- Chain mechanics
- How networks actually keep time, order things, and decide something has finished. Most of the work here is building a machine that behaves the way a chain behaves, then noticing that the machine is easier to argue with than the documentation.
- Protocol behaviour
- What happens to a working system when one of its assumptions is removed on purpose. We build the narrow, unreasonable version of a design, put it somewhere real, and record where it goes wrong. It goes wrong one layer above the thing we changed, almost every time.
- Wrapper studies
- Tokenized real-world assets, examined as paperwork rather than as instruments. Three structures that look identical in a wallet give you three completely different things to hold. That difference is the whole subject.
- Autonomous agents
- Small units with a budget, one instruction, and no supervision, observed at population scale. The same finding keeps coming back: an objective is not an intent, and a unit that reads an instruction literally enough is hard to tell from one that is attacking you.
- Settlement pathology
- Where value gets stuck on the way between two systems that each believe they are finished. We build deliberately long routes so the failure has somewhere to happen, and then we let it stall rather than retrying, because a retry destroys the evidence.
- Failure archaeology
- Reading our own logs back. Every experiment here is filed with the log that was written while it ran, including the entries that were wrong at the time. FAILED SUCCESSFULLY is a real status in our system and roughly a fifth of the work ends up under it.
- Containment
- ████████████████████████████████████████████████████████████████████████████████████████████. ██████████████████████████████████████████████████████.