Palm Springs (2020) and learning loops [SPOILERS]











Wife and I watched Palm Springs over the weekend. The plot is basically:

  • Andy Sandberg is stuck in a time-loop where he relives the same day over and over

  • Cristin Milioti gets stuck in the loop with him

  • Andy has given up and is just sorta nihilistically fucking around in the time loop

  • Cristin decides she needs to get out and studies quantum physics every day until she devises a plan to bust out of the time loop

I see it like this. At t=0, Andy knows some about the time-loop and Cristin knows nothing. She catches up to his level of knowledge. She acclimates to his mindset and stays relative static with him. Then she gets fed up and decides to take advantage of the infinite time loop to learn rapidly. She breaks through a threshold that allows her to break them both out of the infinite loop.

While we don’t have the luxury/hell of an infinite time-loop in our lives, I thought the movie was a cute way to reinforce growth-mindset and tight feedback loops.

She takes her learnings from day(n) to day(n+1) and so on, realizing compounding earnings that she can use in the next day via a sorta learning loop. We don’t get to repeat the same exact day, but we do get to bring learnings from t(0) to t(n+1). The shorter that time interval, the more we can compound learning.

Going back to the movie, Cristin’s learning loops (1-day) were much tighter than Andy’s (potentially thousands of days).

By taking advantage of the learning opportunities afforded by your circumstances, you can break out of whatever time-loop you’re in. Be like Cristin!

YFI Ponzinomics

How things get designed to only go up

HUGE DISCLAIMER: this is not a recommendation to buy YFI. I hold YFI but it’s not a substantial position.

This is going to be sorta rushed because it’s my wife’s birthday (everybody wish @annmhairi happy birthday!) and I need to get this out within 20 minutes. So I’ll keep it short and sweet.

I’m not sure who to credit with the term “ponzinomics” but it’s a brilliant term for the financial engineering that’s happening in DeFi. There are good and bad things about this practice but treat my take today as completely neutral and just amused by its impact on the behaviors of everybody in the system.

I interpret ponzinomics as setting up systems such that the price of the coin cannot help but go up.

YFI offers a few examples of how this is possible. In short, ponzinomics induce sell-side liquidity crises. This means there’s not enough supply on the red side of the order book so every marginal dollar entering the system keeps pushing the price up and up.

How do you do this? Reduce supply and increase demand duh. Here are three ways this played out with YFI.

1. Limit circulating supply

Most coins have at least two of the three supply drivers in the figure above: premine, initial issuance, and ongoing inflation.

  • The premine typically gets allocated to the team and early investors and can be moderated by vesting schedules.

  • Initial issuance is what is available to the early community of investors and users whether that’s through yield farming or a crowdsale

  • Ongoing inflation is what you earn from staking or using the platform on an ongoing basis

Do the exercise on your favorite coins to get a sense for this.

YFI just had an initial issuance of 30K coins distributed over seven days to liquidity providers. Because no coins were on the market, the only way to earn them was to farm them. The sellers in the market were limited to those that had already farmed them.

2. Drive demand

No supply doesn’t matter if there’s no demand. I simplify demand drivers into three categories:

  • Invest: holding the coin over a long time horizon because you believe it will appreciate due to growth of the network

  • Trade: holding the coin over short time frames because you think you can make money on short term volatility

  • Work: holding the coin to do productive work in the system to earn returns

YFI is mostly being acquired to work. This is an oversimplification and over time there will be more investors and traders in YFI markets but because YFI can be staked to earn rewards from the iearn ecosystem, users have a strong incentive to park their YFI to passively earn from the network and if they care enough, use the YFI to govern the network. (Governance of YFI is fascinating topic and something I’d like to talk about another time).

3. “Rigged” pools

So yea less supply plus more demand = higher price… bruh you’ve discovered supply and demand I’m unsubscribing.

Okay but it all comes together in “rigged” systems that continuously lock up supply and drive demand.

For YFI, the best example is the 2nd farming pool where users could deposit 98% Dai and 2% YFI to farm YFI. Because there was no YFI on the market, people were depositing tens of millions of Dai into the pool, which automatically converted it to YFI (effectively buying YFI at the market price). A 98/2 mix maximizes this effect (and is the most inbalanced the Balancer protocol allows).

Users are buying YFI to farm YFI. I’d venture a guess that at least half of the Dai entering that pool did not consciously realize they were “buying” YFI when staking.

These mechanics are at play with assets like SNX too. Large rewards in minting sUSD by depositing SNX simultaneously reduces SNX supply and drives SNX demand.

What’s next?

For YFI, I hope we see one of the most exciting experiments in community governance in blockchains history. It certainly feels like that’s possible.

For other protocols, I expect a lot more focus on cracking ponzinomics for their systems.

I’d be remiss if I didn’t mention risk. There’s a lot of it. In addition to the usual suspects, these systems are extremely reflexive. The same things that make prices gap up when things are chugging will make things gap down when the unwind. So good luck have fun but be safe.

Readers: what did I miss about YFI’s ponzinomics? What are the other great examples live today? What hasn’t been done that’s going to blow peoples’ minds?

Censorable assets are toxic collateral

For the first time ever, a balance of USDC has been frozen by the creators of the asset, the CENTRE consortium. (Worth mentioning that USDT has frozen 22 addresses in total).

