a couple of months ago when the maiinstream business press began to notice companies complaining about the cost of the tokens their employees were burning. Since then the trickle has turned into a flood, which made finishing the post hard. Below the fold I throw up my hands and dump out a small sample from the flood.

Chatbots vs. Ozone

Source

Back in February I posted The Kessler Syndrome, which also included a brief section mentioning the impacts of the proposed megaconstellations on the environment, specifically global warming from CO2 and black carbon, and depletion of the ozone layer. Three months earlier Anton Petrov had examined the last of these in Risk of Ozone Layer Destruction from Internet Satellite Swarms and Rocket Fuel. He has now followed up with SpaceX Is Conducting a Giant Chemical Experiment on Our Atmosphere Without Realizing. Below the fold I survey the papers Petrov cited and a few others.

Library of Congress Storage Architecture Meeting 2026

Once again I attended most of the library of Congress' Designing Storage Architectures workshop remotely. I apologize for the delay in posting this; domestic duties have kept me very busy recently. Below the fold notes on the talks that caught my attention, based on my now somewhat memory and the slide decks for the talks from the Library of Congress website.

AI's PR Problem (Updated)

J.P. Morgan hits photographer with cane

This is just a brief post to explain to my old boss, Eric Schmidt, why he and his ilk are getting booed at college commencements, and why laws against data centers are getting passed. The explanation is below the fold.

Wrench Attacks

XKCD #538

A year ago I wrote The Risks Of HODL-ing sparked by Mitch Moxley's They Stole a Quarter-Billion in Crypto and Got Caught Within a Month. Moxley recounts the kidnapping of Veer Chetal's parents to persuade him to hand over his share of the loot:

the Lamborghini was suddenly rammed from behind by a white Honda Civic. At the same time, a white Ram ProMaster work van cut in front, trapping the Chetals. According to a criminal complaint filed after the incident, a group of six men dressed in black and wearing masks emerged from their vehicles and forced the Chetals from their car, dragging them toward the van’s open side door.

Below the fold I look at Bloomberg updates from last week on why the crypto-bros are having to spend vast sums on defending against the threat of HODL-ing.

Talk for Stanford's EE 292J

Via John Markoff, I was invited to a conversation with Jonathan Dotan and the students of his EE292J course entitled Designing for Authenticity. Below the fold are my brief introductory remarks, and some notes for the discussion.

Flooded Zones Part 2

Source

This is the promised follow-on to Flooded Zones Part 1, which discussed the Distributed Denial of Service (DDoS) attack being mounted by AI against the scholarly publication system. By reducing the cost of generating and submitting a paper or a review, AI has caused a massive increase in the quantity and a significant decrease in the quality of submissions to a system that was already vastly overloaded.

Below the fold I look at AI-enabled DDoS attacks against two other even more important areas; software security and political discourse (as shown in the overview image).

Flooded Zones Part 1

Tom Cowap
CC-BY-SA 4.0

Three years ago in Flooding The Zone With Shit, my first post on the AI bubble, I wrote:

My immediate reaction to the news of ChatGPT was to tell friends "at last, we have solved the Fermi Paradox". It wasn't that I feared being told "This mission is too important for me to allow you to jeopardize it", but rather that I assumed that civilizations across the galaxy evolved to be able to implement ChatGPT-like systems, which proceeded to irretrievably pollute their information environment, preventing any further progress.

The post title was a notorious quote from Steve Bannon. Below the fold, I look into scholarly publication, the first of three areas whose zones are currently being flooded with AI output in what can be considered DDoS (Distributed Denial of Service attacks:

A distributed denial-of-service (DDoS) attack occurs when multiple systems flood the bandwidth or resources of a targeted system, usually one or more web servers.

A subsequent post will examine two more flood zones, political discourse and software security.

The Permissionless Catch-22

Potential Attack Target

Suppose some genre of content is under attack by powerful adversaries. Lets take political satire as a thought experiment in which powerful politicians are attacking sites and Web archives hosting it by sending bogus DMCA takedowns, suing for defamation, buying up their hosting platforms, getting their flying monkeys to flood them with spam, and so on. Below the fold I discuss the problem facing the defense.

Dormant Digital Assets

PsiQuantum's computer

Four and a half years ago I wrote The $65B Prize about the potential reward for developing a "sufficiently powerful quantum computer" capable of cracking Bitcoin's encryption. It was based on work by Aggarwal et al, who were then projecting it would happen between 2029 and 2044. The $65B was the notional value of the wallet containing the million Bitcoin Satoshi Nakamoto mined originally.i But I noted that:

Chainalysis estimates that about 20% of all Bitcoins have been "lost", or in other words are sitting in wallets whose keys are inaccessible. That is around another 3.6 million stranded Bitcoin or at the current "price" about $234B.

So the potential prize was almost $300B.

