a world an agent can act on. i'm building vivacity with aditya bhatia and tanish. we work on the execution layer between an agent and the environments it needs to reason about.
the runtime gives an agent a persistent world: inspect its state, apply an action, simulate forward, fork alternatives, verify constraints, and commit a branch. different domains can use different simulators behind that interface.
my focus is product, pipelines, and the path from an action to a replayable trace. the public surface is tryvivacity.com. the orbit above is an independent browser example of the same verbs. no production api connection.
the execution loop
a branch starts from a parent state. actions produce a trajectory. checks decide which outcomes can be committed.
world = create(spec). state = world.observe(). branch = world.fork(). branch.act(action). trace = branch.simulate(horizon). report = branch.verify(constraints). if the report passes, world.commit(branch). rollback restores a checkpoint. render is one observation, not the world itself.
that is the conceptual interface from vivacity's public runtime contract. this page is not claiming a live backend behind the canvas.
state, observations, and rendering
a world state needs enough information to continue execution. in the orbital example that is position and velocity under a fixed gravitational model. an observation reads some view of that state. the diagram is one such view.
keeping those concerns separate lets an agent inspect quantities directly. a planning loop can compare energy or clearance without rendering a frame. rendering becomes useful when a person, or a vision model, needs to see the result.
routing an execution
vivacity's public architecture puts a router behind the runtime contract. backend families include exact physics, scientific solvers, robotics simulators, learned world models, game engines, and private domain backends. the agent sees one interface. the router sends each step to the engine suited to its domain.
physics and scientific solvers. domains with explicit laws, numerical methods, and quantities that can be checked.
robotics and interactive environments. contact, kinematics, sensors, geometry, and actions inside an environment.
learned world models. predicted observations and transitions, with model uncertainty carried into the decision. those backends are a category, not a list of logos. some of those companies will be competitors. some might be clients. hwahwhahwa
game engines for interactive geometry at interactive rates.
private backends.plant models, warehouse twins, and proprietary solvers can stay inside the owner's network.
the design question is which execution semantics a branch needs: what state the backend accepts, what an action means, and what guarantees accompany its output.
verification belongs in the loop
checks depend on the domain. this orbit uses negative specific energy, periapsis above 1 R, and energy drift below 1e-3. a warehouse would use clearance and units. a circuit would use conservation at a node. a fork keeps the parent available while an alternative is evaluated. a failed branch can be discarded or revised. a passing branch can become the new state. the consequence of each decision stays inspectable.
model, units, and what this page is
experiment 001 integrates newtonian two-body motion in 3d with velocity verlet. μ and the central radius R are both 1. the parent starts circular at 1.8 R, inclined, with speed √(μ/r). each branch is simulated for 20 dimensionless time units at Δt = 0.0125. integration stops if the body meets the surface. there is no drag, no third body, no production backend. act scales velocity by +10%. fork opens three futures: two exact impulses at ±12%, and one noisy learned backend. verify reports energy and periapsis. commit promotes the selected branch only if the checks pass. the learned fork is meant to fail. rollback restores the stored parent. reset returns the circular orbit.
apply an impulse, inspect the trajectory, and compare counterfactuals. then talk to the team if your agent actually needs a world it can branch.