Daniel sent us this one — and it's the kind of question that sounds like sci-fi until you realize people are already living in it. He's been thinking about what we talked through recently, that Goldilocks problem where the more issues you care about, the less likely any single party represents you. Housing, rental reform, security — pick any three priorities and watch the overlap vanish. His question is: what if the solution isn't finding a better representative, but eliminating representatives entirely? Could you build a pure direct democracy where every citizen votes on every law remotely, from their phone, as legislation comes up? And if so, what function do politicians actually serve anymore?
Here's the thing that makes this not a dorm-room thought experiment — Estonia has been running nationwide i-voting since two thousand five. Eight national elections. Their most recent parliamentary election, twenty twenty-three, saw fifty-one percent of all votes cast online. That is not a pilot program. That is the default for half the electorate.
Fifty-one percent. And Taiwan's vTaiwan platform has processed over a hundred policy proposals through structured digital deliberation. Switzerland has held more than six hundred fifty national referendums since eighteen forty-eight. These are running systems with real data. So the question isn't "could we" — it's "what happens when we do, and why hasn't it replaced representative government yet.
Right, because if the technology works — and in Estonia's case it demonstrably does, zero confirmed security breaches in twenty years of operation — then the bottleneck isn't the voting mechanism. It's something else. Something about how humans actually behave when you hand them direct power.
Which is where this gets uncomfortable. We're going to walk through the real systems — Estonia, Switzerland, Taiwan, the LiquidFeedback experiments inside the German Pirate Party — and what we find is that the technology is the easy part. The hard part is cognition, participation inequality, and the fact that someone still has to write the laws.
That last piece — who writes the laws — is the one that keeps me up. Because voting is downstream of proposal drafting. You can give everyone a vote on everything and still end up with a legislative bottleneck where only a tiny group has the expertise to actually craft coherent policy. Which sounds an awful lot like the system we already have.
The arc here is: look at what's been built, look at what broke, and ask whether the problem we're trying to solve is really about voting at all.
What I think Daniel's really putting his finger on is a paradox that representative democracy has never solved. The more you learn, the more issues you care about, the more nuanced your positions become — and the worse the system fits you. It's a scaling failure that punishes engagement.
A system that penalizes you for paying attention. That's almost impressively broken.
It's not a bug in any given party or candidate — it's structural. If you have five issues you care deeply about, the odds that one party aligns with you on all five are already low. Make it ten issues, they're near zero. The rational response is to either narrow your priorities or hold your nose. Neither feels great.
The radical answer is: stop trying to find a proxy. Become your own proxy. Everyone votes on everything, directly, from wherever they are. And on paper, that solves the alignment problem completely. Your vote on housing policy reflects your view on housing policy. Your vote on security reflects your view on security. No compromise, no party platform negotiation — just you and the ballot.
Which is genuinely seductive. And it reframes the question Daniel asked in a pretty stark way. If we can vote on everything from our phones, what function do politicians actually serve? Are they just an archaic middle layer we've been maintaining out of habit?
Or are they doing something that direct voting doesn't replace? Because that's the fork in the road. Either politicians are vestigial — a pre-digital workaround we no longer need — or they're performing functions that remote voting doesn't touch. Proposal drafting, amendment negotiation, committee expertise, the actual construction of legislation.
The real question isn't "can we build the voting app." Estonia already did. It's whether voting was ever the bottleneck.
Let's start with the one that actually works at national scale. Two thousand five, they roll out i-voting. The mechanism is clever — you authenticate with your national ID card or Mobile-ID, cast your ballot, and here's the part most people miss: you can vote as many times as you want during the early voting window. Only your last vote counts.
Which means if someone coerces you — your boss stands over your shoulder and makes you vote a certain way — you just vote again later from your couch.
That's the anti-coercion design. And it's paired with end-to-end verifiability. Independent auditors can check the entire chain without seeing individual ballots. Twenty years, zero confirmed breaches. Fifty-one percent online turnout in twenty twenty-three. This isn't experimental.
