Technical Guide: Organizing Personal Archives with Story of Us

Practical Summary

Story of Us is a methodology for transforming large, unstructured digital archives—such as years of text messages and image attachments—into a coherent, chronological storybook. Rather than relying on automated summarization, the process treats raw data as evidence to be modeled. By breaking archives into time-based packs and using AI as a tool for editorial structure rather than content generation, users can create a private, readable narrative that preserves the nuance of a relationship or personal history. The approach emphasizes data separation, where the source repository contains sensitive working files, while the final output is a static, sanitized artifact.

Prerequisites and Data Preparation

Before beginning, users must aggregate their digital archives, which may include SMS, WhatsApp, and various image formats. The scale of such projects can be significant; for instance, an eight-year archive might contain over 50,000 messages and thousands of attachments. The first step is to quantify the data to understand its shape and scale. This involves calculating message counts, identifying active streaks, and determining the busiest periods. This quantitative analysis is not the final product but a necessary foundation to understand the density of the archive, which helps in identifying gaps where communication was sparse because the individuals were physically together.

Evidence Modeling and Timeline Construction

The core of the workflow is the transition from raw data to an evidence-based timeline. Because thematic categorization often flattens the complexity of life, the narrative must be strictly chronological. Users should organize the archive into time-based evidence packs. Each pack acts as a container for specific periods, incorporating anchors such as travel context, family milestones, and side-channel clues from different messaging platforms. This structure ensures that the narrative accounts for periods of silence, recognizing that a lack of digital communication often indicates high-quality, in-person time rather than a lack of activity or emotional significance.

Integrating Visual Anchors

Attachments are essential for filling gaps in the text record. While many images in an archive are noise—such as receipts or screenshots—others serve as vital visual anchors. The workflow involves filtering these attachments to identify photos that provide context for a specific chapter, such as trips, family dinners, or seasonal changes. By using these images as structural elements, the narrative gains a visual dimension that prose alone cannot provide. This process requires careful selection to ensure that the images support the story without relying on the model to interpret the emotional weight of the content.

The Editorial Workflow

AI should be used as a system for decision-making rather than a primary author. The workflow requires a source-of-truth layer to track chapters, drafts, evidence packs, and deployment states. This prevents the project from becoming a collection of arbitrary model outputs. Users must define constraints for the AI, providing it with specific evidence anchors and clear instructions to preserve uncertainty. If a draft produces overly sentimental or generic prose, it must be discarded. The human role is to maintain the voice, ensuring the final output remains grounded in the evidence while avoiding the pitfalls of automated, generic storytelling.

Privacy and Security Considerations

Privacy must be built into the pipeline from the start. Because the process involves processing private material, users should maintain a strict separation between the source repository and the deployed artifact. The source repo contains sensitive working material, including SQLite databases and raw message logs, which must never be included in the final deployment. Furthermore, users must strip metadata from all images before they are included in the final package, as files often contain device, timestamp, and location data. The final output should be a static, sanitized site that contains only the necessary files for the intended audience.

Deployment and Public Previews

When sharing the project, it is recommended to create a public-facing version that stands on its own without exposing private content. This can be achieved by generating a clean folder that excludes sensitive chapter text and real photos. For the public preview, placeholders or blurred stand-ins can be used to demonstrate the structure and design of the storybook without compromising the privacy of the original archive. This approach allows others to understand the methodology and the final format of the gift without gaining access to the underlying private data or the complex, non-public pipeline used to generate it.

Limitations and Use Cases

This methodology is best suited for individuals who have a large, messy, or emotionally loaded archive they wish to organize into a meaningful narrative. It is not a tool for those seeking a quick, automated solution, as the process requires significant manual editorial work and careful data management. The primary limitation is the potential for AI to over-interpret sparse data or produce generic prose if not strictly constrained. Users should approach this as a system-building exercise rather than a prompt-engineering task. For more information on the project's philosophy and technical approach, see the Story of Us documentation.