Result for Mac is a beta app for people who want open-weight models on hardware they control, with computer use on the Mac and a way to reach that compute from an iPhone. You download weights and run them locally, connect private inference such as DGX, Halo, or Ollama in one click, or call Result’s cloud models. Written in Swift and Rust, the app is free today, and those cloud models are free to use without limits for a limited time. The Mac app follows, by a day, an iPhone agent that browses, fills forms, books, and edits reminders and calendar on the device. Sign-ins, payments, and CAPTCHAs are handed back to the user, so passwords and card numbers stay on the phone. The company says models small enough for an iPhone were too slow and not capable enough, so you pair a Mac and tunnel inference to it, including from an iPhone anywhere, the part CEO Aaryan Kushwah has called his favorite. Requests to that Mac pass through a relay that does not store them, and if the Mac does not answer, the cloud model does. While it is serving the phone, the app keeps the Mac awake and uses its processor, memory, and power. Kushwah and CTO Savio Martin are building this as a two-person Y Combinator company from the Spring 2026 batch. Cloud chat currently goes to open-weight models including Qwen, GLM, and DeepSeek through Vercel’s AI Gateway, with zero data retention enabled. Local weights come from Hugging Face.