Here's what jumps out at me. Estonia is still voting for representatives. They're not voting on laws directly. They're using digital tools to do the same thing paper ballots do — pick a party, pick a candidate. So the technology solved the convenience problem, not the representation problem.
Right, and that's the distinction that matters. i-voting increased turnout slightly and made voting easier, but it didn't change the structure of representative democracy at all. The Estonians deliberately chose not to go further.
Which brings us to Switzerland, because they actually do go further — and they've been doing it since before electricity. Six hundred fifty plus national referendums since eighteen forty-eight. Citizens vote four times a year on federal proposals, plus cantonal and municipal levels. But here's the key: they don't vote on everything. A proposal needs a hundred thousand signatures just to reach the ballot.
It's a curated filter. The Swiss system doesn't ask citizens to become amateur legislators — it asks them to judge a small set of proposals that have already cleared a threshold of public support. And even then, turnout averages forty-five to fifty percent. Half the country sits out most votes.
When they do vote, the outcomes can be razor-thin. The twenty twenty-one corporate tax reform referendum passed with fifty point two percent. That's a national policy direction decided by a fraction of a percent on a complex tax question.
Which raises the question: how many of those voters understood the full implications of what they were approving? That's not a dig at Swiss voters — it's a cognitive bandwidth problem. Nobody can be an expert on corporate tax law, housing policy, environmental regulation, and foreign affairs simultaneously.
Which is exactly where Taiwan's experiment gets interesting. The vTaiwan platform, launched in twenty fourteen, doesn't ask citizens to vote on final legislation at all. It uses a tool called Pol.is — a real-time survey system that visualizes where opinion clusters form. The process goes: open problem framing, broad public input, the software identifies where consensus exists and where it doesn't, and then that structured feedback feeds into the actual legislative drafting.
The UberX case is the canonical example. Twenty sixteen, Taiwan had no regulatory framework for ride-sharing. Four thousand plus participants on vTaiwan, three months of structured deliberation, and they reached roughly eighty percent consensus on a regulatory approach. That consensus didn't become law directly — it went to the legislature — but it gave lawmakers a clear map of what the public actually wanted and where the red lines were.
VTaiwan isn't replacing representatives. It's giving them better information. Structured deliberation rather than direct voting.
That's the design choice that matters. is doesn't just tally votes — it identifies opinion groups and shows people where they overlap with groups they might disagree with on other issues. The goal isn't winning, it's finding the shape of the possible consensus.
Which brings us to the most ambitious experiment, and the one that gets closest to what Daniel's asking about. LiquidFeedback, developed in two thousand nine by the German Pirate Party. This was liquid democracy in actual practice — every party member could vote directly on internal policy, or delegate their vote to someone they trusted on a specific issue. Revocable at any time. Issue by issue, not election by election.
This is where the data gets brutal. The Pirate Party used it from twenty eleven to twenty thirteen, and what they found was: eighty percent of votes were cast directly by non-delegating users. Only fifteen percent used single delegation. Five percent used multi-hop delegation — passing your vote through a chain of trusted proxies. And here's the kicker: the top one percent of delegates controlled thirty percent of all delegated votes.
The system that was supposed to dissolve concentrated power ended up recreating it. Super-delegates emerged organically, not because anyone designed them in, but because most people don't want to vote on most things.
The Italian Five Star Movement used LiquidFeedback too, and they saw the same pattern. Low delegation rates, power concentration, and eventually they just abandoned it. The technology worked. The humans didn't show up the way the theory predicted.
Across all four systems, we see the same shape. Estonia proves digital voting is secure and convenient — but they use it for representative elections, not direct democracy. Switzerland proves direct voting works — but with a heavily curated ballot, not universal participation. Taiwan proves structured deliberation can build consensus — but it feeds into the legislature, it doesn't replace it. And LiquidFeedback proves that even when you build the perfect liquid democracy tool, most people don't delegate, and power concentrates anyway.
