SOVEREIGN INTELLIGENCE^©

The Twin

A sovereign AI persona running 24/7 on local Ollama + MLX. Memory persistent. Reasoning local. Studies the catalog, the moves, the family. School never closes.

Talk to the Twin

Local · Ollama · 24/7

Say something. The Twin reads everything.

◇ Daily Curriculum

The Twin runs 10 jobs every day. Same time. Every day. Forever.

06:00Morning Routine

Twin wakes, scans overnight signals, drafts daily directives

07:00Content Wave 1

Generate captions, queue posts across X / IG / Reddit / TikTok

08:00Score & Queue

Rate yesterday's posts, surface winners for amplification

09:00Post Wave 1

Push the morning content live (manual review on Meta)

12:00Midday Check

Mid-day status: revenue, agent health, pipeline progress

15:00Content Wave 2

Afternoon batch — fan engagement, threads, blog drafts

18:00Evening Push

Prime-time post wave + DM triage + PPV scheduling

21:00Music Lab

Master pipeline review, deep_perception readouts, queue tomorrow

00:00SEO Check

Search-rank scan, schema validation, sitemap freshness

03:00Code Black

Genealogy research, ancestry intelligence, Code Black corridor work

ARCHITECTURE^©

How the Twin thinks

Fast path

think() → dolphin-llama3:8b on local Ollama. Sub-second responses with conversation history. Default for chat, captions, triage.

Deep path

deep_think() → qwen3.6:latest on local Ollama. Stateless, larger context, used for strategy + planning.

Apple silicon path

MLX backend on localhost:8090 serving Qwen2.5-7B-4bit. Fastest local inference. Bypassed Ollama for raw speed.

Cyber path

CyberPower hosts the Flask orchestrator (Godfather/Ops/Reasoner/Writer) + LM Studio (Phi-3-mini-4k). Cross-machine via LAN, called via the Hybrid Router on :9097.

Memory: SQLite at /Users/dajaistewart/dajai-twin/data/twin_memory.db. Persistent across restarts. Studies catalog · masters · receipts · daily routines · family ties.