Patrick McKenzie (patio11) is joined by Emmett Shear, co-founder of Twitch, former interim CEO of OpenAI, who now runs Softmax AI alignment. Emmett argues that current AI safety approaches focused on "systems of control" are fundamentally flawed and proposes "organic alignment" instead—where AI systems develop genuine care for their local communities rather than following rigid rules. –Full transcript available here: www.complexsystemspodcast.com/ai-alignment-with-emmett-shear/–Sponsor: MercuryThis episode is brought to you by Mercury, the fintech trusted by 200K+ companies — from first milestones to running complex systems. Mercury offers banking that truly understands startups and scales with them. Start today at Mercury.com Mercury is a financial technology company, not a bank. Banking services provided by Choice Financial Group, Column N.A., and Evolve Bank & Trust; Members FDIC.–Links:Softmax - https://www.softmax.com/–Timestamps:(01:26) Understanding AI alignment(04:42) The concept of universal constructors(13:45) AI's rapid progress and practical applications(19:08) Sponsor: Mercury(20:19) AI's impact on work(34:59) AI's sensory and action space(42:10) User intent vs. user request(44:35) The illusion of a perfect AI(49:57) Causal emergence and system dynamics(55:19) Reflective and intentional alignment(01:01:08) Engineering challenges in AI alignment(01:04:15) The future of AI(01:26:40) Wrap
--------
1:27:28
--------
1:27:28
Building software that survives contact with reality, with Will Wilson
Patrick McKenzie (patio11) is joined by Will Wilson, CEO of Antithesis, to discuss the evolution of software testing from traditional approaches to cutting-edge deterministic simulation. Will explains how his team built technology that creates "time machines" for distributed systems, enabling developers to find and debug complex failures that would be nearly impossible to reproduce in traditional testing environments. They explore how this approach scales from finding novel bugs in Super Mario Brothers to ensuring the reliability of critical financial and infrastructure systems, and discuss the implications for a future where AI writes increasingly more code.–Full transcript available here: www.complexsystemspodcast.com/software-testing-with-will-wilson/–Sponsor: This episode is brought to you by Mercury, the fintech trusted by 200K+ companies — from first milestones to running complex systems. Mercury offers banking that truly understands startups and scales with them. Start today at Mercury.com Mercury is a financial technology company, not a bank. Banking services provided by Choice Financial Group, Column N.A., and Evolve Bank & Trust; Members FDIC.–Recommended in this episode:Antithesis: https://antithesis.com/––Timestamps:(00:00) Intro(01:23) Database scaling and the CAP theorem(08:13) Abstraction layers and hardware reality(15:28) The problem with traditional testing(19:43) Sponsor: Mercury(23:16) The fuzzing revolution(30:35) Deterministic simulation testing(42:36) Real-world testing strategies(47:22) Introducing Antithesis(59:23) The CrowdStrike example(01:01:15) Finding bugs in Mario(01:07:37) Property-based vs conventional testing(01:09:51) The future of AI-assisted development(01:14:51) Wrap
--------
1:15:41
--------
1:15:41
Defense, drones, and military procurement, with Bean of Naval Gazing
Patrick McKenzie (patio11) is joined by Bean, a pseudonymous defense industry expert, to explore the intellectual crossovers between military and civilian domains. The conversation reveals how the defense industry's fundamental constraint of having only one customer (a monopsony) creates entirely different incentives than tech, leading to conservatism and 30-50 year product lifecycles. Bean argues that drones are largely modern iterations of cruise missiles we've had since the 1950s, and explains why current anti-drone defenses make swarm attacks less threatening than headlines suggest.–Full transcript available here: www.complexsystemspodcast.com/defense-with-bean-of-naval-gazing/–Sponsor:This episode is brought to you by Mercury, the fintech trusted by 200K+ companies — from first milestones to running complex systems. Mercury offers banking that truly understands startups and scales with them. Start today at Mercury.comMercury is a financial technology company, not a bank. Banking services provided by Choice Financial Group, Column N.A., and Evolve Bank & Trust; Members FDIC. –Recommended in this episode:Naval Gazing: https://www.navalgazing.net/––Timestamps:(00:00) Intro(00:29) The overlap between tech and defense(01:35) Operations research in World War II(02:55) Mathematical insights and military strategies(05:28) The role of operations research in modern warfare(16:59) Tech and defense (Part 1)(19:48) Sponsor: Mercury(21:00) Tech and defense (Part 2)(26:07) Economics behind the defense industry(32:07) SpaceX's early challenges and achievements(33:00) The Super Hornet development story(34:39) Military procurement lessons(37:42) Aerospace industry retention rates(38:42) Lockheed Martin's dominance and supply chain(40:55) Drone technology and military applications(46:53) Anti-drone defenses and future warfare(48:01) Naval warfare and historical perspectives(01:01:03) Wrap
