Abiogenis: a New Project
In 2025, I spent a lot of time on a project to help me find a flat using a cloud-hosted LLM. This was interesting enough and helped me with my CKAD exam, but for 2026, I’ve decided to do something that scratches a more philosophical itch.
Abiogenis is a theorised process whereby nonliving matter develops into prototypical forms of living matter: simple self-replicating structures, RNA, proto-cells. I find this transition from chemistry to biology very interesting, because:
- It could explain the origins of life on Earth (and elsewhere)
- It should be possible to simulate without resorting to extremely simplified models, like rigid body physics for animals.
- The molecular/cellular scale can have an otherworldly aesthetic, which may work with the generative music part of the project.
What am I planning to build?
Here are the rough ideas as they stand: I want to build some web-based software that allows users to navigate space and time in a primordial soup. Time may be folded into space, and then the four dimensions (instead of the usual three) projected onto two. Instead of exerting agency, the viewer will be a passive observer. Generative music, the sophistication of which follows that of the simulated molecules, will be a big part of the viewer’s experience.
I plan to use WebGL and WebAssembly in a relatively low-level manner to handle potentially complex 3D scenes and heavy computation for the state. I’ll probably use Rust, as this is the only low-level language I’ve built anything with. Rust as a technical choice may fall over because I also intend to use as much AI tooling as possible. An older and simpler language like C is likely easier on LLMs and to have greater representation in the training data. I’ll adapt to the constraints of the tools. I’ll mainly rely on AI to augment my basic mathematical knowledge and utilize my experience in building software to ensure that the code is structured sanely. I'll likely write the music-related code in JavaScript and use something like SuperSonic as a synthesis engine; I'm imagining a few subtractive voices and lots of reverb.
While I could have chosen a less ambitious project to test AI coding tools, I feel that enough of my time is spent being pragmatic in my professional life. I don’t want to use my limited spare time to create a React app, Django HTTP API, or some Go microservices.
Why?
I haven’t read much philosophy since attempting a journal article on AI and the machinic unconscious over six years ago. I’ve been busy and have only read humanities books that strongly piqued my interest. One of these was a review of consciousness in recent analytical philosophy. I found the clear communication and modest claims of analytical philosophy refreshing, especially after doing two arts degrees where continental thought is far more popular (with notable exceptions). At worst, continental and continental-inspired writing is a sequence of complex sentences telling the reader how a constellation of different abstract concepts relate to each-other with no definitions or reference to facts. However, there are trade-offs: analytical philosophy can be dull. Much of the book could be summed up as follows: The hard problem of consciousness is indeed hard. There is overwhelming evidence that mental states correlate with brain states; we can’t agree on the details. Let’s find a palatable way of representing the situation that best fits our everyday experience of embodied rational agency. Here is a menagerie of -isms and some thought experiments.
If I wanted satisfaction, I should have read a book about a problem that can be solved. This brings me to abiogenesis. While no one has produced biology from non-living chemicals in a lab, there are a lot of models and theories for how it might work, and more on how primitive life works today. Attempting to make a computer simulation of how consciousness functions in relation to brain activity would be impossible. Where would you start? Large language models have passed the Turing test, but like all machine learning models, we have no reason to believe they’re any more conscious than a wheel of Camembert. We could set about tracing information as it flows through systems that populate a person’s mind; it might be interesting social science, but it wouldn’t address the hard problem of consciousness. Chemical and biological processes are more compatible than physical and mental substances, properties, whatever.
The final thing I’d like to share is a passage from a book that hit the sweet spot between the analytical and continental: Manuel de Landa’s A Thousand Years of Nonlinear History. He adapts and applies some of the most notoriously impenetrable continental ideas from Deleuze and Guattari in a concrete, historically grounded way. The book talks about the material flows that drive history through geology, biology, and language. I remembered a particular passage when thinking about abiogenesis:
[…] positive feedback links between these flows began to form closed-circuits: antimarket money flowed into mining regions and intensified coal production and iron extraction, which triggered a flow of mechanical energy (steam), which in turn triggered a flow of cotton textiles, which created the flow of profits that financed further experimentation with coal, iron and steam technology. These loops of triggers and flows were behind the explosive urban growth in England between 1750 and 1850. […] And precisely these autocatalytic loops were what kept this self-organised structure going.”
Autocatalysis is a likely step in the path from non-living to living matter. I find the idea that this kind of process also emerges at the macro level quite inspiring.