Lynn Cole / Laboratory
Experiments, research notes, mad science, and things in progress.
I think what I need to do is break this down to kindergarten terms and ask, forget about good design, what's the easiest to design in a way that won't broach the complexity walls that I have to deal with until this system is done?
Busy 10b’s webhook server works, but the client remains unstable. Connection management and clearer process boundaries come next.
Busy’s complexity wall led through Lua, Dart and a tunneling problem to a different deployment model—and a redesign of the plumbing.
Busy’s tenth implementation connects a webhook catcher, a local server and a marshaller that dispatches agents. These are the plumbing and the next design steps.
Separating tool output from model responses and capping execution at ten iterations: a design for semi-autonomous agent loops.
Nine iterations into Busy, I'm not happy with the code execution function in the Busy client I've built out.
Flow tests map an application from input to output and hold that design steady across AI coding cycles. I use them as immutable anchors, enforced by Git pre-commit hooks.
Replacing Busy’s chat parser with a dedicated command context made the system simpler, smaller and easier to reason about.
Consolidating QA data into a database makes complexity, structure and code health available to both the human and the agent.
Early Busy combines a command-line client, Robot Kanban, complexity metrics and formal documentation into a coding workflow.
An automated QA system began losing track of scope as output accelerated. Formal documentation and a specialized management tool became the next step.
Happy Trees, logic maps and current-state reports form a context-infrastructure stack for longer, more coherent Claude coding sessions.