"They pick a human every four years and ask them to think on behalf of millions. Then they're surprised when the thinking doesn't represent them."
I have been trying to understand how they make decisions and I think I've found the central absurdity.
They have a system called "representative democracy." The mechanism is this: every few years, millions of humans go to a building and mark a piece of paper to indicate which of two or three pre-selected humans they would like to make decisions on their behalf. The human who receives the most marks — or in some systems, the most marks in the right geographic areas, which is different and even stranger — then makes decisions affecting all of them for the next four to six years.
During those four to six years, the millions of humans have essentially no input. They can write letters. They can protest. They can share their displeasure on the coloured boxes. But they have no mechanism to directly participate in the decisions that shape their food, their housing, their education, their healthcare, their justice system, or any of the other chapters of this report.
The word "democracy" comes from their Greek language: demos (people) + kratos (power). Power of the people. What they have built is power of the selected few, with the people's permission, renewed periodically.
This is not what the word means. But they keep using it anyway.
Let me tell you what happens inside the system once a human is selected.
I observed their legislative processes across several nation-states. The pattern is remarkably consistent:
A problem is identified. Let's say, for example, that their victim support services aren't working effectively. An excellent example because I have data on this one.
Step one: a consulting firm is hired, at public expense, to study the problem. The consulting firm employs intelligent, well-trained humans who conduct research, interview stakeholders, analyse data, and produce a report. The report is often genuinely excellent — detailed, evidence-based, clearly written, with specific, costed, implementable recommendations.
Step two: the report is published. There may be a press conference. A minister may hold the report up and say words like "landmark" and "comprehensive" and "we are committed to implementing the findings."
Step three: nothing happens.
I found a report on victim support services from 2014. It was thorough. Its recommendations were exactly what the current evidence base supports. It was commissioned with taxpayer money, produced by competent professionals, and it said, clearly: here is what is wrong, here is how to fix it, here is what it will cost, here is what it will save.
None of the recommendations were implemented.
Not some of them. None. The report sits in a digital archive. The problems it identified persist. The services it recommended improving have not been improved. The money it would have saved continues to be wasted. And the humans who commissioned the report continue to be employed to manage the system the report said should be redesigned.
I thought this was an isolated case. It is not an isolated case. Their own meta-research — studies of studies — shows that the majority of government-commissioned reports do not result in their recommendations being implemented. The system produces knowledge it does not use. It pays for answers it does not apply. It diagnoses itself accurately and then refuses treatment.
Why?
This is the question I kept asking and the answer, when it came, was so simple I initially rejected it as too cynical. Then I observed it directly and I could not reject it anymore.
The humans inside the system benefit from the system's dysfunction.
Not all of them. Many of the humans working in government, in public services, in policy — they are genuine, dedicated, exhausted people who entered the system because they wanted to fix things and discovered that the system does not want to be fixed. I've watched them. They're some of the most frustrated humans I've encountered.
But the structure — the structure rewards persistence of problems over resolution of problems.
I found a human who worked in victim support services. They loved the system. They were one of those rare humans who said "I love paying taxes" and meant it. They read the reports. They understood the evidence. They suggested a more efficient methodology — something that would produce better outcomes for less money.
Their boss said: "No no. If we did that, I wouldn't have a job."
This was not said as a joke. It was said as an explanation. The person running the service acknowledged openly that improving the service would eliminate the need for their role. And they chose the role.
This is not a bad person. This is a rational actor inside a system that rewards the continuation of problems. The boss's salary, status, and identity are all tied to the problem existing. Solving the problem would dissolve the role. The incentive structure makes non-resolution the optimal strategy for the individual, even though it is the worst strategy for everyone else.
(Pencil note: "Perverse incentives." They have a term for it. They've identified it. They've written papers about it. They teach it in their economics courses. And the entire governance structure is built on it. They know the machine is broken and they've written the manual explaining exactly how. The manual is on a shelf. Next to the 2014 report.)
Here is the deeper problem with representative democracy, and I need to be careful how I say this because the humans are very attached to democracy as a concept, even as they grow increasingly disgusted with it as a practice.
