watch me if you can

The description states that watch me if you can is a macOS application built with Python and local AI models (Moondream 2), using macOS Vision for OCR redaction. It captures foreground window changes, structures them into an encrypted timeline, and allows retrieval via Codex through a local MCP server.

• The system observes only changed foreground windows.
• It excludes certain apps, private browser windows, and secure fields before capture.
• Sensitive information is removed with local OCR.
• Observations are stored in an encrypted append-only ledger with exact and semantic search capabilities.
• A menu-bar app provides controls; a dormant LaunchAgent ensures stable macOS permissions.

Inference The product appears to be a personal productivity tool, not a commercial offering. It is built for the author’s own use, not for distribution or monetization.

The description states that the author built watch me if you can as a private memory layer for their computer, aiming to preserve context behind work so they can switch between projects without manual handoffs and remain accountable when working.

• It is positioned as a contextual continuity tool, not an AI agent or task manager.
• The system is described as non-invasive, never blocking apps or typing for the user.
• It integrates with Codex to make memory useful, but does not replace it.
• The author emphasizes that it’s built around existing tools (Codex) rather than replacing them.

Inference The positioning reflects a personal tool for developers or researchers, not a commercial product. The claim of “accountability” and “context preservation” is framed as personal utility, not market demand.

The description does not state any specific customer base beyond the author’s own use case.

• The system is built for individuals working across multiple tools, research, code, conversations, documents, meetings.
• It targets users who are already using Codex or similar AI engineering loops.
• The author notes that it works around the Codex engineering loop, suggesting a niche audience of developers or technical professionals.

Inference The ICP is likely technical professionals or researchers who use AI tools like Codex and manage complex, multi-project workflows. No evidence of broader customer segments or personas.

There is no evidence of any business model or pricing structure in the description.

• The author states that watch me if you can was built for personal use.
• It is described as a self-hosted, local application, not a SaaS or subscription product.
• No mention of monetization, licensing, or distribution channels.

Inference There is no evidence of a business model. The tool appears to be a personal prototype, not a commercial offering.

The author reports building watch me if you can using:

• Python (managed with uv)
• macOS Vision for OCR redaction
• Moondream 2 for local visual understanding
• Codex as reasoning backend
• pyobjc for accessibility and foreground window capture
• AES-GCM encryption with macOS Keychain

Inference The system is highly technical, built for macOS, and designed with privacy and local processing as core principles. It shows a strong understanding of macOS permissions, AI inference, and data security.

The description makes no claims about external traction or adoption.

• The author states it works on a real Mac and passes 82 automated tests.
• It is described as a personal project, not a product for others.
• No mention of users, customers, or usage metrics beyond the author’s own experience.

Inference There is no evidence of traction or adoption. The tool appears to be a prototype or personal experiment, not a mature product with market validation.

The description does not reference any competitors or similar products.

• It is described as a personal solution to a problem the author faced.
• No mention of existing tools for context preservation, memory layers, or focus management.
• The integration with Codex suggests it’s tailored to a specific ecosystem, not a general-purpose market.

Inference There is no evidence of competitive analysis or market positioning. The tool appears to be unique in its approach, but without external validation or comparison to existing solutions.

• No commercialization or monetization strategy: The tool is personal, not a product.
• No external users or feedback: No evidence of traction, adoption, or customer validation.
• Highly technical and niche: Likely only useful to developers or researchers using Codex.
• Self-reported only: All claims are unverified; no third-party data or metrics.

Inference The biggest risk is that this is a personal prototype, not a scalable or market-ready product. It lacks any commercial due-diligence signals.

1. What is the actual scope of your personal use case? Is it a general solution or specific to your workflow?
2. Have you shared this with others, and if so, what feedback did you get?
3. Are there any plans for broader distribution or monetization?
4. How do you plan to scale beyond macOS or support other platforms?
5. What are the technical limitations of local inference that you’ve encountered?
6. Do you have a long-term vision for how this might evolve into a product, or is it purely personal?