The Three Pillars of Truth: How a True Idea Establishes Itself in an AI World

Status: Original concept, first publication 11 July 2026. Completes the Bayesian Trust series.

Anchor on consensus. Get trusted intermediaries to engage. Build a stable information layer.

Being right gets you nowhere on its own

There’s a version of being right that gets you precisely nowhere: you see the true thing, you say it clearly, and the world, and now the machine, carries on believing what it believed before. Being right is necessary and nowhere near sufficient, and in an AI-mediated world the gap between having a true idea and getting it established has widened, because the machine leans toward the crowd. So the real question was never whether you’re right. It’s how a true thing establishes itself against a consensus that doesn’t yet include it, and over fifteen years of doing exactly that, I’ve found it takes three things, in order: you anchor on the existing consensus, you get trusted intermediaries to engage, and you build a stable information layer. The first gets you in the door, the second gets you heard, the third gets you understood and repeated. Miss any one and the idea stalls; miss the third and it never lasts.

Pillar one: anchor on the consensus, and the door opens

You start where the machine and the market already agree, and you turn that agreement a few degrees rather than smashing into it. I’ve made this case already, so I’ll keep it short: flat contradiction reads as noise, to humans and machines alike, and noise gets filtered out. Build on what’s believed, add your turn on it, and you’re admitted into the conversation instead of shut out of it. That’s the door, it’s the cheapest of the three, and it’s the one most people get right by instinct, because most of us don’t actually enjoy picking a fight with everything the audience believes.

Pillar two: get trusted intermediaries to engage, and you’re heard

Getting in the door isn’t the same as being heard. To be heard, you need voices that already carry a thick prior to engage with you, to publish you, cite you, put you on a stage, argue with you in public, back some of what you say and push on the rest. When a trusted source engages, its trust flows into you: the machine reads the proximity, the market reads the endorsement, and your own prior climbs on the strength of theirs. You don’t build your standing alone, you borrow established standing until you’ve got enough of your own to stand without it.

Here’s the part that matters for the years ahead, and it’s the thing most people will get wrong: an intermediary is an intermediary, whether it’s a human or a machine. Today it’s mostly humans lending their standing, an editor who publishes you, a peer who cites you, an institution that hands you its stage. Tomorrow it’s increasingly machines corroborating you to other machines, and the mechanism is identical, a trusted source vouching for you and passing on a little of its trust. Build for the humans alone and you’re building for the intermediaries who are on their way out; build for the principle, trusted sources of any kind engaging with you, and you’re building for the world that’s arriving.

I can show you the principle from my own record, because it’s exactly how I got heard. My series runs on a major industry platform, under an editor who chose to publish it. A journal peer-reviewed and published my academic work. Google put me on a keynote stage this year. Companies hired my consultancy because they trusted my judgement enough to pay for it, and peers credit the ideas on the record. None of that was me shouting into the void until the void agreed, it was trusted voices, one at a time across fifteen years, each lending me a little of their standing, until the ideas had standing of their own.

Pillar three: build a stable information layer, and you’re understood and amplified

The first two pillars get you in and get you heard. The third gets the machine to understand you well enough to trust you and repeat you on its own, with no human in the loop, and it’s the one almost nobody builds properly.

Here’s the pillar, stated carefully, because the careless version dates within a year: you build and maintain a stable information layer, a version of the truth about you that holds still everywhere it’s read, by every kind of reader. Not infrastructure for the web, and not infrastructure for AI, but infrastructure for stability, whatever the layer happens to be. Right now that layer is largely the web. Soon it’s the web plus the endpoints assistants read from and the channels agents pull through. Later it’s something nobody has built yet. The substrate keeps changing, and the principle doesn’t: stable information is trustable information, because a human and a machine both grant their trust to what stays consistent everywhere they look, and both withhold it from what wavers, contradicts itself, and drifts. Stability is the trust signal at the level of infrastructure.

That’s the pillar that gets you understood and amplified, it’s where my company’s work lives, and it’s deliberately substrate-independent, which is the whole point: the job doesn’t expire when the web stops being the main channel, because the job was never about the web. It was always about keeping your truth stable across whatever the information layer becomes.

The three run in sequence: belief, then trust, then confidence

Notice the shape of it. Consensus gets you in the door, so the machine will tolerate you. Intermediaries get you heard, so the machine and the market take you seriously. Infrastructure gets you understood and repeated, so the machine carries you forward on its own. In, heard, repeated: a sequence, not a menu you pick from.

And it maps onto something I’ve taught for years, quietly, without forcing the fit. The machine has to understand you first, to know what’s true about you: call that belief. Then it has to weight you as reliable: call that trust. Then it has to be certain enough to actually put you forward: call that confidence. Belief, then trust, then confidence, built in that order, because you can’t trust what you don’t understand and you can’t confidently recommend what you don’t yet trust. The three pillars are how you walk a true idea up that ladder.

Pillar three is the one almost nobody builds, which is exactly why it matters

If you want to know where you’re exposed, it’s the third pillar. Most brands get the first by accident, because they don’t naturally contradict everything their audience believes. Most get the second with effort, because chasing coverage and citations is familiar work. The third they build badly, in a junky, temporary way that falls apart the moment the substrate shifts, because they built it for the web instead of for stability, and the web is only the current layer. Without the stable information layer underneath, the first two pillars don’t hold: the door opens, the voices vouch for you, and then the machine looks at a fragmented, contradictory picture of who you are and quietly lowers its confidence anyway.

That’s the argument for doing the third pillar properly, and it’s the reason there’s a permanent job in it. The layer will keep changing, and keeping your truth stable across it, whatever it becomes, will not.

One strategy for a converging audience

We’re heading into a world where the line between trusting a person and trusting a machine is thinning, where you already trust a search engine and an assistant roughly the way you trust a knowledgeable friend, and that convergence means you don’t need a human strategy and a separate machine strategy. You need one thing, done for both: to be understood, to be trusted, and to stay stable, everywhere you’re read, by every kind of reader. Build the three pillars and a true idea establishes itself, slowly, then all at once. Skip the third, and you’re the goat with its head against the wall, right and unheard, still wondering why being right was never enough.

Publication note: Completes the Bayesian Trust series. First articulation of the Three Pillars of Truth (anchor on consensus, trusted intermediaries, stable information layer) as the method for establishing a true minority view in an AI-mediated world.