Karpathy once said a thought-provoking sentence: "Remove yourself from the bottleneck and maximize your leverage. Input only a few tokens, and let a lot of things happen on your behalf."

But in reality, most people's AI has almost no working memory. You attend meetings, read discussion threads, make decisions, and your brain quietly loses half of the content the next week. Then you have to spend time re-reading, re-asking questions, and re-explaining the context to your AI again.

Rowboat is built to solve this problem. This open-source project is based on Markdown + Obsidian but further extended to work scenarios. It connects emails, meetings, decisions, commitments, and deadlines into a knowledge graph, and this graph becomes denser every day over time with almost no manual intervention from you.

### Why You Need a Second Brain

A major limitation of current AI is the lack of persistent memory. Each conversation is like restarting a blank session, requiring re-explanation of the context. This is not only inefficient but also limits the potential of AI as a work partner.

Rowboat solves this problem in the following ways:

1. Automatically extract information from sources such as Gmail and meeting minutes

2. Save decisions, commitments, and deadlines as separate Markdown files

3. Establish connections between people, organizations, and projects

### Technical Implementation

Rowboat runs entirely locally, with data stored in the `~/.rowboat/` directory using Markdown format compatible with Obsidian. This means:

- You can continue using Obsidian to edit notes

- All data remains on your machine

- No risk of vendor lock-in


### Practical Application Scenarios

- **Meeting Preparation**: Type "Prepare for my 2 PM meeting with Sarah Chen", and the AI will extract past relevant decisions, unresolved issues, and discussion threads

- **Daily Summary**: Ask "Create a voice summary for me today", and the AI will organize the day's meetings and emails

- **Long-term Tracking**: Notes marked with `@rowboat` will be automatically updated to keep information current

### How to Get Started

1. Download the corresponding version from the [Rowboat official website](https://www.rowboatlabs.com/downloads)

2. Configure model connections (supports Ollama local models or APIs like OpenAI)

3. Set up Google service integration (optional)

4. Start using

GitHub Repository: [rowboatlabs/rowboat](https://github.com/rowboatlabs/rowboat)

### Thoughts

Rowboat represents a trend: AI is evolving from a mere conversation tool to a work partner with persistent memory. This shift may change how we collaborate with AI, making it truly a "leverage" rather than another tool to manage.

However, a potential challenge for such systems is how to handle outdated or incorrect information. As time passes, the knowledge graph will become increasingly complex, and ensuring that the information within it remains accurate and relevant may be a problem to solve in the future.