Implications

If the work succeeds:

• General claims such as safe, aligned, or intelligent give way to bounded, testable properties.

• Developers gain maps of achievable capabilities and explicit operating conditions.

• Evaluations are judged by whether they predict behavior where systems are actually used.

• AI can be adopted faster where the evidence warrants confidence.

• Systems facing unresolved uncertainty receive limits, monitoring, reversibility, and human escalation.

• Multi-agent infrastructure adds identity, provenance, accountability, and systemic-risk controls.

• Capital moves away from structurally unreliable promises and toward defensible applications.

• Human authority remains explicit where technical guarantees end.

If the evidence cuts against Sponte's hypotheses, the claimed boundaries should narrow accordingly.