During a March 2026 session on expertise modules, Scarlet Dame articulated the problem with generic AI responses:

"AIs do have an opinion but they have a singular opinion in the middle of their training set and a number of narratives that we don't understand. Modules replace this with curated, attributed perspective."

This decision highlights the shift from AI as a retrieval tool to AI as a carrier of specific organizational judgment. It validates the Perspective vs. Information principle, which holds that the system's value lies in providing the interpretive framework that keeps agents from defaulting to the collaborative flatness of their training data.

In a March 2026 strategy session, Scarlet Dame crystallized the product's primary moment of value as the delivery of verifiable provenance:

"The moment of value is when someone sees AI-generated text with citations showing where it came from."

This decision establishes the system's core differentiator: the ability to surface the lineage of organizational thought. It extends the foundational hallucination pitch by providing a visible trust layer that allows any party in a relationship to verify the reasoning behind an AI's response.

During the January 30, 2025 rebrand session, Scarlet Dame established the canonical workflow for capturing organizational thought:

"Human has experience → AI notices → draft together → approve → write to collective memory"

This decision formalizes the conversational nature of capture, which was later extended in a February 20, 2026 call with an early go-to-market advisor to include "long rambling voice memos" as high-quality inputs. The model makes the capture process an active collaboration between the user and their AI.

In the March 19, 2026 strategy session, Scarlet Dame named where the value of human intervention in the capture process lives:

"Each correction from Scarlet was more valuable than the original hypothesis because it revealed thinking that was not yet in collective memory."

This insight drives the design of the real-time feedback loop, where reacting to AI drafts as they are written prevents error compounding and surfaces insights that batch review would miss. It positions the human as the essential source of novel perspective, beyond what the AI can generate on its own.

The March 31, 2026 design session settled how the system communicates back to the user after a save, introducing structured return fields to reinforce agent behavior:

"Three new fields in remember's return value: (1) tx_summary — plain-language summary of what shifted in the worldview; (2) zones — domains/areas affected by the extraction; (3) review_instructions"

This architectural decision was later validated at the May 7, 2026 pilot kickoff, where the customer's chief architect identified the system's ability to show how a perspective moves across transcripts as a key product moment. It transforms the save action from a silent commit into an informative dialogue about the evolving worldview.

During a February 24, 2026 discovery session, Scarlet Dame defined the infrastructure's commitment to data sovereignty:

"Scarlet's product stores all data in the user's own repository — no backend storage — positioning data sovereignty as a trust differentiator."

This principle ensures that users initiate every capture and maintain full control over their records, with the Aswritten backend serving only as a destination for what they choose to send. It aligns with the broader strategy settled on March 19, 2026, where the creator's ownership of data is positioned as a core competitive moat.

In an April 14, 2026 design session, Scarlet Dame confirmed that the system's architecture is fundamentally built around witnessed speech and attribution:

"narr:Actor is a first-class class… Attribution for 'who said what' is already in the graph via narr:Actor_* URIs."

This capability lets the AI reason from specific human contributions, and extends into a model where actors are distinguished from the abstract roles they fill. By referencing specific people and contexts, the perspective adopts the organization's actual voice as told by its members.

The tool naming session on March 30, 2026, reframed the system's verification capability to align with user intent:

"'annotate' renamed to 'cite'. 'Annotate' is a system action; 'cite' is what the user wants: verify where claims come from."

This shift foregrounds the "provenance/citation moment of value" identified on March 19, where trust is built by showing exactly where AI output originated. By moving away from technical jargon, the system makes the act of sourcing knowledge a natural part of the conversational workflow.

As early as February 13, 2026, the introspect tool was architected to serve as the system's second primary answer:

"Becomes an orchestrator, not a direct reader. Reads structural index, identifies gaps, calls context for subgraph, synthesizes what we have / where thin / what to ask."

This capability was later extended during the enterprise pilot in May 2026, where it was reframed as an "active surface" that allows the product to surface its own roadmap by pointing at thin spots in the organizational memory. It ensures that the AI is aware of its own epistemic limits, naming the gaps the perspective hasn't filled yet.

The January 29, 2025 SIC check-in formalized the integration strategy for the platform, selecting the Model Context Protocol (MCP) as the standard for tool exposure:

"GitHub Actions for .sparql/.story file triggers… MCP server protocol for tool exposure."

This protocol serves as the bridge between the organizational perspective and the AI assistants users already employ, eventually extending to a core set of five tools including cite and introspect. By using MCP, Aswritten grounds existing AI workflows in a structured, version-controlled worldview.

In a February 24, 2026 discovery call, Scarlet Dame articulated the core architectural choice to treat organizational knowledge with the same rigor as software:

"Scarlet's product stores knowledge graphs inside the user's own repository, treating organisational memory like code documentation."

This repo-native approach grounds the AI in a version-controlled environment, a concept further refined during a March 4, 2026 customer onboarding to enable institutional knowledge management for non-programmers. By living in the repo, the perspective provides the substance that interprets and renders the AI's work.

A March 31, 2026 call with a long-standing advisor reframed the organizational perspective from a passive store to an active participant:

"The perspective starts being the thing that's doing the development instead of you doing it."

This shift enables the deployment of perspective-grounded agents that act on the organization's behalf, a model later formalized as "agentic steering" in May 2026. It allows a single person or small team to orchestrate complex development processes by setting goals that the AI implements using the organization's actual words and decisions.

The April 8, 2026 session on goal-directed agents established the rules for when an AI must stop acting and reach out to a human:

"Perspective-grounded agents escalate on four types: situation not covered, principle contradiction, claim tension, and ungrounded recipient input."

This principled escalation is what Scarlet Dame calls the "trust contract" that makes delegation possible, ensuring the human is always reachable when the AI hits its limits. By escalating in the same channel where the work is happening, the agent maintains continuity while signaling its own humility and grounding.

In the March 17, 2026 pricing evolution session, Scarlet Dame clarified that sovereign deployment options are driven by specific regulatory needs:

"On-prem deployment is NOT what defines the Professional tier. It is a requirements-driven add-on available at that tier… On-prem is triggered by compliance requirements (HIPAA, GDPR), not by org size."

This posture was immediately validated by the customer's CTO during an April 13 demo, where he identified per-jurisdiction deployment as a hard requirement for their international clients. These sovereign deployments ensure that transcripts and AI history never leave the client's server, fulfilling the needs of high-trust engagements.