I wrote about this possibility almost two years ago in “Use regulated stablecoins, get censorship.” And over the course of the next few months wrote three other pieces on the properties of USDC/GUSD, DAI, and USDT (unfortunately behind the old paywall).

Properties of stablecoins

This figure basically sums up my thinking from that time:

The fiat-backed stablecoins on the market were censorable, but at least USDC and GUSD were regulated and thus held verifiable reserves. But if you needed non-sovereign money tracking the value of USD, then you need to go with something like Dai.

My three positions were:

(1) USDT demand would go away because it's neither uncensorable or verifiable.

Regulated stablecoins are less risky than USDT for users that don't need a censorship resistant store of value

Censorship resistant stablecoins are less risky than USDT and Regulated stablecoins for users that need censorship resistance

So that leaves you with USDT in the awkward middle where there's no market.

Wrong so far. USDT demand has ballooned. I continue to be surprised that Asian traders favor USDT to other stablecoins.

(2) Cautiously supportive of USDC adoption because I thought it could convert speculators on Coinbase to users of web3:

Do we want Coinbase to successfully onboard their millions of users to USDC? As I've writtenbefore, a programmable and stable money is a boon to crypto adoption as long as we remember that it's censorable. Coinbase has the opportunity to increase the probability that any one of their users goes from crypto speculator to crypto user--the most important step in the funnel--so I guess I'm cautiously cheering them on.

Directionally correct. USDC demand has also ballooned and has become a core piece of DeFi and what seems to be a primary on-ramp for users.

(3) Only censorship resistant stablecoins could compete for non-sovereign money

This one is too soon to know but the majority of today’s users seem happy using censorable fiatcoins. We’ll have to see if this changes as awareness of censorability increases.

Stablecoins as collateral

The biggest thing that has changed since I wrote those posts is DeFi. Back then, people were using stablecoins almost entirely as a reserve currency for trading on centralized exchanges. Now stablecoins are being pooled and lent and leveraged through a complex system of smart contracts.

Stablecoins are the biggest form of collateral in what should be “trustless” systems.

Bad assets were one of three risks I outlined in my thread on yield farming which I described as “the value in the asset itself gets robbed.”

The toy example I gave was a fiatcoin that wasn’t fully backed by reserves. Getting your balance frozen is another. But what was difficult to articulate at the time was how somebody else’s bad asset could lead to your losses due to the interconnectedness of DeFi.

Here’s a toy example of a hypothetical pool of stablecoins.

The user deposits USDC, USDT, and DAI into the pool. We later learn that a horrible criminal has also deposited USDC into that pool and as a result, the USDC in that pool gets frozen. When the user goes to withdraw their funds, they don’t get their USDC back. Depending on how the pool is constructed, they might not get anything back!

This example isn’t likely to happen (it’s a toy example!), but I wanted to illustrate how unsafe collateral can take down entire link of DeFi. And given how DeFi is a network of links, taking down one link could lead to failures in many other places.

Be careful out there!

Unblocking creative output with technology leads to category defining platforms

What mediums can blockchains and smart contracts unlock?

Here’s a theory: for any given medium, there are moments in time where the cost to create gets sufficiently low to onboard a huge wave of new creators. It’s easy to see this with the big social networks of the last couple decades:

We’re seeing this in every medium, with the hottest battlegrounds in streaming video, mobile video, and (not pictured) long form audio and streaming audio.

So my question is: where do blockchains and smart contracts fit into this framing? What are the relevant mediums of content?

Markets are an obvious one. Because smart contracts can be deployed permissionlessly and users can trade with one another without needing permission from financial institutions or governments, anybody can create a market now.

  • Status quo: NYSE, CME

  • Disruption: Uniswap, Binance

  • New normal: a market for everything

What else? What are the underrated opportunities? Leave a comment with your thoughts.

Crypto use-cases by liquidity and throughput

I’m working on a post gaming out the scenarios for DeFi to scale and as I was writing it I came up with this 2x2 exploring the properties required for a few common crypto use-cases. It doesn’t fit in that post but I’m curious what you guys make of it so posting it here.

In the figure, I suggest that certain use-cases are only attractive once sufficiently high throughput or liquidity has been achieved.

Speculating: buying something very early speculating on its possible success down the line does not require high liquidity or throughput. When I GPU-mined, I did not expect a market to emerge for months or years (much less a liquid one). Or when investing in startups, VCs don’t expect to transact until there’s a “liquidity event.” This is probably the direction yield farming is going.

Gaming: game economies tend to have huge catalogues of assets, most of them very illiquid. Because the assets aren’t often used as investments, players are tolerant of slippage but they want responsiveness because latency takes them out of the experience.

Investing: when investors deploy capital to investments, they are expecting those investments to play a specific role in their portfolio (most commonly beat an index). They also need to manage risk, which makes liquid assets more attractive because they can get in and out of the asset easier if something changes. (This is in large part why the biggest institutional investors talk about BTC and don’t care about anything else). But because investors are buying and holding until conditions change, they do not need high throughput as they do not plan to get in out with high frequency.

Trading: traders want liquidity and throughput to decrease their costs and increase their profits. Low liquidity means slippage and low throughput tends to mean higher transaction costs not to mention risks that the prices get away from them as their transactions are confirming.

Lots of implications you can start to draw from this chart if you buy the premise. Curious what you all think. Leave a comment with your feedback and predictions.

Loading more posts…