From Cognitive Flow to Speech and Back
Speaking is not the end of cognition. It is part of the loop.
Papers 01-09 built an agent with closure, phenomenology, attachments, planning, and self-monitoring. Communication was mentioned along the way - speech acts as actions, language as cultural medium - but never developed as its own architectural problem.
Paper 10 develops it.
What’s in This Release
Paper 10 - Communicative Agency: Speech Acts Within OODA Control argues that communication is not a capability layered on top of cognition. It is part of the same OODA loop that drives every other action. A speech act is constructed during Orient, committed during Decide, and executed during Act, exactly like any other action.
The cognitive grounding of speech runs through the propositional layer of Paper 06. When the agent speaks, the words point back at the contents that produced them - perceptions, evaluations, decisions already in the stream. This is what makes communicative behavior cognitively grounded rather than confabulated.
The paper introduces a clean distinction between three forms of verbalized stream content: reflection (the agent’s running internal commentary, not an action), internal speech (a committed self-directed speech act), and external speech (a committed speech act delivered to another agent). Reflection accompanies cognition; internal and external speech are actions that the agent commits to. All three reach into the same prior cognitive flow when they ground what they say.
What the agent says does not end its cognitive process. The words enter the stream and shape what flows through the loop next. At the multi-agent scale, speech couples agents in the present; stories couple them across time.
A Long Road to Formal Grounding
This paper closes a loop that began long before OAT existed. In pre-LLM agent systems, the author built architectures in which speech acts were tied directly to the agent’s cognitive state - communicative surfaces of problem-solving steps rather than dialogue moves on a script. Those systems used the NLP technology of their time to approximate what Paper 06 now provides formally: stream contents that speech acts can point back at as their grounding. The integration principle was right; the architectural machinery is only now in place.
The Progression
Papers 01-04 built the subject. Paper 05 gave it a point of view. Paper 06 gave it the means to examine that point of view. Paper 07 gave it reasons to care. Paper 08 gave it a relationship with time. Paper 09 gave it the capacity to learn from its own failures. Paper 10 gives it the means to speak from its own cognitive flow - and to let what it has said return into the stream as input to what comes next.
An agent that speaks from its own cognition - and lets its words return as new cognitive flow - is no longer merely autonomous. It has become a participant.