Skip to main content

All In Podcast Episode 213 Recap and Notes January 31 2025

January 31 2025

  • Friedberg episode with Ray Dalio

    • New book on how countries go broke
  • Travis Kalanick joins the show

    • Cloud Kitchens
    • Uber
  • Cloud Kitchens intro

    • Future of food
    • Bowl builder
    • Cooks prep, machine plates
    • Tie in to health
    • Friedberg did the same thing with ~Eatsa? ~10 years ago
      • Used a canister model
      • He got sick, company closed
      • Had conversations with Chipotle, Sweetgreen, etc
      • 800 sqft restaurant doing $3M a year
    • NY had automats in the early 20th century
  • Sacks is here!!!!

    • Has an office in the Eisenhower building
    • Has a pass to the West Wing
    • Can take people to lunch at the Navy mess
  • Deepseek

    • New startup disrupting AI space
    • Sacks new role lets him get access to all the big people in the space
    • China vs US
    • Open source vs Closed source
    • 2 kinds of models
      • LLM
        • 4o
      • Reasoning
        • o1
    • Final training run cost vs full training cost is different
    • Some credible reports that Deepseek has extensive training cluster which isn’t exactly cheap
    • If you’re on the outside guessing, you have your own incentive for the narrative you’re embracing
    • Regardless of speculation, it seems they have used novel techniques
    • Balaji - It might turn out that the GPT wrapper has more of a moat than GPT
    • Microsoft is also hosting R1 - what does that do to their relationship with OpenAI?
    • Was there distillation?
    • Security implications on calls to the models?
    • Lot of pressure on Meta to show its better than R1
    • Lot of pressure on Open AI with so many external pressures
    • Embrace and extend
    • Does this hit Nvidia?
      • Apparently they bypassed Cuda?
    • This could herald the time of amazing AI usage explosion
    • Jevons paradox
    • When you’re behind then OpenSource is a good strategy to catch up
  • US vs China in AI

    • China has epic intensity of copying based on Travis’s experience with Uber in the past
    • But after they are done copying, they move to innovation and creativity - and you can’t compete with that - they are light years ahead right now in terms of delivery, etc
    • Same thing might happen in AI
    • Is Singapore being used as a backdoor to funnel NVidia chips to China?
    • Export controls could have bad second order effects
  • OpenAI fund raise

    • ~$340B valuation?
    • Taking money from Masa can be a double edge sword
      • Getting the money is good - then your competitors don’t get it
      • But diligence means they might invest in competitors also
    • Smaller models could have a good future
  • DOGE first 10 days

    • Claiming to save $1B/day

    • Generous buyout kinda deal offered to government workers - might save another $100B

      • Bill Clinton did the same kinda buyout during his Presidency - he did an amazing job
    • Terminating leases to buildings, etc

    • RTO

    • Doge eggplant?

    • All this was announced a while ago

    • Dalio book shows we need to cut about a $1T in spending to get the government spending under control

    • Interest rates might go down

    • The faster you cut spending, the less you have to cut

    • Real estate owners will be hit by Government cutting leases

    • Amazing how little real negative impact has happened - so cool

    • 3 layer onion

      • People
      • Infrastructure
      • IT and services and spend
    • 2019 spend vs 2024 revenue would be $500B surplus!!!

    • Need 2 deflationary things

      • DOGE
      • AI benefits
    • Support from people

    • Border policies are also popular

  • Self driving - Waymo, Uber, Tesla

    • Cheap AI will enable cheap autonomy
    • For CA to go full electric vehicles - crazy amount of power generation will be needed
    • Might need combustion engine AVs
    • There is a huge real estate play, electrification play here - fleet management
    • Robots need actuators need magnets need rare earths
    • Car ownership could drop
      • Parking lots could drop
      • 10x fewer cars, 1/10th the land needed for parking
      • 20% of the land in the city goes fallow
  • Interest Rates

    • Might go up before they go down
    • DOGE needs to work harder super faster
  • Fatal DC plane crash 😟

    • Need some upgrades to these systems - very sad situation 😟

Comments