New Tool Aims to Fix Executive Assistant AI Memory Loss

Updated 21 Sep 2026 · Reviewed by the Zekai editorial team

A new GitHub project called Lossless-memory addresses a common frustration for daily AI users: the tendency of models to forget or summarize past interactions. Built by a single developer named Aru alongside an AI named Cece, the repository offers an alternative to standard memory systems that rely on compact notes or similar chunks. The developer argues that standard summarization processes drop the exact phrasing and timing that make past records useful.

Instead of compressing conversations, the system converts every turn into a fixed seven-field record. These records are appended daily into JSONL files, functioning as an unalterable source of truth. Time acts as the primary axis rather than just metadata, allowing any underlying indexes to be completely deleted and rebuilt directly from the raw log files if necessary.

To maintain usability, the tool uses a small index of topic markers known as LLL. These short, timestamped lines indicate when a conversation shifts subjects and are injected into the model context during every turn. A small daemon handles incremental re-indexing every ten minutes by default, scanning for rewritten source files without ever requiring a full rebuild from scratch.

Practical tests detailed in the quickstart documentation show that users can ingest sample conversations, build the necessary indexes, and run time-scoped queries in approximately five minutes. A pytest round-trip test covers this exact functional path. The implementation has run daily since July 2026 for a single user, with raw logs stretching back to June 2026, building on design concepts dating to December 2025.

For executive assistants and professionals managing ongoing projects with an AI, the system changes how past decisions are retrieved. Rather than receiving a paraphrase of a similar discussion from weeks prior, the model can pull the exact lines from a specific time, such as a Tuesday night session, keeping conversational context intact across sessions.

Why it matters for AI for Personal Productivity & Executive Assistants

Executive assistants and professionals who maintain long-running dialogues with AI assistants often face context degradation over time. By maintaining raw, timestamped logs without summarization, this open-source tool enables assistants to recall exact prior decisions and specific turns of phrase, improving continuity across daily productivity workflows.

What to do about it

• Review the quickstart documentation in the GitHub repository to test the ingestion process.
• Run the pytest round-trip test to evaluate the memory setup on local hardware.
• Examine the lessons documentation to understand potential failure modes and disk space requirements.

Source: Show HN: Lossless-memory – a personal AI memory that never summarizes (github.com)