--------
1:02:47
--------
1:02:47
AI and the great developer speed-up, with Joel Becker of METR
This week on Complex Systems, Patrick McKenzie (patio11) is joined by Joel Becker from METR. They discuss groundbreaking research on AI coding assistants.Joel et al’s randomized controlled trial of 16 expert developers working on major open source projects revealed a counterintuitive finding: despite predictions of 24-40% speed improvements, developers actually took 19% longer to complete tasks when using AI tools, even though they retrospectively believed they were 20% faster. The conversation explores why even sophisticated professionals struggle to accurately assess their own productivity with AI tools, the industrial organization of software development, and the implications for AI's recursive self-improvement in research and development. It also touches on other perspectives from software developers using these tools professionally, and where we can expect them to improve rapidly.–Full transcript available here: www.complexsystemspodcast.com/the-great-developer-speed-up-with-joel-becker/–Sponsor: This episode is brought to you by Mercury, the fintech trusted by 200K+ companies — from first milestones to running complex systems. Mercury offers banking that truly understands startups and scales with them. Start today at Mercury.com Mercury is a financial technology company, not a bank. Banking services provided by Choice Financial Group, Column N.A., and Evolve Bank & Trust; Members FDIC.–Recommended in this episode:METR: https://metr.org/ Joel Becker’s site: https://joel-becker.com/ –Timestamps:(00:00) Intro(00:34) Understanding AI evaluation methods(02:04) METR's unique approach to AI evaluation(03:10) The evolution of AI capabilities(06:44) AI as coding assistants(09:15) Research on AI's impact on developer productivity(13:55) Sponsor: Mercury(15:07) Challenges in measuring developer productivity(20:38) Insights from the research paper(31:26) The formalities of software development(32:07) Automated tools and human discussions(32:47) AI and style transfer in software(34:35) The role of comments in AI coding(36:51) The future of AI in software engineering(40:25) Economic implications of AI in software(46:53) Challenges and risks of AI in software(59:03) Security concerns with AI-generated code(01:04:59) Wrap
--------
1:04:04
--------
1:04:04
How Banks Actually Work (And Don't Work)
In this solo episode, Patrick McKenzie reads his classic essay "Seeing Like a Bank," exploring why financial institutions often appear to have no memory of previous customer interactions despite being excellent at tracking money itself. He breaks down the complex web of legacy systems, tiered support structures, and regulatory constraints that create Kafka-esque experiences for bank customers. Using the lens of institutional legibility borrowed from "Seeing Like a State," Patrick explains how banks' technical architecture and organizational design choices—from core processing systems to customer service tiers—systematically generate the dysfunction that customers experience when things go wrong.–Full transcript available here: www.complexsystemspodcast.com/how-banks-actually-work/–Recommended in this episode:Patrick’s Bits about Money essay, Seeing like a bank: https://www.bitsaboutmoney.com/archive/seeing-like-a-bank/–Timestamps:(00:00) Intro(03:52) Recordkeeping systems(10:20) Sponsor: Safebase(11:50) Human accountability and its malcontents(22:57) Two embedded surprises about bank staffing(27:47) Society has goals which conflict with banks being good at banking(30:52) So what can be done about this?
About Complex Systems with Patrick McKenzie (patio11)
We live in a world where our civilization and daily lives depend upon institutions, infrastructure, and technological substrates that are _complicated_ but not _unknowable_. Join Patrick McKenzie (patio11) as he discusses how decisions, technology, culture, and incentives shape our finance, technology, government, and more, with the people who built (and build) those Complex Systems.
New Zealand