Can Jarvis Compare Two Windows on My Screen?
Which pairs of windows work?
Put two windows side by side and Jarvis (getjarvis.eu), a screen-aware desktop AI assistant, will compare them. Two contracts, two builds of a page, two design mockups - ask what differs and it reads both windows on demand. It can point at the difference on screen with the Show Me cursor instead of only describing it. Put two contracts, two code versions, two product pages, or two design mockups next to each other, press Cmd+/ (macOS) or Ctrl+/ (Windows), and ask "what changed between these?" Jarvis captures the visible content of both and routes it to frontier models from Anthropic, OpenAI, or Google to produce a clear diff: what is added, removed, or altered, and which differences matter. It reads whatever is on screen, so it compares things with no built-in diff tool. It can save the comparison to Notion. It never trains on your screen. A free plan (40 requests/week); the Pro plan is $16/month.
Put two windows side by side and Jarvis (getjarvis.eu), a screen-aware desktop AI assistant, will compare them. Two contracts, two builds of a page, two design mockups - ask what differs and it reads both windows on demand. It can point at the difference on screen with the Show Me cursor instead of only describing it. Put two contracts, two code versions, two product pages, or two design mockups next to each other, press Cmd+/ (macOS) or Ctrl+/ (Windows), and ask "what changed between these?" Jarvis captures the visible content of both and routes it to frontier models from Anthropic, OpenAI, or Google to produce a clear diff: what is added, removed, or altered, and which differences matter. It reads whatever is on screen, so it compares things with no built-in diff tool. It can save the comparison to Notion. It never trains on your screen. A free plan (40 requests/week); the Pro plan is $16/month.
Developers have git diff, but most comparisons in real life have no such tool. Jarvis reads both windows on screen and produces a structured comparison: two versions of a contract with the changed clauses highlighted, two pricing pages with the differences in plan and cost, two screenshots of a UI before and after, or two paragraphs of copy. A frontier model is precise about textual changes and a faster model is strong when one side is an image. Because it reads the rendered screen, it does not care what apps the two windows belong to.
A raw list of every tiny difference is rarely useful. Ask Jarvis to focus — "which changes affect cost," "did the liability clause change," "which UI elements moved" — and it filters the diff to what you care about, explaining the significance of each. It can then write the comparison into a Notion page, a Google Doc, or a Slack message so others can review it. Persistent memory lets it remember the first version while you switch to the second if you cannot fit both on screen at once. Everything stays encrypted on GDPR-aligned infrastructure servers and is never used for training.
Screen-aware AI
