Governed by design, not by discipline

Most trading systems ask one actor, a person, a model, or an agent, to both find the opportunity and decide whether to take it. AUMERA deliberately splits that in two, and neither half can override the other.

Opportunity and authority are different jobs

The specialist agents on the desk exist to build the best possible case for what the market is doing right now. Their job is to be persuasive, fast, and willing to change their mind.

The risk kernel has a narrower, less glamorous job: hold the line. It does not judge whether a thesis is interesting. It checks whether acting on it would break a limit that was set before anyone had a position. That check ignores how confident the agents are, and it cannot be argued down.

"When conviction and control sit in the same place, a strong enough thesis eventually talks its way past the limit that was supposed to stop it." AUMERA Systems Governance Desk

Why the separation matters

When conviction and control sit in the same place, a strong enough thesis eventually talks its way past the limit that was supposed to stop it. Keeping them apart means the system’s worst days are bounded by a rule written on a calm one.

It is also why AUMERA runs as a zero-capital, public-data research build today. The governance model is the product, and proving that the veto holds under a live, plausible-looking thesis comes before capital ever enters the picture.

You can watch it happen: open Agent research and Risk and controls in the paper workspace, and follow one proposal from thesis to verdict.