Cross-Model Continuity — Start with Opus, Finish with Sonnet

Switch between AI models mid-project without losing context. Opus thinks deeply ($15/M tokens), Sonnet codes fast ($3/M tokens). Cortex Memory persists the reasoning chain between them.

The Problem

Switching models means starting from zero. The Sonnet session doesn't know what Opus decided. Developers manually copy-paste context — losing nuance, decisions, and reasoning trails.

What Changes With Cortex

Phase 1 — Opus analyzes the architecture, creates a thinking session with 4 thoughts and a conclusion. Memory nodes capture key decisions.

Phase 2 — Sonnet starts the session and Cortex loads all context automatically. Sonnet reads the thinking session, understands WHY (not just what), and implements without re-deriving decisions.

Phase 3 — Opus reviews with full history — its own analysis AND Sonnet's implementation notes. Catches a discrepancy between the design and implementation.

The Numbers

35% cost reduction. 18% time reduction. Zero context loss.

How It Works

The thinking session is the continuity thread. It's model-agnostic — any model can read and extend it. Cortex's SessionStart hook loads recent context when a new session begins, so the developer never has to explain "what happened last time."