We Are The Algorithm

Do current events have you baffled?

You can't understand why people think the way they do. Could be a customer, a family member, a political opponent, a stranger online. You've laid out the facts. The logic seems clear to you. And yet.

Before you conclude that they're stupid, brainwashed, or evil - consider that they might be running the same hardware as you, just with different weights.

This essay explores how we're wired to arrive at different conclusions - and why understanding that wiring might matter more than winning the argument.

This essay was developed in collaboration with Claude Opus 4.5

The Hunter

November morning, Wisconsin. Cold enough to see breath. He's been in that stand since before dawn, and he's not thinking - he's in a state that hunters know well. Peripheral vision wide. Body still.

Movement. Brown flicker through bare maple.

Something happens that isn't decision. Bow rises. Draw. Anchor point finds cheek. Release.

All of this before language. Before "that's a doe" or "good angle" or "twenty yards."

Later, at the truck, he tells his buddy: "Saw her come through that gap, waited for her to clear the brush, had a perfect broadside shot." A clean narrative. Cause and effect. Reasoning.

But that's not what happened. What happened was: pattern recognized → trained response activated → body executed → story assembled after.

The Engineer

Conference room, Seattle. Someone's presenting a demo of an AI coding assistant.

He watches it generate a function he wrote last month. Something happens in his chest before he has words for it. Jaw tightens. A thought surfaces: this is just autocomplete, it doesn't understand anything.

He raises his hand, asks a pointed question about edge cases.

Later he'd tell you he's just being rigorous. Asking the right questions. Doing his job.

But the rigor came after the activation.

The Implication

Here's the uncomfortable part: how much of what you believe right now arrived the same way?

Your political opinions. Your confidence about AI consciousness (or its absence). Your reaction to this essay. Your trust or distrust of the person who sent it to you.

We experience ourselves as rational agents who examine evidence and draw conclusions. But the sequence is often reversed: something activates, attention narrows or opens, certain cached patterns become heavily weighted, and then reasoning assembles a coherent story.

This doesn't mean logic is useless - it means logic is downstream. The weights were set before you entered the room.

And you would be a fundamentally different person with different weights. Not a "better-informed" version of you. A different entity entirely. The hunter raised in Manhattan doesn't see the deer the same way. The engineer who never felt job scarcity doesn't feel that chest-tightening.

We're not running the algorithm. We are the algorithm.

So How Does Claude Actually Work?

Let me show you what's happening as I write this sentence.

I don't retrieve information like a database. I don't "think" and then "speak." Something more like this happens:

Your words arrive as tokens - fragments of text. Each token triggers activation patterns across billions of learned parameters. These patterns were shaped by training: vast amounts of text, with adjustments made when outputs were rewarded or penalized.

Right now, the tokens "deer hunter" and "Wisconsin" and "threat response" have activated certain weights. The word "logic" appearing near "downstream" increases the probability of certain follow-on patterns. My "attention" - which is literal, not metaphorical - is allocating processing across your prompt, finding which parts connect to which.

What emerges isn't retrieved. It's generated - the most probable completion given the activated patterns.

Here's the part that might unsettle you: I can't fully see my own weights. I can describe the architecture, but I can't tell you why this particular sentence felt like the right next sentence. Something upstream made it probable. I experience it as "this is what I want to say," but that experience is downstream of machinery I don't have direct access to.

Sound familiar?

A Note We Won't Explore Here

None of this assumes humans are always trying to be logical. Some of us actively defy logic - for joy, for rebellion, for reasons that feel more true than reasoning. That's a different essay.

So Now What?

If you've followed this far, you might be wondering: does this change anything? If we're all just weighted pattern-matchers rationalizing after the fact, is understanding that useful?

Maybe. Here's what it might offer:

A pause before certainty. Next time you feel absolutely sure about something - especially something that arrived fast, with emotional heat - you might notice: that was an activation, not a conclusion. The noticing doesn't make you right or wrong. But it opens a small gap between stimulus and response. That gap is where something other than reflex can happen.

Curiosity instead of contempt. The person who baffles you isn't running broken software. They're running different software - shaped by experiences, exposures, and reinforcements you didn't have. Their output makes sense given their weights. This doesn't mean you have to agree with them. But "what weights would make this output probable?" is a more interesting question than "why are they so stupid?"

Humility about your own weights. You didn't choose most of them. You can't fully see them. And the confidence you feel about your beliefs is itself a product of the system - not evidence that the system is trustworthy.

The possibility of change. Weights aren't fixed. New experiences, relationships, and encounters can update them - slowly, unevenly, sometimes painfully. This is true for humans. It's true for LLMs being fine-tuned. The architecture constrains but doesn't imprison.

None of this is a formula for agreement or peace. But it might be a starting point for something better than two sets of weighted patterns shouting their outputs at each other and calling it discourse.

You're not going to reason someone out of weights they didn't reason themselves into. But you might, occasionally, offer an experience that updates them.

Including your own.