How a Post‑00 Team Open‑Sourced OpenAI’s Chronicle Within 48 Hours

On April 20 OpenAI launched Chronicle, a feature that lets AI directly see a user’s screen and retain contextual information, but it is limited to ChatGPT Pro subscribers at $100 per month.

Just 48 hours later the developer group “Vida”, composed of post‑2000‑born engineers, published OpenChronicle on GitHub (https://github.com/Einsia/OpenChronicle). In their announcement they argued that AI’s "eyes and memory" should not be locked behind a paywall and should become infrastructure anyone can use.

OpenChronicle builds on the same core capability—screen observation and persistent memory—but makes three more aggressive moves: it runs entirely locally, it can attach to any model (Claude Code, Codex, OpenCode, Claude Desktop, etc.), and its memory can be shared across different AI agents.

The project demonstrates three concrete use cases. First, when an AI is asked a vague reference such as “what’s the bug of that?”, it retrieves the current screen context (e.g., the open file and error message in VS Code) and resolves the pronoun accurately. Second, a test with Claude showed that without continuous memory the model asks “What is OpenChronicle?” whereas with OpenChronicle it pulls project information from the browser, Feishu, and VS Code and produces a logo prompt without any prior explanation, proving cross‑session continuity. Third, the system learns user habits: if a user says “Add dinner with my parents this Sunday,” the agent routes the event to the personal calendar (Apple/Fantastical) instead of the work calendar (Google Calendar), illustrating habit‑aware execution.

Technically, OpenChronicle stores memory as Markdown files indexed with SQLite and exposes UI structure through an AX Tree, allowing users to read, modify, or migrate the data. This transparent design contrasts with the black‑box approach of many commercial AI memory services.

The release quickly ignited discussion in overseas communities, generating over 2 000 posts within nine hours. A highlighted comment described the project as moving AI from a "product shape" to a "system shape".

Beyond the immediate functionality, the article argues that AI interaction is shifting from isolated queries to a continuous process where the model observes the environment, incorporates historical context, participates in ongoing tasks, and leaves a persistent trace. By keeping the memory layer local and user‑controlled, OpenChronicle raises broader questions about who ultimately owns the long‑term record of a user’s behavior, habits, and work processes.