The Siri Overhaul
As of early 2026, Apple has successfully deployed "Project Linwood," a total reconstruction of Siri's logic. By replacing legacy code with a massive-scale reasoning engine, Siri has evolved from a voice-trigger assistant into a proactive autonomous OS agent [1].
1.2T Parameter Scale
Moving beyond the 150B parameter foundation, Siri utilizes a unbranded Gemini-based 1.2-trillion parameter model licensed for $1B annually [2]. This "Invisible Brain" enables deep planning and world-knowledge reasoning, while 80% of personal data processing remains local to the device.
App Intents & Awareness
Siri can "see" on-screen content and execute multi-step workflows. It uses Semantic Indexing to understand the relationship between your files, contacts, and calendar events [3].
Private Cloud Compute
Complex reasoning that exceeds the local A19 Pro NPU (45 TOPS peak) is sent to PCC Enclaves. These stateless nodes purge all data immediately post-execution [4]. This ensures user data is cryptographically isolated from all engineers.
| Requirement | Minimum Specification | Intelligence Tier |
|---|---|---|
| Device Memory (RAM) | 8GB RAM Minimum | Full Agentic Access |
| Neural Engine (NPU) | A17 Pro / M1 (25 TOPS) or Newer | Full Agentic Access |
| Storage | 10GB Reserved SSD Space (for local LLM cache) | NPU Cache Active |
| Legacy Devices | Standard iPhone 15 / 6GB RAM | Standard Siri Only |
.jpg)
No comments:
Post a Comment