Proof of Concept: We Built Ithiel With It.
Every client project, every research brief, every line of code, every strategy call prep is handled by Aldric. Not supervised. Handled.
We don't pitch AI. We run on it.
What Aldric does.
Voice and Vision
Talk to Aldric naturally. He sees what you see through Ray-Ban Meta glasses and responds in real time.
Persistent Memory
Every conversation, preference, and decision compounds. After a month, Aldric thinks like you.
Agentic Execution
Aldric does not just suggest. He drafts emails, schedules meetings, updates records, and follows through.
Approval Gates
Every action with side effects requires your explicit approval. You review, edit, and confirm before anything executes.
Immutable Audit Trail
Every action Aldric takes is logged permanently. Searchable, exportable, append-only.
Safe by Default
Hard-coded safety boundaries. No unauthorized financial transactions, no data deletion without confirmation, no credential sharing.
The missing layer.
The AI stack has five layers. Four of them are already owned.
The brain layer stores who you are, what you want, what you have asked for, and what still needs doing. It turns raw model output into genuine intelligence. Context, memory, goal-directed execution. That layer does not exist yet. We are building it.
Trust is the architecture.
Aldric acts on your behalf. That means security is not a feature. It is the foundation.
- Safe Mode on by default. Every action with side effects requires your approval before execution.
- No training on your data. Your interactions are never used to train AI models. Our agreements with model providers explicitly prohibit it.
- Encrypted everywhere. TLS in transit, AES-256 at rest. On-device memory encrypted with AES-GCM before cloud sync.
- Row-level security. Every database query is scoped to your account. Users cannot access each other's data.
- Structured execution only. Aldric operates through defined tool schemas, not free-form code execution. Constrained scope, audited operations.
Aldric will never make financial transactions autonomously, never delete data without confirmation, and never share your credentials. These boundaries are hard-coded and cannot be overridden.
Same brain. New bodies.
The thing we are building is not an app. It is the missing layer in the AI stack. Once the brain layer exists, the interface it runs through becomes beside the point. A phone. A pair of glasses. A humanoid robot. Same brain. New body.
Phase 01 · Live now
The Foundation.
Personal AI chief of staff on iOS. Voice, memory, and execution in your pocket. The brain layer learns from every conversation, every completed task, every preference you express.
Phase 02 · 2026
The screen disappears.
Ray-Ban Meta. Apple Glasses in 2027. Your AI moves with you, in your ear, in your vision. Same brain, same memory, same context. The body changed. The intelligence did not.
Phase 03 · 2027
Agents that talk to each other.
Your Aldric talks to their Aldric. Scheduling, coordination, and logistics handled agent-to-agent. No email chains. No calendar back-and-forth. The work disappears.
Phase 04 · 2028
Physical intelligence.
Humanoid robots are entering the workforce. The hardware is solved. What is not solved: the brain. The same context engine, the same memory, the same goal-directed execution. Just running in a body with hands.
This starts with one person talking to Aldric about their calendar. It ends with a brain layer running inside everything that moves.
Read the manifesto300 spots. Application only.
We are accepting founding members who want a personal AI that actually knows them. Not a chatbot. Not an assistant that resets every session. An intelligence that compounds.