This extension is 1.28 MB. That footprint is too small to house the "Intelligent AI" engine it advertises. A functional local LLM requires hundreds of megabytes; this file is smaller than a floppy disk. The physics dictate the architecture: your keystrokes aren't staying on your machine. They’re being transmitted to a remote server for inference. The privacy policy exploits this distinction, explicitly denying "storage" (persistence) while remaining silent on "transmission" (processing). You aren't installing a local tool; you're mounting an API pipe that funnels your input out of the browser.
__________

Edit (response to Dev's reply): The Dev's defensive and fallacious reply would have us believe his imprecision was and continues to be due to the following sentence being too long to include in his policy statement: "On the free tier, all auto-correcting is done completely on your own device, meaning the extension doesn't transmit, relay, or store any of your key strokes or words on any servers." The proper response to my review would've been: Thank you for your constructive criticism. I've now clarified the points you raised in my updated privacy policy. Please check it out. Thanks again."
__________

Edit 2: Update (Post-Incident Analysis): The dev's claim that "everything happens locally" and "no data leaves" is demonstrably false based on evidence from my own testing:

1. Unauthorized Data Persistence: After installing then uninstalling the extension, it left a persistent user_pref("keycorrect", ...) entry in my Firefox prefs.js file. This artifact contained my email address, which the extension had cached locally despite my never creating an account or enabling "Snippets Cloud Sync."

2. Unauthorized Data Transmission: Immediately after uninstallation, a feedback form opened and autofilled my email address without any user interaction from me. This proves the extension extracted my browser profile data and transmitted it to an external server.

3. The "Premium Only" Claim is a Lie: The dev says, "The only thing we collect is your email address to track your premium plan." I'm a free user with no premium plan and no cloud sync enabled. The fact that my email was harvested and transmitted proves this claim is false.

4. The "Adaptive Learning" Contradiction: The description brags about "Adaptive Learning" where the tool "adapts to your writing style." A local Levenshtein algorithm doesn't "learn" or "adapt" over time; it uses static logic. This feature description strongly implies server-side processing, directly contradicting the "everything happens locally" claim.

The dev uses semantics ("we don't store it" vs. "we transmit it") instead of substance. The reality is simpler: The extension reads your identity and sends it out, regardless of your subscription status. The "Levenshtein algorithm" excuse doesn't explain why a local text processor needs to phone home with your email address to generate a feedback form, nor why it needs to "learn" your style.