Paloma

Paloma is a macOS application that delivers intelligent autocomplete by reading what's on your screen and using that context to predict what you'll type next. Instead of relying solely on what you've typed in the current window, it pulls from content across your other applications to offer suggestions that actually make sense for the work you're doing right now. The result is an autocomplete experience that feels less like guesswork and more like having a coworker who already knows what you're working on. When you start typing a message, Paloma already has the names, terminology, and phrases from your recent context ready to surface as suggestions.

The problem Paloma addresses is the disconnect between standard autocomplete and the way people actually work. Most text prediction tools only see the document you're in, ignoring the browser tab with your research, the Slack thread with context, or the email you just read. They make suggestions based on general language patterns rather than the specific information sitting right in front of you. That's why you end up typing the same project name a dozen times when it's already visible in another window. By watching your screen, Paloma bridges that gap and makes predictions that reflect your immediate context rather than generic statistical patterns.

Everything runs locally on your Mac. There's no cloud processing, no syncing your content to external servers, and no data leaving your machine. The team behind Paloma emphasizes that they never see what you write. For anyone cautious about typing sensitive information while using an AI tool, that's a meaningful distinction. You get the benefits of context awareness without handing your work over to a third party. No API calls to external services, no network requests carrying your screen content, just local processing on your own hardware. This approach makes it viable for people working with confidential documents, financial data, legal materials, or anything else they wouldn't want leaving their laptop.

You have full control over where Paloma operates. You can pick which applications it runs in, and you can pause it anytime. It's designed for people who want the assistance but don't want it everywhere or all the time. If you're writing something personal or confidential, you can shut it off with a single click and turn it back on when you're ready. The per-application controls mean you can keep it active in your email client and code editor while disabling it in your notes app or messaging platform. That granular control makes it practical for mixed workflows where privacy matters in some contexts but not others.

Installation happens through a simple terminal command using curl. You paste a single line, the script downloads the application, and you're running within minutes. It requires Apple Silicon and macOS 13 or later. This isn't a cross-platform tool or a browser extension; it's built specifically for the Mac environment and takes advantage of the permissions model there. That focus keeps it lean and privacy-respecting, though it does mean Windows and Linux users are out of luck. The decision to target only recent Apple Silicon Macs likely reflects performance requirements for the local screen reading and context processing.

Paloma fits writers, developers, support agents, researchers, and anyone else who spends their day moving between applications while composing text. If you've ever copied and pasted names, project codes, or technical terms from one app to another just to get them into an email or document, Paloma aims to eliminate that friction. It learns what's relevant from your screen and surfaces it when you need it. The value compounds the more you context-switch during your day. Someone who works in a single application all day might not notice much difference, but knowledge workers bouncing between browser tabs, documents, chat threads, and code editors will find the contextual predictions noticeably more useful than standard autocomplete.

There's no pricing information published on the site currently, which suggests it may be in an early access or beta phase. The founding team operates it as an unincorporated venture for now, with updates promised if that changes. You can reach them at their published email for questions or access requests. The privacy policy and terms of service are already in place, indicating they're serious about the product even in its early stage, and the Delaware jurisdiction suggests eventual US incorporation is on the roadmap.