Marcus Kline: Ben, catch — you ready for this one?
Ben Okonkwo: Ready as I'll be — I've been staring at distributed systems papers since Tuesday, so my bar for 'ready' has shifted considerably.
Marcus Kline: Well, here's what got me thinking about this. My nephew — he's seventeen — he came to me genuinely proud. Said he'd switched his crypto holdings from Ethereum's old setup to the new one because it was, quote, 'the green choice now.' And I didn't know what to say to him, because... he's not wrong. September 2022, Ethereum flips a switch — the Merge — and overnight, its energy consumption drops by ninety-nine point nine-five percent. That's not an improvement. That's a phase change.
Ben Okonkwo: Ninety-nine point nine-five — that number still gets me every time.
Marcus Kline: But here's what I couldn't quite explain to him. The thing that went away — the mining, all that computational brute force — that was concentrating power in the hands of a few large mining pools. And what replaced it, Proof-of-Stake, validators locking up capital... that concentrates power in the hands of wealthy validators. So the energy is gone. But did the underlying problem move? Did we solve it, or did we just give it a different address?
Ben Okonkwo: Right — and that's the thing. The mechanism changed. The type of actor who holds power changed. But concentration itself, that didn't resolve.
Marcus Kline: Which pulls us back to something older. Satoshi Nakamoto, 2008 whitepaper — Proof-of-Work wasn't presented as a workaround. It was framed as the solution to Byzantine fault tolerance for a permissionless ledger. A forty-year computer science problem, answered. And now, sixteen years later, we're still... standing in front of the same shape of problem.
Ben Okonkwo: And that shape has a name. The blockchain trilemma — the claim that no system can simultaneously maximize decentralization, security, and scalability. And what we're really trying to work out today is whether that constraint is real, irreducible, mathematical — or whether it's something murkier than that.
Marcus Kline: Whether it's gravity, or just a very stubborn assumption.
Ben Okonkwo: Exactly that — and the evidence, honestly, keeps pointing both ways depending on what you count as evidence. Which is the problem.
Marcus Kline: Then that's where we start.
Ben Okonkwo: But here's where I want to back up — because before we can even weigh those two mechanisms against each other, we have to understand why the problem is hard at all. And I think most people skip that step. So imagine — you and a group of friends splitting a dinner bill over text. Simple enough. Except two of them are quietly lying about what they ordered. And nobody knows which two.
Marcus Kline: And you still need to land on a real number.
Ben Okonkwo: Right — you need a rule that gets you to the correct total even if some of the inputs are lies. And the brutal part is, the more friends you add to the chat, the harder that rule gets to run. More people, more potential liars, more messages to cross-check. That is Byzantine Fault Tolerance in two sentences. A system that keeps working correctly even when some participants are failing — or actively lying.
Marcus Kline: And that's not something someone invented for blockchains. That goes back to 1999, at minimum.
Ben Okonkwo: 1999 exactly — Miguel Castro and Barbara Liskov, Practical Byzantine Fault Tolerance. PBFT. And what they showed is that you can get fast, definitive agreement — not probabilistic, actual deterministic finality — but only if you're working with a small, known group. Everyone in the room is identified. The moment you open the door to strangers, the math gets... okay, 'brutal' is the right word. The communication overhead scales roughly with the square of the number of validators.
Marcus Kline: Squared. So adding participants doesn't just slow things down — it compounds.
Ben Okonkwo: Compounds, yeah. And that's actually — I mean, that's the load-bearing fact that everything else rests on. Castro and Liskov weren't writing about blockchains, those didn't exist yet. They were solving a coordination problem in distributed computing. But Satoshi's 2008 whitepaper essentially says: here's a permissionless version of this, open to anyone, no one identified. And to make that work, you need Proof-of-Work — computational cost as the barrier to lying. Because you can't verify identity, so you make dishonesty expensive instead.
Marcus Kline: So Proof-of-Work is the price of opening the door.
