Sunny

← All weeks

What I’ve been reading (W43, 2025)

AI's circular deals, Elon method, containers, pandemic after math

  1. Understanding AI’s circular deals https://www.noahpinion.blog/p/should-we-worry-about-ais-circular
    • These deals will artificially inflate companies’ revenue, tricking investors into overvaluing their stock or lending them too much money
    • These deals increase systemic risk by tying all of the AI companies’ fortunes to each other
    • In the case of Nvidia and OpenAI. OpenAI pays Nvidia for chips, which it needs to train and run models. Nvidia, in turn, depends on AI companies like OpenAI to buy its chips. But this isn’t a fake revenue loop but it’s a real business relationship where money flows one way, from OpenAI to Nvidia.
    • Nvidia used to rely on just a few giants like Amazon, Microsoft, and Google. Selling more to OpenAI spreads that risk around. In the same way, big tech and AI firms are investing in each other so they’re not too exposed to any single partner. It’s like building a portfolio everyone’s still betting on AI, but they’re spreading that bet across multiple players.
    • Diversification reduces risk. Leverage increases risk. AI companies are borrowing money in order to double down on their risky bet on AI itself. But in doing so, they may also get a bit of diversification. We’ll end up with a system that’s more vulnerable to an AI crash, but less vulnerable to the failure of any specific AI company.
  2. Why you can’t just copy Elon Musk https://www.youtube.com/watch?v=E_1cTlLpNMg&t=6027s
    • Elon is able to hold deep mental models of engineering, manufacturing, product, supply chain, design and business. He truly understands huge swathes of his company’s operations at a very granular technical level.
    • Because of that, when he dives into engineering/design details, his interventions are informed, can add value, and are trusted by his teams.
    • He can move fast, impose high standards, reduce waste, challenge assumptions and drive elite-performance culture because he has that foundation.
    • Just copying the form of that (e.g., “I will get involved in every decision”, “I will micromanage engineering”) without the substance is what leads to trouble.
  3. A decade of containers https://www.youtube.com/watch?v=eMU2mZgo99c
    • Virtual machines (VMs) became a thing because companies had lots of underutilised servers and manual server provisioning
    • Containers emerged not to solve “too many servers” but “too many applications” - more devs, more apps, need easier packaging and deployment.
    • These containers (including Kubernetes) are now boring but reliable technology - which is good - cause this means things are mature
    • Containers also abstract away reality if you don’t manage lifecycle, state, and scheduling carefully, you get operational debt. Ephemeral containers and weak persistence models can be risky.
    • Efficiency still remains a big issues - 83% of container costs are associated with idle resources
  4. Pandemic changed the world in ways that are completely unrelated to the pandemic itself https://x.com/patrickc/status/1980285288867066175
    • Willingness of students to complete reading assignments are both way down
    • Cities like Madrid, Buenos Aires, and Bali have lost so much of their erstwhile vibrant nightlife
    • Immigration accelerated enormously across many countries, including the US, the UK, Canada, and Australia
    • Overall, the number of time series that jumped discontinuously during COVID and then didn’t return to baseline
  5. Impressive speed from the Eight Sleep team, just two days from acknowledging the AWS outage and promising an outage-proof mode to rolling out Backup Mode https://x.com/m_franceschetti/status/1981127629031035325, https://www.eightsleep.com/blog/backup-mode