What makes an agent the same agent?
Start with identity when you are thinking about ownership, memory, or changing models. Then follow the interoperability essay into the practical problem of handing work between systems.
Read about continuityTHE VISCA JOURNAL
Notes from the boundary of intelligence and work.
Technical arguments, field observations, and unresolved questions from building agents that persist across software, laboratories, machines, and bodies.
No trend summaries. No synthetic certainty. Work in enough detail to be challenged.FIELD NOTES
Questions that survive contact with the world.
02 · Continuity
The agent is not the model
A model can be replaced overnight. Identity, responsibility, memory, and history cannot. The durable unit of work must live above the intelligence that powers it.5 min read
03 · Scientific work
Evidence is part of execution
In a laboratory, a plausible answer is not an outcome. The instrument state, protocol, observations, anomalies, approvals, and interpretation must survive together.5 min read
04 · Governed machines
Authority is an interface, not a prompt
As agents cross from software into instruments and machines, safety cannot depend on polite instructions. Authority must become visible, bounded, revocable infrastructure.5 min read
05 · Architecture
Interoperability without continuity is portability theater
Moving prompts and tool calls between providers is useful. It is not enough to preserve a worker across models, environments, organizations, and bodies.5 min readA way into the questions.
Start with the problem you are working through. Each essay develops an argument; together, they examine what an agent needs to carry from one situation to the next.
How do humans stay responsible?
Start with authority when you are deciding what an agent may do. Continue to evidence to consider how someone else can review the result after the action has finished.
Read about authorityWhat happens between the tools?
Start with the substrate essay for the broader architecture argument. Read its sources and scope alongside the proposal; the existence of a protocol is different from proof of an end-to-end working system.
Read the architecture argumentEDITORIAL PRINCIPLE
Proof before promise.
Our essays distinguish sourced context, design proposals, and illustrative scenarios. A proposed test is not a reported result; unresolved questions remain part of the argument.
Explore the research program