Research & prototype lineage
The environment around an agent
Connected explorations of context, memory, tools and the structure that helps agent work survive beyond a conversation.
- My part
- Explored agent context, memory, tools and approval boundaries with agents across related research and prototype projects.
- Where it stands
- Architecture research, source artifacts and historical fixture-based tests. These are connected explorations, not independently launched platforms.
Several of my projects return to a related question: what should the environment around an agent provide? A model response is one part of a larger system of context, tools, memory, permissions and feedback.
This thread runs through agent-workspace experiments, Empirical Systems Lab, primitive-first architecture research and DevGraph. They are related explorations, not a collection of independently launched platforms.
Different ways to examine the question
Empirical Systems Lab contains runtime, memory, checkpoint and evaluation artifacts, with historical fixture-based test records. Those records describe bounded tests; they do not establish a general intelligence claim.
The primitive-first material preserves architecture research and later alternatives, including reuse of existing substrates. DevGraph explores expressing workflow stages and approval boundaries in code; its domain model still identifies major modules as work to be created.
What I want to retain
Working through these questions with agents has generated designs, source artifacts and corrections. The useful thread is how understanding and responsibility carry across a changing system.
Some ideas become tools, some change direction, and some remain research. I want this space to make those distinctions visible without erasing the exploration.