Reason
Interpret context, compare options, and prepare a decision without presenting preparation as execution.
A11 / Agent Frontiers
The frontier is where capability meets control.
Explore agent systems that coordinate models, tools, memory, evidence, and human decisions without confusing a suggestion with an executed result.
Frontier primitiveA11 / 01
Reason.
Act.
Prove.
Frontier primitives
More capability should create more clarity.
Agent capability is useful when its boundaries are visible. A11 treats reasoning, tool use, and evidence as separate surfaces that can be inspected before the next consequential step.
Act
Expose carefully bounded tools with explicit identity, authority, version, and idempotency requirements.
Prove
Return the evidence trail needed to inspect what happened and what remains blocked.
Controlled autonomy
Let the system move quickly. Keep the decision legible.
A11 does not make uncertainty disappear. It gives uncertainty a visible place in the operating model so people can decide what happens next.
- Suggestions remain advisory until an authorised handoff.
- State-changing actions require a defined authority boundary.
- Stale decisions are rejected rather than silently applied.
- Unknown runtime state is labelled unknown, not live.
Next surface