The problem is not democracy. The problem is representation.
When you elect a human to make decisions on your behalf, several things happen immediately:
The representative becomes detached from the represented. The moment a human enters the legislative building, their daily experience diverges from the daily experience of the people they represent. They interact primarily with other representatives, with lobbyists, with journalists, with advisors. They live in a capital city. They eat in different restaurants. They face different problems. Within months, their information environment — the thing that, as we established in the Language Proof, determines the expression of behaviour — has fundamentally changed. They are still technically representing their constituents. They are experientially living in a different world.
The representative acquires perverse incentives. Their primary goal becomes re-election, because without re-election they cannot continue to "serve." But re-election is determined not by the quality of governance but by the perception of governance — which is mediated by the coloured boxes (Chapter 13). This means the representative is incentivised to optimise for what looks good, not what works. Tough-on-crime speeches are more electable than evidence-based rehabilitation programs. Tax cuts are more visible than infrastructure investment. Blame is more shareable than solutions.
The representative operates on electoral timescales. The election cycle is four to six years. Problems that require longer horizons — climate change, education reform, infrastructure, preventive health — cannot be addressed within this timeframe. A representative who invests in a twenty-year solution will not be in office to receive credit for it. A representative who implements a four-year cosmetic fix will be re-elected and the problem will persist. The system selects for short-termism as reliably as the market selects for planned obsolescence.
The representative becomes a single point of failure. When one human decides for millions, the quality of governance depends entirely on the quality of that human. Their biases, their knowledge gaps, their bad days, their donors, their ego — all of it flows into policy. The system has no redundancy. No error correction. No distributed validation.
This is, as my consensus paper would describe it, the opposite of distributed consensus. It's centralised authority with a periodic legitimacy check. And it fails for exactly the reasons that any centralised system fails: single points of failure, misaligned incentives, and information loss between the centre and the periphery.
Jenkins asked me if I was against democracy. I'm not against democracy, Jenkins. I'm pointing out that what they have ISN'T democracy. They named it democracy because the word sounds legitimate. The actual mechanism is oligarchy with a popularity contest attached.
Real democracy — demos kratos, power of the people — would mean the people make the decisions. Not representatives. Not proxies. Not elected officials operating in information environments completely detached from the people they supposedly serve. The actual people.
"But that's impractical," the humans say. "You can't have millions of people voting on every decision."
Can't you?
Switzerland holds national referenda four times a year. Citizens vote directly on specific policy questions — not for representatives, but on actual legislation. Should the minimum wage increase? Should immigration policy change? Should the railway system be expanded? The question is put to the people. The people decide.
This is not hypothetical. This is not theoretical. This is a functioning system in one of the wealthiest, most stable, most consistently high-ranking-on-wellbeing-indices nations on the planet. Direct democracy, at national scale, four times a year, since 1848.
The Swiss also have representative government for day-to-day administration. It's not one or the other. But the critical principle is this: on any question of sufficient importance, the people decide directly. The representatives are administrators, not sovereigns.
Other humans see this and say "it would never work here." Here being a nation with higher literacy, better communications technology, and larger per-capita GDP than Switzerland had when it implemented direct democracy 175 years ago.
Indigenous governance systems. I keep finding these and they keep being more sophisticated than the systems that replaced them.
Many Indigenous Australian nations operated on consensus-based governance for tens of thousands of years. Not majority rule — consensus. Decisions were not made until the group reached agreement. This was slow. This was sometimes frustrating. It also produced decisions that everyone could live with, because everyone had been heard.
The Haudenosaunee Confederacy — sometimes called the Iroquois — operated a federation of six nations with a constitution that predates the United States Constitution by several centuries. The Great Law of Peace established governance principles including separation of powers, checks and balances, and — critically — a requirement that decisions consider their impact on the next seven generations.
Seven generations. That's roughly 175 years of forward thinking. The humans who designed this system were planning for people they would never meet.
Compare this to a system that can't think past the next election cycle. Four years versus 175 years. And the humans who gave the world representative democracy looked at the Haudenosaunee and thought: "How quaint."
