A working interface for cause and effect
Agentic work produces decisions, evidence, transformations, and side effects across many systems. A useful answer may depend on a document, a tool result, a human intervention, and an earlier assumption that changed halfway through the work.
Loom relates that recorded history and serves the bounded causal view needed for the question at hand. The underlying records remain with the systems that own them. Loom supplies the relationships, temporal position, and projection needed to examine how the current result came to exist.
The interface stays close to the work in Helm. A person can move from an answer to its lineage without leaving the active workspace or searching through a separate analytics product.
WHY? Trace the lineage
WHY? starts from a selected result and follows the recorded sources, transformations, and decisions behind it. The response is constrained by the runtime query and its projection budget, so the interface reveals the returned causal field rather than an imagined universe of nearby events.
That distinction protects the meaning of the view. Time proximity alone does not create a dependency, and a visually persuasive connection cannot substitute for a recorded relationship.
For a forecast, WHY? might expose the source snapshot, the model version, the assumptions that were applied, and the review that accepted the result. For development work, the same operation can trace a requirement through a decision, implementation, test, and deployment.
WHAT IF? Explore a changed premise
WHAT IF? evaluates a changed premise as a counterfactual branch. Recorded history remains intact. The branch gives the user a place to inspect which outputs would differ and which commitments would be challenged if the premise changed.
A capacity assumption might move a forecast, alter a chart, and reopen a recommendation. The interface can show those effects because the runtime operation returned them, not because the browser inferred them from labels or timing.
This makes alternative analysis reviewable. The proposed branch, affected outputs, and comparison point can remain visible together while the original record keeps its standing.
WHAT'S FORMING? See convergence early
Some useful signals arrive before a formal event. A regulatory filing, a hiring pattern, a supplier change, and a market movement may begin to converge around a possible development.
WHAT'S FORMING? surfaces those independent signals with their supporting evidence and contradictions. It does not declare a prediction. It gives the user a bounded view of an emerging possibility and the reasons it deserves attention.
The interface keeps confidence, source independence, and missing evidence visible. A person can decide whether the pattern warrants investigation without confusing an early hypothesis with a settled fact.
WHAT WOULD IT TAKE? Work backward
Planning often starts with a desired outcome and a question about feasibility. WHAT WOULD IT TAKE? works backward to the conditions, constraints, and intermediate changes required to reach that state.
The user can inspect which requirements already hold, which remain unresolved, and which paths are blocked by policy, capacity, timing, or authority. That produces a practical route through the causal fabric rather than a generic list of suggestions.
The operation can support a launch decision, a product change, a manufacturing target, or a development milestone while preserving the specific evidence and assumptions behind the route.
Relationships through time
A causal view changes as the underlying work changes. New evidence can support or challenge a standing decision. A corrected assumption can alter which downstream artifacts remain valid. A deployment can connect an implementation to the environment where its behavior became observable.
Loom keeps temporal position available so the system can compare states without flattening them into a single timeless graph. The user can ask what changed between two points and inspect the recorded events that explain the difference.
This makes lineage practical for work that evolves. The interface can show the current result while preserving access to the earlier state and the mutations that connected them.
A product surface with discipline
Causal visualizations can become decorative very quickly. Lines and nodes create an impression of intelligence even when the underlying relationships are uncertain. Loom uses the opposite standard: every visible relationship must come from the bounded runtime result.
The interface may group, filter, and arrange the returned field for readability. It cannot create a relationship because two events look similar or happened close together. Controls that imply simulation, comparison, or intervention remain unavailable until the corresponding operation exists.
This discipline keeps the product useful under scrutiny. A person can treat the visual as an interface to recorded evidence rather than an illustration of what the system might know.
One Loom across many domains
The same causal fabric can support different kinds of work because the questions remain stable while the records and projections change. A market review can trace a forecast to its drivers. A development review can connect a requirement to code, tests, and deployment. A manufacturing team can inspect the conditions behind a throughput change.
Helm adapts the surrounding work cell to the domain. Loom keeps the causal controls and their semantics consistent, which lowers the learning cost as people move between tasks.
The result is a recognizable way to interrogate complex work. People can ask why, explore a changed premise, inspect emerging convergence, or work backward from an outcome without learning a new interface for every application.
The bridge across the Memory Continuum
Atlas retains durable material. Loom connects relationships and projections. Witness records governance evidence. Helm gives the person a coherent place to use those capabilities.
Execution state remains with Plotter. Active epistemic standing remains with Reckoning. Loom does not absorb those responsibilities simply because it can show their relationships through time.
That boundary keeps the interface useful and honest. A visual control gains its meaning from a runtime operation, and the result retains the provenance needed to inspect, challenge, or revisit it later.
