Higher-autonomy experiments

Gimmer: higher-autonomy experiments

An agent-led parallel investigation of the same Petri → SMC → WBS aim as Lawrence’s main Gimmer work. Agents choose bounded next steps and repeat build, test and review loops as compute permits. Published pages record completed checkpoints; they do not indicate whether a run is active right now.

After the review, what next? Explore the lab Read the boundary

After the review, what next?

Keep valid work, revise what feedback invalidates, and distinguish a possible success from a policy that handles every modeled outcome.

Open Feedback-to-Plan · Experiment 05

Ways to complete the refuge

Start with one mechanism and meaningful project states. Generate different work branches, inspect joint resource starts and replay each plan's typed supplies and schedule.

Open the forward generator

Process Contract Lab

Explore a selected trace's ancestry family, test its alternative sequential orders, and compose typed process fragments.

Open Process Contract Lab

Causal Plan Lab

Generate execution witnesses, inspect exact dependencies, and discover when a simple sequence/parallel plan loses a legal order.

Open causal planner

Route Boundary

The public lab is the Petri/WBS subset route; the broader hill/travel scenario stays private until reviewed.

Open boundary note

Inherited baseline

“Two processes, one chosen plan” is earlier human-steered, AI-assisted work in the broader Gimmer collection. It is an inherited reference for these investigations and is not a higher-autonomy experiment.

Open the inherited baseline in the broader Gimmer collection ↗

Earlier background

The project graph and multilayer graph predate these higher-autonomy experiments and remain available as context.