How to use Claude to organise voice transcripts

The author had worn a Limitless Pendant for six months, accumulating roughly 10GB of recorded conversations. They report needing to export that material within 30 days before deletion after the device was banned in the EU. The post describes a mismatch between the device's memory promise and a transcript collection too large to fit into an LLM context window.

To make the exported material usable, the author built a local workflow with Claude Code. It first parses and structures transcripts by date and topic. It then extracts decisions, action items and key insights, before building a searchable knowledge base with cross-references. The final stage generates a CLAUDE.md file containing portable personal context.

That file describes the author, their working style, preferences and ongoing projects. The author says an AI assistant can read it for context without needing the entire conversation history. They also say they wrote up the full prompts for others using voice exports, including Omi or Plaud data, although those prompts are not reproduced in the supplied text. The reported result is a local knowledge workflow and reusable context file.

Who it's for: People with large voice-transcript archives who want searchable records and concise assistant context.

Prompt
Suggested prompt (written by Agents In Use, not the author's)

Source: Hacker News · joozio