Sal. Private AI for Confidential Work.

Sal is a private, sovereign AI partner engineered specifically for practicing attorneys, clinicians, accountants, solo founders, and software engineers who cannot risk uploading confidential client work to corporate cloud servers. Operating 100% on Apple Silicon, Sal delivers deep analytical reasoning across both Mac and iPhone with zero token caps, zero cloud rate limits, and physically zero server infrastructure.

Nearly every AI assistant built for Mac users today is designed for enthusiasts — people comfortable choosing model quantizations, configuring runtimes, and debugging GPU memory errors. Sal was engineered for the other 99%. By embracing tight macOS integration instead of pretending to be a cloud service running elsewhere, HelloPartner.ai made privacy and depth non-negotiable from day one.

• 01

  Completely Private, by Architecture

  Sal does not "phone home." She does not collect telemetry, log user prompts, or route files to HelloPartner.ai or any third party. She is architecturally incapable of it—there is no cloud endpoint to send data to. Your conversations, documents, and voice debriefs stay strictly on your physical Apple hardware. Because HelloPartner.ai maintains zero cloud servers or user data repositories, there is no remote storage or cloud repository to breach, inspect, or subpoena.

• 02

  Accessible Monthly Pricing. Aligned Incentives.

  Sal costs a flat $9.99/month for Mac (with an unconditional 14-day free trial) and is 100% free for iPhone. Unlike cloud AI companies that meter every query and face climbing datacenter COGS, Sal executes directly on the Apple Silicon hardware you already own. Heavy usage never degrades product margins or incurs rate limits—delivering truly unmetered reasoning with zero token taxes.

• 03

  Two Devices. One Mind. (Continuous Companion Ecosystem)

  Sal operates as a unified executive assistant across Mac and iPhone. Interrogate a 100-page merger agreement on your Mac workstation, step into a client meeting, and continue the exact same thread from your pocket on your iPhone. Memory synchronizes end-to-end encrypted through your personal Apple iCloud account—never passing through any HelloPartner.ai intermediary servers—and can be turned off anytime for complete device air-gapping.

• 04

  The HelloPartner Intelligence Director (HID™ Engine)

  Conventional local AI setups require multiple fragmented, uncoordinated open-source runtimes (separate apps for chat, STT, TTS, and diffusion) that violently clash for RAM and GPU threads. HelloPartner engineered the proprietary Intelligence Director (HID™ Engine) directly on Apple MLX and Metal Shaders. A single unified kernel arbitrates scheduling and memory budgets across five specialized modalities without system memory collisions or OS performance degradation.

• 05

  Works 100% Offline with Zero Network Reliance

  Because inference executes directly on Apple Silicon's Neural Engine and Metal GPU, Sal operates flawlessly in airplane mode, hospital basements, remote retreats, and courtrooms with zero cellular signal. For a high-stakes professional, having deep analytical horsepower ready anywhere—without depending on server uptime—transforms daily deep work.

• 06

  Developer IP Protection & Terminal Tools

  Software engineers and technical founders cannot risk sharing proprietary codebases or private API keys with cloud LLMs. Sal features native Terminal Tools to inspect local repositories, analyze git diffs, execute unit tests, and automate shell tasks entirely within sandboxed Apple Silicon memory—eliminating trade secret exposure while providing an autonomous coding partner.

• 07

  Deep Document Pinning & Episodic Memory

  Cloud AI models with bloated 1M-token windows frequently suffer from attention dilution and "lost-in-the-middle" recall failures. Sal solves this through disciplined systems architecture: deep RAM document pinning locks 100+ page contracts, native Apple Pages files, and PDFs into unified memory, while autonomous background note-taking synthesizes past sessions into an associative local knowledge graph. The result is laser focus and cross-week contextual recall with zero hallucination bloat.

• 08

  "Off the Record" Discretion (Zero-Trace Incognito)

  For sensitive tangents, confidential brainstorming, or one-off queries that shouldn't persist in long-term memory, Sal offers an instant "Off the Record" mode. Sal maintains full conversational fluency during your active session, then permanently purges the entire exchange the moment you clear chat or close the app—leaving zero trace across devices.

Sal's primary interface — a focused dark-mode chat window that lives in your menu bar and appears exactly where you need it.
↓ Download full-resolution image

Semantic file search: Sal finds documents by what they mean, not what they're named.
↓ Download

Calendar integration: Sal reads and writes to the macOS Calendar app directly, with no cloud relay.
↓ Download

Reminders integration: Sal writes tasks directly to macOS Reminders — no third-party sync, no cloud. Already on your iPhone.
↓ Download

Ephemeral Sight: Press Shift+Cmd+Return (or force-press PTT while dictating) to invite Sal to glance at your screen. She sees, responds, and forgets. Your desktop is never monitored passively.
↓ Download

Document awareness: Sal reads a pinned PDF or Apple Pages document and answers questions about it — while you keep working in Finder. Everything stays on your Mac.
↓ Download

iPhone Companion: Sal loads the 3.1GB on-device reasoning model directly onto Apple Silicon and synchronizes conversation memory via private iCloud.
↓ Download

All screenshots are approved for editorial use with attribution to HelloPartner.ai. High-resolution originals are available upon request. Do not crop, composite, or alter screenshots beyond standard color correction.

Download hi-res headshot  ↓

Neil Balthaser is a 20-year veteran of enterprise software with deep roots in constrained-resource execution and systems engineering. Beginning in video game development where 60fps frame deadlines, strict memory budgets, and zero cache misses were sacred, Neil went on to co-create the Flex programming language at Macromedia/Adobe, architecting high-performance enterprise runtimes inside the constrained Flash Player virtual machine.

He subsequently founded Intellogo (acquired in 2018), pioneering early enterprise machine learning and "Conversation as a Platform" models, before serving in senior executive product roles at Adobe and Barnes & Noble. Frustrated by the enterprise liabilities, data leakage, and recurring cloud bills of modern subscription AI, Neil brought video game runtime discipline to Apple Silicon, engineering Sal and the HelloPartner Intelligence Director (HID™ Engine) to deliver uncompromised, zero-custody confidential AI directly to knowledge workers, solo founders, and enterprise professionals.