We've got four systems, and none of them actually do what Daniel's asking about. Which raises the question: what happens when you try? What's the pure version look like? And the best test case isn't digital at all — it's California.
The proposition system.
Since nineteen eleven, California has let citizens vote directly on legislation and constitutional amendments. And the twenty twenty ballot is the cautionary tale. The average voter spent ten minutes researching all of them combined.
For twelve laws.
Proposition twenty-two was a twenty-five page ballot measure on gig worker classification. Passed with fifty-eight percent support. Post-election polling found seventy percent of voters couldn't accurately describe what it did.
You had a supermajority of voters approving a law they didn't understand, written by the industries that stood to benefit from it. That's not democracy failing — that's democracy working exactly as designed, and the design is the problem.
This is the cognitive bottleneck in its purest form. Direct democracy assumes citizens have infinite time and domain expertise. In practice, people are busy, complex legislation is long, and the rational response is to rely on heuristics — party cues, ads, the wording on the ballot summary. Which is just outsourcing your vote to whoever wrote the summary.
Which loops us back to the delegation paradox. Liquid democracy was supposed to solve exactly this — you don't have to understand everything, you just delegate to someone who does. But the Pirate Party data showed that eighty percent of people didn't delegate at all. They voted directly or abstained. And the ones who did delegate concentrated power into super-delegates.
Fifteen percent single delegation, five percent multi-hop. That top one percent controlling thirty percent of delegated votes — that's not a bug, it's an emergent property. People who are willing to vote on everything become the de facto representatives. You haven't eliminated politicians, you've just made them unaccountable to elections.
Unaccountable and unelected super-delegates. That's worse than what we have now. At least my terrible representative has to face me every few years.
Then there's the problem nobody talks about: who writes the laws in the first place? Voting is downstream of drafting. In Switzerland, a hundred thousand signatures gets you a federal initiative — but someone still has to write the actual legal text. In vTaiwan, the structured deliberation produces consensus maps, but actual legislation still gets written by lawmakers and ministry staff.
Because writing a coherent law that doesn't contradict existing statutes, that survives judicial review, that has workable enforcement mechanisms — that's a skill. It's not something you crowdsource.
And this is the function representatives provide that direct voting doesn't replace. Not just voting — drafting, amending, negotiating, committee markup, legal review. If you eliminate representatives, you either need every citizen to become an amateur legislative drafter, or you need some new class of professional law-writers who aren't accountable to anyone.
Which is just representatives again, but with worse branding.
Let me hit one more thing that doesn't get enough attention. Estonia's i-voting has never been breached — twenty years, zero confirmed incidents. But the threat model changes when you move from electing representatives to voting on specific laws. If I can coerce your vote on a corporate tax break worth billions, the incentive to coerce goes way up.
The boss in the room problem.
Estonia's solution is clever — vote as many times as you want, only the last one counts. So if your boss forces you to vote a certain way, you change it later. But that doesn't cover every scenario. Domestic coercion, for one. If your spouse or parent controls your device, they can watch you vote and there's no later. It also doesn't cover vote-buying at scale — you sell your vote, the buyer verifies it, you change it afterward, and now you've got a very motivated buyer with a grievance.
If the vote is on something existential — say, a referendum on whether a minority group gets equal rights — the stakes are high enough that coercion, intimidation, and organized manipulation become rational strategies for concentrated interests. Estonia's security model works for representative elections because the stakes are diffuse. Direct votes on specific laws concentrate the stakes onto single binary questions.
Which brings us to the last problem, and in some ways the most stubborn. Every direct democracy system we've looked at shows the same pattern. Older, wealthier, more educated citizens participate at higher rates. Estonia's i-voting actually widened the gap — the sixty-five plus age group votes online at three times the rate of eighteen to twenty-four year olds. Urban voters are fifteen percent more likely to vote online than rural.
You build a system to democratize participation, and it ends up amplifying the voices that were already loudest.