Nearly a year ago I followed up with The $740B Prize. There are two reasons why the prize was then bigger but is now smaller than that:

• Bitcoin's "price" had then increased from about $65K to around $107K, but it is now around $76K.
• Because the "market cap" of Michael Saylor's Strategy was 1.6 times the "market cap" of its stash of Bitcoin, it was possible to use Saylor's algorithm to amplify the prize. But the factor has decreased from 1.6 to 0.81, so the algorithm no longer works.

But the threat to Bitcoin, and other cryptocurrencies, is far worse than I described in either of these two posts. The date is closer and the range of threats much broader. Follow me below the fold for the details.

Angels in America

I have wanted to write this post for a long time, but I was waiting until I could visit the invaluable Royal National Theatre Archive to check my memory of their early productions. It doesn't look like I'll be in London any time soon, and I have the time now to write a long post about a long play, so here goes.

Growing up in London meant that theatre has always been an important part of my life. I have seen a great many plays including some legendary performances and magnificent productions, such as Royal National Theatre's 2014 King Lear. One of my particular theatrical interests is long-form plays. Highlights of this genre have included:

Play Text

• Ken Campbell's Science Fiction Theatre of Liverpool's 9-hour Illuminatus Trilogy
• The Magic Theatre's all-night production of Murray Mednick's Coyote Cycle
• The Royal National Theatre's 6-hour adaptation of Phillip Pullman's His Dark Materials
• Taylor Mac's 3x8-hour A 24-Decade History of Popular Music

But there is one such play that is very special to me, Tony Kushner's 7+ hour Angels in America. It is clearly among the greatest plays of the 20^th century. I was there at the beginning, and I have seen many productions since. Below the fold I recount my history with this masterpiece.

The Handoff Problem (Updated)

Source

Around twelve years ago, Google figured out the fundamental problem facing Tesla's Fake Self Driving. Almost nine years ago in Robot Cars Can’t Count on Us in an Emergency, John Markoff wrote:

Three years ago, Google’s self-driving car project abruptly shifted from designing a vehicle that would drive autonomously most of the time while occasionally requiring human oversight, to a slow-speed robot without a brake pedal, accelerator or steering wheel. In other words, human driving was no longer permitted.

The company made the decision after giving self-driving cars to Google employees for their work commutes and recording what the passengers did while the autonomous system did the driving. In-car cameras recorded employees climbing into the back seat, climbing out of an open car window, and even smooching while the car was in motion, according to two former Google engineers.

Gareth Corfield at The Register added:

Google binned its self-driving cars' "take over now, human!" feature because test drivers kept dozing off behind the wheel instead of watching the road, according to reports.

"What we found was pretty scary," Google Waymo's boss John Krafcik told Reuters reporters during a recent media tour of a Waymo testing facility. "It's hard to take over because they have lost contextual awareness."

Follow me below the fold for a wonderful example of Tesla's handoff problem, and a discussion of the difference between Tesla's and Waymo's approaches to self-driving.

Metastablecoin Fragmentation (updated)

A fundamental problem for decentralized systems like permissionless blockchains is that their security depends upon the cost of an attack being greater than the potential reward from it. Various techniques are used to impose these costs, generally either Proof-of-Work (PoW) or Proof-of-Stake (PoS). These costs have implications for the economics (or tokenomics) of such systems, for example that their security is linear in cost, whereas centralized systems can use techniques such as encryption to achieve security exponential in cost.

Shin Figure 3

Now, via Toby Nangle's Stablecoin = Fracturedcoin we find Tokenomics and blockchain fragmentation by Hyun Song Shin, whose basic point is that these costs must be borne by the users of the system. For cryptocurrencies, this means through either or both transaction fees or inflation of the currency. The tradeoff between cost and security means that there is a market for competing blockchains making different tradeoffs. In practice we see a vast number of competing blockchains:

Tether’s USDT sits on 107 different ledgers. ... USDC sits on 125.

The chart shows Ethereum losing market share against competing blockchains.

Shin's analysis uses game theory to explain why this fragmentation is an inevitable result of tokenomics. Below the fold I go into the background and the details of Shin's explanation.

Skynet Progress Report (updated)

Source

I, for one, welcome our new insect overlords

Kent Brockman in "Deep Space Homer", The Simpsons

In recent months Cyberdyne Systems Corporation and its many subsidiaries have made very encouraging progress towards removing some of the major road-blocks standing in the way of the initial deployment of Skynet. Below the fold I report on the most significant ones.

Tesla's Not-A-Robotaxi Service

Source

I have now seen the fabled CyberCab three times in real life. It has two seats, one of them fully equipped with human driver interface equipment. In each case a human was using them to drive the car, which is necessary in California because Fake Self-Driving is a Level 2 driver assistance system that requires a human behind the wheel at all times. A Robotaxi that requires a human driver and can carry at most one passenger isn't going to be a economic success.

Fred Lambert has two posts illustrating the distance between Musk's claims and reality. Below the fold I look at both of them:

The Kessler Syndrome

LEO in 2019 (NASA)

In 1978 Donald J. Kessler and Burton G. Cour-Palais published Collision Frequency of Artificial Satellites: The Creation of a Debris Belt. Wikipedia notes that:

It describes a situation in which the density of objects in low Earth orbit (LEO) becomes so high due to space pollution that collisions between these objects cascade, exponentially increasing the amount of space debris over time.