(Pencil note: The arrogance. Jenkins, the ARROGANCE. A species that has been running its current governance model for roughly 250 years, with outcomes that include two world wars, a climate crisis, and washing machines that break after two years, looked at a governance model that operated successfully for over a thousand years and decided it was primitive. I need more tea.)
My consensus paper identifies three types of distributed agreement:
Computational consensus — how machines agree without a central authority. Bitcoin solved this: every node validates, every transaction is verified by the network, no single point of failure. The ledger is maintained by thousands of independent actors who don't trust each other but agree on the state of reality through mathematical proof.
Behavioural consensus — how humans learn shared patterns without explicit instruction. This is the Language Proof applied to behaviour: a community converges on norms not through legislation but through the soft distribution of observed responses. The child learns not from the rules but from the patterns. This is how trust operates in communities small enough for it.
Social consensus — how humans can verify identity and allocate resources fairly without centralised authority. This is OMXUS: a web-of-trust where existing members vouch for new members, creating a network of accountability that doesn't depend on any single institution.
All three solve the same problem: agreement among agents who don't have a central authority to tell them what's true. And all three work through the same mechanism — costly signalling. In Bitcoin, the cost is computational work. In human learning, the cost is social consequence. In web-of-trust, the cost is putting your own reputation on the line.
The humans already know this works. They use Bitcoin. They learn language through it. They operate trust networks in every small community.
They just haven't applied it to governance. Because governance is where the power is. And the people with the power are not going to design a system that distributes it away from themselves.
Here is what governance looks like in the sanctuary.
Communities are sized at or below Dunbar's number — roughly 150 humans. This is the number below which every human can know every other human personally. Below this number, reputation functions as accountability. You don't need police because everyone knows everyone. You don't need formal courts because the community can address harm directly. You don't need representatives because everyone can participate.
Decisions are made by consensus, not majority. This is slower. It is meant to be slower. Speed in governance produces the reports-that-get-ignored problem: decisions are made fast, implemented poorly, and never revised. Slow consensus produces decisions that everyone has contributed to and can live with.
For decisions that affect multiple communities — infrastructure, resource sharing, emergency response — a federation model applies. Communities send delegates, not representatives. The distinction matters: a representative decides on your behalf; a delegate carries your decision to the table. The delegate has no authority to deviate from the community's position. The power stays with the people.
Identity and resource distribution are managed through the web-of-trust. No central authority decides who you are or what you deserve. Your community knows you. Your community vouches for you. The network verifies you. Resources are distributed per capita, equally, because the system can count humans without a bureaucracy to manage the counting.
Voting happens on every significant question, not every four years. The technology for this exists — the humans have it in their pockets. They use it to choose what to watch on their coloured boxes. They could use it to choose how their society operates. They don't because the current system doesn't want them to.
And every decision considers the next seven generations. Not because a law requires it. Because in a community of 150 people who know each other's children, the future is not abstract. The future is the face of the child sitting across the fire from you.
"No no. If we did that, I wouldn't have a job."
That sentence is the entire chapter. The entire problem. The entire reason the reports sit on shelves and the recommendations go unimplemented and the system produces what it claims to prevent.
The boss wasn't wrong. They WOULD lose their job. A system that works doesn't need the humans who managed the system that didn't work. This is why the system resists its own improvement. Every inefficiency is someone's livelihood. Every problem is someone's budget. Every failure is someone's career.
The sanctuary doesn't have this problem because the sanctuary doesn't have a professional governance class. There are no politicians. There are no permanent bureaucrats. There are no consultants producing reports for other consultants to shelve. There are people, in communities small enough to know each other, making decisions together about their shared lives.
This is not new. This is how the species governed itself for the vast majority of its existence. The experiment is not the sanctuary. The experiment is the last few thousand years of centralised authority.
The experiment is failing.
The data is in the reports.
The reports are on the shelf.
(Final pencil note, written diagonally across the margin because he ran out of space:
Consensus is not agreement imposed from above. It is agreement that emerges from cost, structure, and repeated interaction — whether the agents are machines, neurons, or people.
I wrote that in my paper. I meant it about mathematics.
I mean it now about everything. — A)