The bench was officially out for the season.

We came to that conclusion on a Sunday in early February, after sitting down for less than four minutes before Dev's hands began turning a color I had not previously associated with hands. Moolah, who appeared genuinely unbothered by the cold but who had taken pity on us, suggested a kissaten she said she had been visiting since the seventies.

I did not ask which seventies.

The kissaten was on the second floor of a narrow building two blocks from the park. Wood paneling, jazz records, an old man behind the counter who nodded once at Moolah and twice at us and then poured three coffees without being asked. None of us said anything. There are some places where you understand, immediately, that questions would be a kind of impoliteness.

We sat at a small table by the window. Snow was starting to fall outside, lightly, the kind of snow that hangs in the air without quite committing to landing.

"Today," Moolah said, "we talk about staking."

"Like, locking up tokens to earn yield?" I said.

"Yes. But it has gotten complicated, in the way things get complicated when humans get involved with a good idea. There are now three different things people call staking, and they are stacked on top of each other like a set of nested dolls, and most of the people using them do not understand what they own. So we are going to take them apart, one at a time."

She produced - from somewhere - three sugar cubes from the kissaten's sugar bowl and arranged them on the table in a row.

"First sugar cube. Native staking. This is the original. Remember Proof of Stake? To become a validator on Ethereum, you put up thirty-two ether as a deposit. The network randomly picks you, sometimes, to write the next page of the notebook. If you do your job, you earn rewards. If you misbehave, the network slashes your deposit. That's boring old native staking."

"And the rewards are…?"

"Currently around three to four percent a year, paid in ether. Comparable to a savings account, except instead of a bank promising you the money, the protocol pays you for keeping the network honest."

"What's the catch?"

"Three catches. First, thirty-two ether is mid-five, sometimes low-six figure dollars. A high bar. Second, your ether is locked up - you can't trade it, can't spend it, can't use it as collateral while it's staked. Third, you need to run a computer that's online twenty-four hours a day, or you lose money slowly through inactivity penalties. The protocol pays validators to validate and punishes them for sleeping."

"So most people don't do this directly."

"Most people don't. Which brings us to the second sugar cube."

She pushed the second cube forward.

"Liquid staking. Some clever person, a few years ago, said: what if we let people pool their ether, run the validators on their behalf, and give them a receipt for their deposit? A token that represents their share of the staked ether. Then they can trade the receipt, use it as collateral, do whatever they want with it. Their underlying ether is still staked and earning rewards. But the receipt itself is liquid."

"That seems useful."

"Hugely useful. The biggest example is called Lido. They run validators on behalf of millions of people, and the receipts they issue - called stETH - trade almost identically to ether. So you get the staking rewards and you get to use the position. You get the yield without giving up the liquidity."

"Where's the catch?"

"There are several." She picked up the second sugar cube and turned it slowly. "One - you're now trusting Lido. They are a particular organization, with particular validators, and a bug or attack on their system could affect your deposit. Two - the receipt is similar to ether but not identical to ether. In a panic, the receipt can trade at a discount. People who needed to exit fast in stressful moments have sometimes had to take five or ten cents on the dollar of haircut. Three - and this is subtle - when one liquid staking provider gets too big, it starts to look like a centralized point of control over the network they're supposed to be securing. There is an ongoing debate about how big is too big. Lido alone runs about a third of all Ethereum validators. Some people lose sleep over this. Some people don't. Both sides have a case."

"Okay. Third sugar cube."

"Ah." Moolah gave a small smile. "Now we are in the strange country."

She pushed the third cube forward.

"Restaking. This is where the engineering gets too clever and the metaphors start breaking down. Imagine you've already done liquid staking. You have stETH - your receipt for staked ether. Someone says to you: give me your stETH, and I'll let other protocols use the security of your stake to underwrite themselves. In exchange, those protocols will pay you additional rewards."

"What does that even mean?"

"It means your one stake is now backing more than one thing. Your ether is securing Ethereum, as it always was. Your stETH receipt is now also being used to secure - say - a new oracle network, or a bridge, or a layer-2 chain. If any of those services misbehave or get attacked, you could be slashed. Your one deposit is now exposed to the failure modes of multiple systems. In exchange, you earn a yield from each."

"That sounds…"

"It sounds like turtles all the way down. And it is. The biggest restaking protocol - EigenLayer - has at times held many billions of dollars of stake being lent out, simultaneously, to dozens of services. Each layer adds yield. Each layer adds risk. The math becomes genuinely difficult. The risk is no longer just will Ethereum fail - it's will Ethereum fail, or Lido, or the restaking protocol, or any of the dozen services restaking has been lent to. Any one of them slipping can drag your deposit down."

I was writing furiously. Native: 32 ETH, run a node, slow yield. Liquid: pool with a receipt, trust the operator. Restaking: receipts being lent to other systems, layered risk, layered yield.

"The receipt for the receipt for the receipt," Moolah said. "And there are now restaking-of-restaking protocols. Receipts for the receipt for the receipt for the receipt. Each one promises a little more yield. Each one adds a little more counterparty risk that most users cannot see and would not understand if shown."

"Why does anyone do this?"

"For the same reason they do everything else. Because the yield is real, and because the risks are abstract, and because the second is much harder to feel than the first. The yield arrives every day in your wallet, a small comforting number. The risk only arrives once, all at once, on a Tuesday you didn't see coming."

Dev had been quiet. He stirred his coffee. There was, unbelievably, no debris in it for once.

"I tried to explain restaking to a friend last month," he said. "I couldn't."

"Because you don't understand it," Moolah said, not unkindly. "Nobody fully does. The people building it understand the moving parts. The people using it understand the yield. Almost nobody holds the full picture in their head. That should make you cautious. The history of finance is mostly a history of products that were too clever to be safely held by their own makers. I once watched a banker in Vienna, in 1873, explain a structured product to a count, and I could see the count nodding politely at things the banker himself was getting wrong. They both lost a great deal of money the following month."

She did not, as usual, elaborate on Vienna in 1873. I had stopped asking.

"So what should I do?" I asked.

"With staking? At your scale, very little. If you have ether you don't plan to touch for a year, native staking is fine. If you want some flexibility, liquid staking with a major provider is acceptable, with the understanding that you've taken on a counterparty. Restaking, for now, I would leave to the people whose job it is to model these things and who get paid to be wrong sometimes. Yield that you cannot explain in a paragraph is yield you cannot hold safely. That is the rule. Write it down."

I did. Yield you cannot explain in a paragraph is yield you cannot hold safely.

The snow outside had thickened. The old man behind the counter put a record on - something old, slow, full of brass - and refilled our coffees without being asked.

"Same time next week?" Dev said.

"Here," I said. "Not the bench. Not until April."

Moolah lifted her cup, agreeing. The three sugar cubes still sat on the table between us, in their stacked row. She had, I noticed, not used a single one in her latte.

Next: Yield - If It Sounds Too Good to Be True