Ben Okonkwo: That's actually a clean way to put it. And here's where the knobs become visible — more participants gives you censorship resistance, genuinely harder to corrupt the whole group. But more participants means more messages to reconcile, which slows agreement. So you want faster finality? You need fewer validators, or you need them burning more energy to prove they're not cheap fakes. And if you reduce the energy requirement — Proof-of-Stake does this — you're now checking capital instead of computation. Which concentrates along a different axis.
Marcus Kline: The bars don't disappear. They just... rearrange.
Ben Okonkwo: Exactly — and PBFT tells you why they have to. These aren't independent variables. The communication complexity, the identity problem, the cost of Sybil resistance — they're coupled. Turning one knob tightens another somewhere else. That's not a design failure in Bitcoin or Ethereum. That's the geometry of the problem as Castro and Liskov actually formalized it in 1999.
Marcus Kline: So the question isn't whether the trilemma is real. The question is whether every blockchain we're looking at is just — a different choice about which knob to loosen first.
Ben Okonkwo: That's exactly the frame — and now the question is, what does each choice actually look like when it's implemented. Because 'a different coordinate on the surface' is one thing to say abstractly, but the mechanisms are genuinely distinct in ways that matter operationally.
Marcus Kline: Walk me through Bitcoin first. The PoW version.
Ben Okonkwo: So with Proof-of-Work, validators — miners — are competing to solve a cryptographic puzzle. First one to crack it gets to write the next block. And the cost of that computation, that's your Sybil resistance. You can't flood the network with fake identities because each identity costs real electricity and hardware. The security is physically grounded.
Marcus Kline: But the finality is... it's never actually final.
Ben Okonkwo: Probabilistic — right. Your transaction isn't mathematically guaranteed irreversible, it's just that as more blocks pile on top, reversal becomes statistically... okay, practically impossible. But 'practically impossible' and 'mathematically done' are different things, and that gap matters when you're designing a payment system.
Marcus Kline: And Ethereum, post-Merge, that gap closes. The mechanism for finality is different now — checkpoint blocks become final once votes from at least two-thirds of all staked ETH are cast. That's a BFT supermajority rule written directly into the protocol. Not statistical safety. Deterministic.
Ben Okonkwo: Deterministic — once that sixty-six percent threshold clears, it's done. Which is Castro and Liskov's model, essentially. Brought into a permissionless context by substituting capital for identity. Lock tokens as collateral, behave dishonestly, and the protocol destroys that collateral — that's slashing. So the cost of lying isn't electricity anymore, it's your stake.
Marcus Kline: Two different bets about what makes dishonesty painful enough.
Ben Okonkwo: Exactly — and I mean, neither one is obviously wrong. They just... they're pricing dishonesty on different markets. Energy versus capital. One fluctuates with the electricity grid, the other with wealth inequality. And there's a version of this conversation, which we haven't gotten to yet, where those two resource bases turn out to concentrate power in ways that are more similar than they look different from here.
Marcus Kline: Now — this strikes me as strange. Because EOS and TRON took the capital-based logic and then... compressed it further. Delegated Proof-of-Stake. Token holders don't validate themselves — they vote for a small set of delegates who do it on their behalf.
Ben Okonkwo: And you gain speed. Real speed. But you've made an explicit choice — you have traded decentralization for throughput. It's not an accident. It's not a design flaw. The community looked at the tradeoff and said, we want a representative system, not a direct one.
Marcus Kline: Representative government rather than direct democracy. And we know what representative government does to participation over time.
Ben Okonkwo: You get efficiency and you get concentration in the same package. Which is — I mean, that's the thing. Proof-of-Work, Proof-of-Stake, Delegated Proof-of-Stake — these aren't iterations toward a solution. They're communities naming their priority and then living with what that costs.
Marcus Kline: But naming the priority and living with the cost — that's the clean version of it. And I think what we haven't pressed yet is whether the cost of PoS is actually smaller, or just less visible.
Ben Okonkwo: That's the one. Mining pools show up in satellite photos — you can literally look at a facility in Xinjiang and say, that's Bitcoin hash rate. Validator capital lives in ledgers.
Marcus Kline: Harder to see doesn't mean smaller.
