AI is powerful, extractive,
and here to stay.

01 — The Inputs

What AI actually costs.

The conversation about AI almost always starts with what it can do. Rarely does it start with what it takes. Before a single prompt is typed, before a model generates a single token, the physical world has already paid a price.

Training and running large language models requires an infrastructure of extraction that rarely makes the product page.

Large technology companies are deciding the velocity of adoption of this technology without creating plans to address the externalities they create. It is their obligation to address these issues before pushing for mass adoption — and solutions exist. Closed-loop water recycling systems, for example, can reduce freshwater use by 50–70%. Microsoft announced in 2024 that all new data center designs will use zero-water cooling, saving over 125 million liters per facility annually. The technology is there. The question is whether the will is.

None of this means AI shouldn't exist. It means the people building with it — including me — have a responsibility to use it with intention, not just convenience.

02 — The Outputs

What it actually produces.

AI outputs are only as good as the humans directing them. I've spent the past year building with AI across game design, 3D character creation, AR prototyping, and production workflows — and what I've learned is that the tool amplifies whatever the operator brings to it.

Bring clarity, taste, and domain knowledge: the output is remarkable. Bring vagueness and hope: you get convincing-looking noise.

On this site alone, I've used AI to:

The outputs are impressive. They're also incomplete without the human judgment that shaped every prompt, reviewed every result, and decided what was worth keeping.

But there's a harder truth underneath: the inputs to these models were not ethically obtained. AI systems have been trained on the art, writing, code, and ideas of creatives — often without consent, credit, or compensation. Those same creatives are now being asked to pay for access to tools built from their own labor. That is not a minor concern. It is a structural injustice that the industry has not resolved, and in many cases has actively avoided addressing.

It's also worth naming what AI currently does poorly: it propagates the biases baked into its training data. It has been trained on decades of racist, sexist, and classist writing — and it reflects that back unless actively directed otherwise. The problem hasn't been solved. It's been papered over.

And there's the question of truth. What you see on a screen increasingly may not reflect reality. As AI-generated content floods every medium, there will be — already is — a meaningful shift toward valuing things you can see, feel, and touch in person. Real experiences. Provable presence. This isn't a trend. It's a correction.

03 — Who Chose This?

The accountability question.

Most conversations about AI treat it as weather — a natural phenomenon that arrived, and now we adapt. But AI wasn't inevitable. It was chosen, funded, deployed, and accelerated by specific people with specific interests. That matters.

I don't think everyone building with AI is complicit in harm. I do think everyone building with AI has a responsibility to ask the accountability questions — and to demand that the platforms and companies they work with answer them.

04 — The Impact

What changes for everyone.

The effects of AI aren't abstract or future-tense. They're happening now — in hiring decisions, in creative industries, in the economic structures that determine who has stability and who doesn't.