The first real cold snap came the week we talked about consensus.

We were back at the bench, bundled in scarves and layers, and Dev had brought a thermos of something hot. Moolah was, for the first time I could remember, not holding a latte or tea. Instead she had a small paper cup of something that steamed gently in the air. Miso soup, I realized after a moment. She sipped it with both paws.

"Priya. A question."

"Yes."

"When I tell you the shared notebook is verified by thousands of strangers, what have you been assuming about how?"

I thought about it. "Honestly? Nothing. I just trusted you that it worked."

"A fair answer. But today we're going to pull the lid off. Because how a blockchain decides which new page to add - who gets to write, and who checks the writing - is maybe the deepest question in this whole field. It's called consensus, and it's the reason any of this works at all."

"Okay."

"You have thousands of computers, scattered across the world, none of whom trust each other. They all have a copy of the notebook. Every few seconds, a new page of transactions needs to be added. Which computer gets to write it? If they all try to write at once, you get chaos. If you pick one, you've centralized the whole thing and ruined the point. So the question is: how do you fairly and safely pick a winner, over and over, forever, without any central authority?"

Dev blew on his thermos. "I remember when I first learned this, I thought it was impossible."

"Most people did," Moolah agreed. "And then someone figured it out twice. There are two main approaches in widespread use. They solve the same problem very differently. The older one is called Proof of Work. The newer one is called Proof of Stake. Bitcoin uses the first. Ethereum used to but switched to the second some time ago. The switch is one of the most interesting events of the last decade, and we'll get to that."

"Proof of Work first."

"Proof of Work. Imagine I gather all the computers and say: every few seconds, I am going to ask you a riddle that takes enormous computing power to solve. It's a pointless riddle, with no meaning beyond being hard. The first computer to solve it gets to write the next page of the notebook - and gets paid for doing so."

"A bid."

"A bid with electricity. The more computing power you have, the more guesses per second, the more likely you are to win. And because solving the riddle takes real physical resources - electricity, cooling, hardware - cheating is expensive. If a bad actor wanted to rewrite history, they'd need more computing power than half the rest of the network combined. Possible in theory but more than unlikely in practice."

"And that's what Bitcoin does."

"Still, today. Which is why Bitcoin consumes as much electricity as a midsized country. The security is real, but the cost is enormous. Warehouses full of specialized computers, running twenty-four hours a day, solving meaningless riddles, burning power. The joke - a dark one - is that Proof of Work is the world's most elaborate way to convert electricity into trust."

I wrote that down. Electricity into trust. It was a bleak phrase, but I liked it.

"Proof of Stake takes a different approach. Instead of asking computers to burn electricity to earn the right to write, it asks them to put up money as a deposit. To become a validator - a note-keeper - you lock up thirty-two ether. Then, every few seconds, the network randomly chooses a validator to write the next page. If they write an honest page, they earn a small reward. If they try to cheat - write a false page, or try to write two different pages at once - the system takes their deposit. That thirty-two ether is simply gone."

"Burned?"

"Burned. Slashed. The jargon is 'slashing' and it's as brutal as it sounds."

Dev was nodding along. "It's like a security deposit on an apartment. You pay it upfront. If you trash the place, the landlord keeps it."

"It is like a security deposit on an apartment," Moolah agreed, "except the landlord is code, and the rules are public, and if you commit fraud the deposit vaporizes instantly, visible to everyone, with no appeal. A jury of millionaires who lose their own money if they vote wrong. They tend to vote right."

"And Ethereum switched to this."

"In September 2022. An event called the Merge. It was, in some sense, the largest engineering project in the history of open software - swapping out the engine of a system worth hundreds of billions of dollars, live, without downtime, without losing a record. I watched it happen from a river in Chiba. I was genuinely nervous for humanity. When it worked, I ate three clams in celebration."

"Why did they switch?"

"Three reasons. Power - Ethereum went from using as much electricity as a country to using about as much as a small town. Security - it's actually harder to attack Proof of Stake at scale, because the cost of attack scales with the value of the network. And accessibility - anyone with thirty-two ether can be a validator. You don't need a warehouse and an enormous amount of electricity. Most validators, today, are running on consumer hardware. In bedrooms. In closets. In server farms, yes, but also in homes."

"And the trade-offs?"

She smiled, or came close to it. "Thank you for asking. Proof of Stake concentrates power with people who already have a lot of ether. Proof of Work concentrates power with people who have access to cheap electricity. Neither is perfect. Both are better, probably, than trusting a single company or country. It is not a question of which system is good. It is a question of which trade-offs you find tolerable."

I wrote in the notebook. PoW: electricity into trust. PoS: deposit into trust. Under it: the system you pick is a statement about what you think is scarce.

Moolah read it and nodded once. She finished her miso. The wind picked up. Somewhere across the pond, a heron stood very still.

"Next week," she said, "tokens. And why the word 'crypto' is doing violence to the English language."