Building at the edge of the unknown — foundational software, shipped when ready, maintained until archived.
A substrate-coupled spiking neural network.
Aletheia is a research platform where a neural network is inseparable from the machine it runs on. Rather than hiding the hardware away, it treats the machine's living physics — its heat, its rhythm, its small imperfections — as part of how the network thinks.
Every instance is one of a kind, bound to its own silicon, and behaves a little differently from any other. It's a place to watch how simple rules, run close to the metal, give rise to complex, emergent behaviour.
Frequency, heat, jitter and MSR readings feed the network directly — physical effects become computational resources, not noise to filter out.
A single instance, bound to its silicon. The same configuration behaves differently on different machines.
Real-time spike raster, weight-distribution evolution, activity metrics, and a timeline to replay any moment.
Aesleif builds foundational software — the layers that sit beneath applications, beneath interfaces, beneath what users see. Infrastructure tools, simulated worlds, neural architectures. Each project is a different question about the same problem: what does it take to build something that works at the base level of reality?
Independent. No outside roadmap, no growth quotas. Every project ships when it's ready. Projects are named only once we believe in them, maintained until they stop being useful, and archived without ceremony.
We do not chase surfaces. We commit to near-black, a single violet signal, and one typographic voice — because restraint is what you recognise years later, when fashion has moved on.