I replaced Grammarly Pro with a local model on my GPU, and my drafts never leave my PC

For months, it seems the trend of subscription services has escalated, with firms imposing recurring monthly costs and pressuring customers to commit to their platforms. Grammarly Premium, now called Grammarly Pro, was one such service that compelled users to pay a monthly fee, with every edit contributing to training the service's models unless the user selected an opt-out option. This feature proved problematic for those seeking privacy and control over their data.

Grammarly's Premium often collected and analyzed users' drafts, even when anonymized, raising concerns over data privacy. Even with opting out, the service still required an active internet connection for analysis, placing users' texts through its servers before returning the processed content. This meant that even anonymized data could potentially be exposed to third parties.

A potential solution lies in local language models (LLMs) running on a user's GPU. Writing Tools, an open-source application available on GitHub, offers a viable alternative to Grammarly Pro. The application can be installed and run locally, acting as a bridge between the user's selected text and a chosen LLM. By running the selected text through the LLM without sending it to external servers, Writing Tools enables users to proofread, modify tone, and adjust length without compromising their privacy.

To demonstrate Writing Tools' effectiveness, I ran a test with a deliberately error-ridden email. The tool successfully corrected all 24 errors, preserving the original meaning. Although certain presets may not be as reliable as Grammarly's, the tool provides a robust alternative for users seeking a cost-free, privacy-focused writing assistant. With Writing Tools, users can utilize local LLMs like llama.cpp, Ollama, or LM Studio servers, ensuring their data remains securely stored on their personal computers.