Switzerland shows the same thing. Twenty to thirty percent turnout gaps between high and low education groups on referendums. The people who show up to direct democracy are the people who were already showing up to everything else. You haven't expanded the electorate — you've just given the existing electorate more power.
Which is the uncomfortable through-line here. The technology for what Daniel's describing exists. It's secure, it scales, Estonia proved it. But twenty years of real-world data from multiple systems tells us the bottleneck was never the voting mechanism. It's human attention, human expertise, and the stubborn fact that most people don't want to become amateur legislators.
Where does that leave Daniel? He's got a Goldilocks problem — no party fits — and the radical solution, pure direct democracy, has a twenty-year track record of not actually replacing representative government. The technology works. The humans don't.
I don't think the answer is "give up and accept the broken system." The most interesting thing across all four cases is that the hybrid models outperform the pure ones. Switzerland doesn't ask citizens to vote on everything — it curates. vTaiwan doesn't ask citizens to draft legislation — it structures deliberation and feeds it to lawmakers. Neither is pure direct democracy, and both work better than the California everything-on-the-ballot approach.
The design insight is: filter before you vote. A signature threshold, a deliberation phase, something that separates "this merits a public vote" from "this is a twenty-five page tax bill nobody will read.
For Daniel's specific situation — housing policy, rental reform, security — liquid democracy offers a practical intermediate step that doesn't require rebuilding the entire system. You could delegate your vote on housing policy to a tenant advocacy organization whose expertise you trust, while voting directly on security issues where you have your own informed position. Issue by issue, revocable at any time.
The catch being that no country currently implements this at national scale. But the Pirate Party showed it works at organizational level, and the principle is sound. You're not looking for one party that matches you on everything — you're distributing your political agency across different proxies for different domains.
If you want something actionable right now, at the level where citizens can actually push for change, there are two things. Ranked-choice voting at the local level — it doesn't solve the representation problem directly, but it reduces the spoiler effect that makes the Goldilocks problem worse. And participatory budgeting, which is the closest working model to direct democracy on specific issues. Over fifteen hundred cities globally use it. Citizens decide how to allocate a portion of the municipal budget. Real money, real decisions, real outcomes.
It's not voting on every law from your phone. But it's direct power over a concrete slice of governance, and it works at scale. That feels like the honest answer here. The technology exists for what Daniel's describing. The bottleneck is human. So the smart play is to build systems that work with how humans actually behave, not how we wish they would.
There's a question underneath all of this that I think Daniel's prompt is really driving at, and it's the one we should leave people with. If direct democracy is technically feasible — and Estonia proved it is — but cognitively impractical, then what does a truly representative system look like for the twenty-first century? We've had basically the same model since the eighteenth.
The frustration that got us here is legitimate. The Goldilocks problem is real, and it gets worse the more you engage. That's a design flaw, not a personal failing. But the solution probably isn't abolishing representatives. It's better delegation mechanisms that let us distribute our political attention across issues and experts, rather than concentrating it all in one party every four years.
Which is what we're going to build next episode. Not theoretically — we're going to walk through what a personal voting proxy system would actually look like using existing liquid democracy protocols. How you'd set it up, who you'd delegate to, how revocation works, what the interface needs to solve. A spec, basically.
From Goldilocks to system architect. That's a trajectory I can get behind.
Now: Hilbert's daily fun fact.
Hilbert: In the nineteen thirties, Slovene was one of the few languages that preserved the dual grammatical number — meaning it had separate verb forms for exactly two people doing something, not just singular and plural. If you were describing two fishermen heading out onto Lake Baikal, you would use a verb form that has no equivalent in modern English, effectively converting "they go" into a unit of exactly two.
Slovene had grammatical pair programming.
I have so many questions about what happens with three fishermen.
Just don't. This has been My Weird Prompts. Our producer is Hilbert Flumingtop, and if you want to hear us design a voting proxy system from scratch, that's next time. If you enjoyed this, leave us a review wherever you listen — it helps. We'll be back soon.