Ben Okonkwo: Right — and here's where slashing doesn't actually rescue you. Slashing destroys a validator's stake if they act dishonestly. But it does nothing — structurally nothing — to stop a wealthy validator from simply accumulating more staking power over time through legitimate means. The punishment is for bad behavior. It's not a cap on influence.
Marcus Kline: So slashing is a conduct rule, not a redistribution mechanism.
Ben Okonkwo: Exactly — and whether capital-based concentration is actually better than hardware-based concentration, that question is... okay, it's genuinely unresolved in the literature. It's not a settled normative question. It depends entirely on what failure mode you're most afraid of.
Marcus Kline: And now — now there's a paper that says we may not even have the mathematical ground to call this a trilemma at all. Craig Wright, 2025, arXiv:2507.21111. Formally argues the trilemma isn't a proven theorem. It's a heuristic.
Ben Okonkwo: Wait, not just contested — formally disputed on definitional grounds?
Marcus Kline: The argument is that decentralization, security, scalability — none of those terms have rigorous enough operational definitions to actually derive a formal constraint. Without precise definitions, you can't prove a tradeoff. You've just built an intuition and called it a theorem.
Ben Okonkwo: And here's what makes me sit with that uncomfortably — Cosmos and Tezos are both BFT-based PoS chains, right, the kind where block times should be near-constant by theoretical assumption. And empirically, their block times fluctuate. Meaningfully. So the clean model doesn't map cleanly to the running system.
Marcus Kline: Hm. So the empirical behavior breaks the assumption that the theorem would rest on.
Ben Okonkwo: Which doesn't prove Wright's case — I want to be careful there. It might just mean the implementations are imperfect. But it does mean the gap between the theoretical model and the actual system is wide enough that we probably shouldn't be treating the trilemma as settled physics.
Marcus Kline: And that's the thing that genuinely unsettles me. Because if the trilemma is just a frame — a useful story — then every community that sacrificed something in its name, Bitcoin's energy, Ethereum's capital concentration, EOS and TRON's decentralization... they paid a real cost for a constraint that may never have been formally proven to exist.
Ben Okonkwo: Or the constraint is real and we just haven't been rigorous enough about why. Which is almost worse — because it means we've been navigating something genuine without a proper map.
Marcus Kline: And Bitcoin stayed on Proof-of-Work. Deliberately. That community looked at every alternative and said: no. And that's a statement. It's not inertia. It's a revealed preference for decentralization and security, even knowing the energy cost, even knowing the throughput ceiling.
Ben Okonkwo: While Ethereum moved — and EOS, TRON, they moved further. Each one naming a different constraint as the one they'd live without. Decentralization, in EOS and TRON's case, traded explicitly for speed.
Marcus Kline: Nobody stumbled into their consensus mechanism. And I think — I mean, that's the thing I couldn't have said at the start of this conversation with any confidence. But the pattern across Bitcoin, Ethereum, Delegated Proof-of-Stake... it reads less like engineering iterations and more like confessions. What does your community actually believe matters most?
Ben Okonkwo: What I can't settle — and I want to be honest about this — is whether Wright is right. If the trilemma was never formally proven, if decentralization and security and scalability genuinely lack the operational definitions you'd need to derive a theorem... then those confessions were made in response to a constraint that was, actually, never mathematically established. That's the part I can't put down.
Marcus Kline: Hm. A decade of design choices, real costs, real tradeoffs — shaped by something that might be a very powerful story rather than a law.
Ben Okonkwo: Not a fraud — I want to be clear on that. Even a heuristic that isn't formally proven can be genuinely load-bearing. The question is whether the next generation of protocols builds on it because it's true, or because it became infrastructure before anyone checked the foundation.
Marcus Kline: And maybe that's — actually, no, maybe that's the better question to carry out of this. Not 'which protocol wins.' But: for what specific problem, and for whom, is this constraint worth paying? That reframes the whole thing. Satoshi Nakamoto's whitepaper answered one question for one set of conditions. The trilemma, proven or not, was the frame we used to extend that answer everywhere. Maybe it shouldn't be.
Ben Okonkwo: Genuinely unresolved. I'll take that. Good place to stop thinking out loud, I think.