This became known as the Kessler Syndrome. Three decades later, shortly after Iridium 33 and Cosmos 2251 collided at 11.6km/s, Kessler published The Kessler Syndrome, writing that the original paper:

predicted that around the year 2000 the population of catalogued debris in orbit around the Earth would become so dense that catalogued objects would begin breaking up as a result of random collisions with other catalogued objects and become an important source of future debris.

And that:

Modeling results supported by data from USAF tests, as well as by a number of independent scientists, have concluded that the current debris environment is “unstable”, or above a critical threshold, such that any attempt to achieve a growth-free small debris environment by eliminating sources of past debris will likely fail because fragments from future collisions will be generated faster than atmospheric drag will remove them.

Below the fold I look into the current situation.

Mind The GAAP Again

Source

A bit over three months ago I wrote Depreciation and started with this graph from my 2022 post Generally Accepted Accounting Principles about the way Bitcoin miners were inflating their profits through misleading depreciation of their rigs.

The key message of the graph is the contrast between the 5-year straight-line depreciation and the curves showing the value of the remaining Bitcoin that the rig will generate. I suggested that the same mismatch between straight-line depreciation and remaining value generation would apply to AI hardware. I don't claim to be the first to flag this issue; The Economist's The $4trn accounting puzzle at the heart of the AI cloud was about a month earlier.

About a month later I returned to AI economics with Mind The GAAP, but that was mostly focused on other parts of the puzzle.But now, thanks to Bryce Elder's Big tech’s $680bn buy-now-book-later problem it turns out that both Michael Burry of The Big Short and Morgan Stanley Research agree with me that there's a problem:

Below the fold I go into the details, with many thanks to Bryce Elder.

Funding Open Source?

$BLEEBZORX chart

Most of the world's software infrastructure is, or is based upon, open source. The developers and supporters of some of it, for example the Linux kernel, and the major compilers, are paid by technology companies because they are critical to their business. Other, less visible but similarly critical parts are supported by lone volunteers. Apart from the unfairness, this can lead to serious vulnerabilities. Back in 2018 I wrote about one such vulnerability, the event-stream hack, in Securing The Software Supply Chain

The attackers targeted a widely-used, fairly old package that was still being maintained by the original author, a volunteer. They offered to take over what had become a burdensome task, and the offer was accepted. Now, despite the fact that the attacker was just an e-mail address, they were the official maintainer of the package and could authorize changes.

The change they authorized included code to steal cryptocurrencies.

In 2020 I wrote a detailed post about this problem entitled Supporting Open Source Software. Recently the topic re-surfaced on an e-mail alias I read. But what triggered the post below the fold was that this coincided with yet another fascinating piece from Matt Levine and his laugh-out-loud follow-up the next day.

Internet Archive's Storage

Internet Archive Staff

Bruce Li introduces his The Long Now of the Web: Inside the Internet Archive’s Fight Against Forgetting thus:

This report delves into the mechanics of the Internet Archive with the precision of a teardown. We will strip back the chassis to examine the custom-built PetaBox servers that heat the building without air conditioning. We will trace the evolution of the web crawlers—from the early tape-based dumps of Alexa Internet to the sophisticated browser-based bots of 2025. We will analyze the financial ledger of this non-profit giant, exploring how it survives on a budget that is a rounding error for its Silicon Valley neighbors. And finally, we will look to the future, where the "Decentralized Web" (DWeb) promises to fragment the Archive into a million pieces to ensure it can never be destroyed.

It is long, detailed, comprehensive and well worth reading in full. Below the fold I comment on the part about storage.

Good Questions

Source

On November 21^st Bryce Elder posed Five questions from an ignorant no-coiner about the crypto crash. Each of his five questions identified some interesting apparent anomalies.

Below the fold I look into each of his questions, asking how anomalous its anomalies really were and whether they have persisted into the New Year.

TL;DR none of them are really surprising but reaching that conclusion took a good deal of research.

Meta: Post #1000

This is the one thousandth post to this blog in the 212 months since the first post. That is an average of 4.7 posts per month, or just over one per week, which is my long-term goal for the roughly half my time that isn't taken up with grand-parenting.

Some posts are a lot of work, and take more than a week. Major talks, such as The Gaslit Asset Class or Lessons from LOCKSS typically represent a month's work, as do long posts such as Sabotaging Biitcoin, Drones or The Dawn Of Nvidia's Technology.

The 1000 posts have gained over 6.88M page views, 7.6% of which were for my EE380 Talk. Less publicized but popular posts get around 30K page views, well above the 6.9K average.

The only one of these statistics that I care about is the goal of a post a week. Having an audience is nice when it happens, but that's not why I'm writing. I write for myself, to understand not necessarily to communicate. Despite this, I'd like to thank those who read and comment.