GIE · 2026–2027
NarraWeave
Dynamic Causal Chain Extraction and Generative Narrative Restructuring
NarraWeave turns unstructured text into a causal Directed Acyclic Graph (DAG). Its nodes represent content items or events; its directed edges represent the conditions and causal order that connect them.
Subject & Background
Operational manuals and multi-character stories are both rich in sequential dependencies, branching conditions and state transitions — yet both stay trapped in static, chapter-isolated prose.
Safety manuals typically describe emergency protocols across disparate chapters — fuel leakage, chemical containment, compartment fire — whereas real-world incidents require evaluating their compound causal interactions. Literary narratives likewise intertwine multiple character timelines, where audiences require coherent perspective filtering without breaking global narrative consistency.
GIE.sh (Global Interconnection Engineering, formerly Generative Interactive Entertainment) is a deep-tech startup focused on next-generation algorithmic storytelling, procedural narrative systems and cognitive graph modeling.
Narrative lineage
The marginal cost of visual generation is approaching zero; logic becomes the only scarce resource in the interactive world.
As world models mature, the supply capacity of narrative assets will determine the industry ceiling. NarraWeave deconstructs literary heritage to provide generative imagery with seamless causal chains.
NarraWeave grew out of GIE’s generative interactive entertainment work and is now a standalone project rather than a company theme: one causal graph carries multi-character storylines and maritime emergency SOPs alike.
- Logic as graphs
- Anchoring scene causality, supporting infinite dynamic branching.
- Semantic vectorization
- Computing story similarity for precise cross-work logic transplantation.
- Narrative assets
- From passively watched short videos to self-directed logic streams.
Worked example: engine-room fire
The original procedure reads as follows: a smoke sensor detects smoke and triggers an alarm; the duty engineer then goes to the site to confirm the fire. After confirmation, the response depends on the fire scale. A small fire can be attacked with a portable dry-powder extinguisher and the incident is then recorded. A larger fire triggers the high-pressure water-mist system, followed by a containment check. If the fire is uncontrolled, the crew sounds the general alarm, evacuates personnel, seals the engine room, and finally releases the CO₂ total-flooding system.
NarraWeave translates this text into a graph: each detection, confirmation, decision, action and result becomes a node. Directed edges show the required order and causal conditions between nodes.
Storylines & the logic market
The same graph reads as an engine-room fire SOP and as a multi-narrator storyline.
In the workbench, every narrator’s manuscript gathers on the left, the system distills it into a storyline, and any node can be opened, edited or branched. The “is the fire contained” decision in the fire procedure and an A/B branch in a storyline are the same structure.
Editor’s picks and community favourites are each composed by many narrators, then distilled in the workbench into logic you can cite and remix. Developers invoke mature story logic like calling an API, audiences filter by perspective, and the graph itself holds global narrative consistency.
What we are building
A causal-graph system, grown in three steps: extract, connect, infer.
Step one
Text to DAG
Given a passage from a maritime training manual, the system outputs a DAG of its causal relationships: nodes are content items or events, edges are the conditions required to enter the next item. Procedures scattered across chapters become computable structure for the first time.
Step two
DAG connection
Two passages from one scenario produce two graphs. When one passage is a prerequisite, consequence or complement of the other, the graphs are linked and the merged result stays acyclic — fuel leakage, chemical containment and compartment fire are no longer disparate chapters.
Step three
Cause and response inference
Given a scenario, the system traces plausible causes back through the graph and projects viable handling actions forward. Past this point the graph is no longer a representation of text: it answers why this happened and what to do next.
Potential
NarraWeave is a causal reasoning engine, not one finished interface. Feed the same graph a different corpus and it changes domain — that is where its ceiling lies.
- Emergency & safety
- Real incidents require evaluating the compound causal interactions of several protocols, not reading manuals chapter by chapter.
- Narrative & interactive entertainment
- Multi-character timelines filter by perspective, branches grow without limit, and the graph itself holds global consistency.
- Reusable logic
- Mature causal chains can be cited, remixed and transplanted across works — invoked like an interface.
Get in touch
Project details and technical implementation are not published here. For partnerships and enquiries, reach us via gie.sh