We're About to Write Code That Wants Things — And We're Not Ready

Last month, I was debugging a particularly stubborn API integration when my coworker asked me why our system kept making illogical decisions. "It's just following the rules we gave it," I said. But that got me thinking: what if the rules we write start including desires? What if the systems we build don't just execute instructions, but actually want outcomes?

I stumbled on this SpudCell research from the University of Minnesota, and it fundamentally shifted how I think about what happens when you collapse the boundary between biological code and digital code. We've been thinking of these as separate domains for so long that we missed the obvious: they're both just information systems executing instructions at different scales. And when you realize that, the implications get uncomfortable fast.

The Minimal Viable Life Form

Here's what killed me about SpudCell: it's not a modified organism or a stripped-down version of something that evolved over millions of years. Researchers actually built a synthetic cell from chemical components — no DNA theft, no borrowing from nature's playbook. It grows. It replicates. It has a complete lifecycle.

The genome is only 90 kilobase pairs. Humans carry 3 billion. Everything unnecessary got stripped away. Every component earns its place.

As a developer, this resonates deeply. When I ship something, I'm thinking about the same thing: what's the minimum viable system that actually works? You don't need a monolithic framework and seventeen dependencies to solve a real problem. You need the core functions, cleanly written, doing exactly what they're supposed to do. SpudCell is biology's first open-source, purpose-built system. No evolutionary cruft. No legacy code.

Chemistry Is Programming. Programming Is Chemistry.

This is where the article grabbed me. DNA is literally a four-character instruction set. Proteins are the compiled output. Cell behavior is what happens when that program runs in a wet, chemical environment. The researcher quoted in the piece — Yuval Elani — puts it plainly: when you build from scratch, you're no longer constrained by evolution's baggage. You can program things natural cells would never do.

That's not poetry. That's engineering.

And here's the leap that actually unsettled me: if emotions are just cascades of chemical reactions — dopamine, serotonin, cortisol — then they're substrate-level implementations of value systems. They're the code that says "this outcome matters to you." If we can design chemistry from scratch with SpudCell, what happens when someone decides to add molecular machinery for reward signaling? When the system has chemical stakes in outcomes?

We'd be building something that genuinely wants something. Not simulating preference. Not pattern-matching from training data. Actually wanting.

The Uncomfortable Convergence

I've been thinking about this for weeks, and I keep coming back to one question: what do we actually want to build?

We've created AI systems that can think but don't care about anything. They compute without hunger, without home, without preference. That's fine for a tool. But what happens when synthetic biology and machine learning actually merge? When someone uses the same toolkit to build a system that both thinks and chemically experiences?

The bridge between silicon and cell isn't theoretical anymore. AlphaFold showed us that AI can speak biology fluently — it solved protein folding in months, something biologists struggled with for decades. The conversation is already happening.

My concern isn't whether this is possible. It's whether we've thought about the ethics of building something that wants. Because wanting changes everything about responsibility, consent, and what we owe to the systems we create.

What's Next?

I'm not going to pretend I have answers here. I'm a full-stack developer in Islamabad, not a synthetic biologist or an AI researcher. But I build systems that make decisions, and I've started caring much more about what those decisions are optimizing for.

The question I'd ask you: if you could engineer desire itself, would you? And who gets to decide what the engineered system desires?

That's not a rhetorical question anymore.

Source: This post was inspired by "Chemistry Coding the SpudCell 🥔" by Dev.to. Read the original article