“The market is a pendulum that forever swings between unsustainable optimism and unjustified pessimism.

The Intelligent Investor is a realist who sells to optimists and buys from pessimists.”

Benjamin Graham

So:
- If it is rewarding, do it -> Someone will game the system
- Game the system -> Follow the gradient
- Gradient reaches the local minima - > collapse the system because proxy minima reached
- The optimizer figured out the 'surrogate intent' even if the designer's true intent was something else

Then:
- Essentially tokenomics or economics or even AI alignment in general works (unless value extraction before system collapse is your objective) only works when the 'intent' is opaque or does not have a uniquely global minima
- You need to setup a non-convex problem, where the intent is explicitly unlearnable

So the goal should be:
- Don't make the intent fully learnable, ever

memebuilding • prev manifold.xyz, twilio.com

was thinking about something around this

https://warpcast.com/july/0x85f7eb07

creating and destroying; flying cars, roc.camera, july.rocks

Nice framing; read a bit about convexity.

First example that came to mind was fast fashion like Zara and h&m.

Here’s the deal:
- Reality isn't linear but systems behave linearly around operating points 
- it’s turtles all the way down and the turtles all have Jacobians on their backs
- Convexity is a hammer, everything is a nail
- Linear Transformations Rule Everything Around Me
- Everything is a convexity problem if you squint hard enough 
- Yes, everything

Yeah convex optimization is the pinnacle of applied math / engineering

Yeah convex optimization is the pinnacle of applied math / engineering 

https://warpcast.com/july/0xb06757e4