The Starlight Accord · Version 1.0
A public constraint on what we are willing to build—and an invitation to test whether the work honors it.
The short answer
Starlight Intelligence exists to turn advanced AI from concentrated technical capacity into distributed human capability.
We build the knowledge, systems, standards, education, and communities that help people direct intelligence without surrendering authorship, privacy, dignity, or responsibility.
The premise
The defining question of the intelligence age is not whether machines become more capable. They will. The defining question is whether people become more capable with them.
A person does not become sovereign because an assistant can answer any question. Sovereignty requires the ability to choose the system, direct its purpose, protect and move personal context, understand consequential outputs, contest decisions, create original value, and leave without losing part of oneself.
After using a Starlight system, a person should be better able to think, decide, create, protect, and contribute.
Advanced intelligence should widen human possibility. It should give a child a patient teacher without replacing curiosity. It should give a creator a studio without dissolving authorship. It should give a founder a capable team without requiring institutional permission.
We reject passive dependency and centralized benevolence as sufficient visions of the future. A healthy intelligence ecosystem needs many models, many builders, verifiable safeguards, portable context, accountable institutions, and people with the capacity to choose among them.
The nine commitments
Each principle carries a question that a team can answer with evidence.
People remain the authors of goals, meaning, and consequential choices. Intelligence may propose and execute; it does not inherit moral responsibility.
Product testCan a person understand, change, refuse, and recover from the system’s action?
Advanced capability should widen participation in education, protection, invention, and economic life—not become a private utility for a cognitive elite.
Product testDoes useful capability reach people beyond the already powerful?
A person should direct their intelligence layer, inspect its memory, move their context, change providers, and leave without losing part of themselves.
Product testCan context be corrected, deleted, exported, and used elsewhere?
A good system leaves the person more able to think, decide, make, and protect—not merely more dependent on the next answer.
Product testWhat durable capability remains when the system is removed?
No single lab, model, state, or philosophy can represent the full plurality of human values. Healthy ecosystems preserve meaningful choice.
Product testCan multiple models, people, and institutions participate without one cognitive center?
Privacy, security, contestability, and recovery belong in the architecture. They are not settings added after trust has already been requested.
Product testWhat happens when the system is wrong, compromised, or used against the person?
People should retain provenance, ownership, audience relationships, reusable workflows, and a legible share of the value they create.
Product testWho owns the catalog, the workflow, and the upside?
Systems that grow beside children carry a higher duty: minimal profiling, age-appropriate explanation, protection from attachment mechanics, and a path toward independence.
Product testDoes the system protect a young person’s right to grow and change?
Intelligence should strengthen the conditions for human and ecological flourishing. Capability without care is not progress.
Product testDoes the full system leave communities and living systems better able to endure?
Constitutional inheritance
Starlight owns the doctrine. Each venture owns a distinct layer of the work. FrankX carries named responsibility across the whole portfolio.
The evidence rule
Every public initiative should be labeled available, in development, research, or horizon. Each product should name the principles it advances and show how a person can verify the claim.
For memory, that means export, correction, deletion, permission boundaries, and portability. For creator systems, it means provenance, approval gates, workflow ownership, and an audience relationship the creator can carry. For learning, it means acquired skill and independent performance rather than completion theater.
The constitution becomes a product operating system.
A living accord
This is not a claim that every tension is solved. It is a public standard against which the work can be judged.