Slide 1
Slide 1 text
Understanding
What We Have Wrought
Patrick McKenzie — Bank of England
October 1st, 2025
Slide 2
Slide 2 text
No content
Slide 3
Slide 3 text
“Hidden” SPOFs
Slide 4
Slide 4 text
No content
Slide 5
Slide 5 text
No content
Slide 6
Slide 6 text
No content
Slide 7
Slide 7 text
No content
Slide 8
Slide 8 text
No content
Slide 9
Slide 9 text
No content
Slide 10
Slide 10 text
Five Whys
• Why did several banks share a software monoculture?
• Why was this configuration change not caught in testing?
• Why was this configuration change so hard to roll back and/or mitigate?
• Why did this configuration change have such a large “blast radius?”
• Why did this configuration change substantially disrupt “important business services”?
Slide 11
Slide 11 text
A Recipe For Software Monoculture
Slide 12
Slide 12 text
No content
Slide 13
Slide 13 text
Defense in Depth: Automated Testing
Crowdstrike Root Cause Analysis
Slide 14
Slide 14 text
Control Plane Issues Are Frequently Sev-0
Slide 15
Slide 15 text
4,500 Banks Couldn’t Have Correlated Failures, Right?
Slide 16
Slide 16 text
This was a Near Miss… Including For You
• Banking system in U.S. largely recovered to normal by extended hours on Friday.
• Most affected service, teller transactions, are societally critical but can be deferred a
short while. Competent, immediate efforts to divert transactions to electronic
channels.
• We had not yet had Crowdstrike completely deployed within the banks
– Saved by our sloth and incompetence!
– Consider an alternate universe in which all US-based counterparties go “dark for a
day”
• You may experience a technical crisis when weather is not normal, either incidental to
the fact of market stress or in a complex casual relationship with that stress.
Slide 17
Slide 17 text
Potential Policy Responses
• Blameless postmortems
– Ask fintechs if you can read
these!
– Particularly the near misses!
• If an engineer can cause an
outage on accident then what
could their laptop do on
purpose?
– Red team exercises
Slide 18
Slide 18 text
Stablecoins
Slide 19
Slide 19 text
No content
Slide 20
Slide 20 text
No content
Slide 21
Slide 21 text
No content
Slide 22
Slide 22 text
No content
Slide 23
Slide 23 text
No content
Slide 24
Slide 24 text
No content
Slide 25
Slide 25 text
No content
Slide 26
Slide 26 text
No content
Slide 27
Slide 27 text
“I think the answer is that it’s messy, but the
funds are real. We see legitimate inflows into
Tether from many sources—large ones—that
result in market makers selling, creating, and
sending billions of dollars to Tether’s bank
accounts. They do this to mint the tokens and
maintain relationships with Tether and its banks.
Everything checks out, albeit in a messy way.”
— Sam Bankman-Fried to Bloomberg
(Aug 8th 2021)
Slide 28
Slide 28 text
“They have the money they say they have … I’ve
seen a whole lot and the firm has seen whole a lot
and they have the money. And so there has
always been a lot of talk ‘Do they have it or not?’
and I’m here with you guys and I’m telling you
we’ve seen it and they have it.”
— Howard Lutnick to Bloomberg TV,
Jan 16th, 2024
Slide 29
Slide 29 text
“Cantor Fitzgerald is not conducting continuous
diligence on Tether’s financial statements, but I
believe my statements were accurate when
made.”
— Howard Lutnick to U.S. Senate,
Jan 25th , 2025
Slide 30
Slide 30 text
No content
Slide 31
Slide 31 text
No content
Slide 32
Slide 32 text
Patrick McKenzie on blog,
October 28th, 2019
Slide 33
Slide 33 text
AI and future of trading
Slide 34
Slide 34 text
Kaplan et al., 2020
Slide 35
Slide 35 text
No content
Slide 36
Slide 36 text
No content
Slide 37
Slide 37 text
No content
Slide 38
Slide 38 text
No content
Slide 39
Slide 39 text
I Spent August and September Falling in Love
Slide 40
Slide 40 text
AI Risks in Trading Space
• ”Every hedge fund will train their own model” not the outcome to bet on.
– Concentration risk among 3 large lab providers, of which one could be (in any given
six month window) effectively sole source to the UK trading community.
• Time to detection and resolution increasingly dependent on whether Claude Code /
OpenAI Codex / etc is up or not.
– Knight Capital: 20 minute critical window
– CrowdStrike: 90 minute time to resolution
• Recursive dependencies put larger chunks of economy on narrower shoulders
Slide 41
Slide 41 text
Thank You
[email protected]
Happy to chat if I can ever be useful, particularly
informally.
https://www.bitsaboutmoney.com Bits about
Money is freely available and relevant to your
interests.