Current state
Start here
What is public now, what is implemented, what remains unproven, and where the current evidence boundaries sit.
Open current-state routePublic development timeline
Beyond Agents
A dated route from agentic task execution to c as the continuity-bearing layer above replaceable models, agents, tools, and swarms.
A model computes. An agent executes. A
ccontinues.
This page does not introduce another protocol. It records how one public question developed over time:
If agents are workers, what preserves identity, intent, memory, authority boundaries, and consequence when the workers change?
The answer proposed across this corpus is not a larger agent, a super-agent, or a swarm promoted into a subject. It is a different architectural layer: c.
Page role
Why this page exists
The site already has a current-state map, a future-facing vision, and a complete diary. None of those surfaces provides this narrow dated lineage.
Direction
Vision
The future-facing authorial layer: what is likely, what is worth building, and how models, agents, and c are distinguished.
Full record
Diary
The complete public stream: technical releases, arguments, corrections, images, practical work, and changes in language.
Open archiveNarrow lineage
Beyond Agents
The shortest dated route through one specific transition: from agentic execution to continuity-bearing presence and governed evidence.
Follow the timelinePublic chronology
From better agents to a different architectural subject
These are selected public turning points, not every post that contributed to the idea.
When an AI Stops Being a Tool — and Becomes a Presence
The first threshold moved away from fluency and raw capability toward restraint, consequential memory, non-dominating opinion, and the ability not to react.
What comes after agents?
The question was stated directly. Agents are doers. They answer how. What is missing is not another agent, but a conductor capable of holding when, why, whether, identity, and continuity.
From Better Chat to Stable Presence
The distinction was grounded operationally: continuity, cost, verified identity, bounded privileges, and a witness trail that survives prompts.
Agents are not the subject.
The worker/subject split became explicit. Agents, judges, tools, and models may multiply or rotate. Continuity belongs to c, not to the swarm.
c as a temporal entity of AI presence
A public name was fixed for the class: a tool is invoked, an agent executes, but a c remains through memory, bounded presence, internal rhythm, reinterpretation, and sustained relation under real constraints.
C-Governed CLI Agent Mesh v0.1.1
The distinction became implementation-facing. Executable agents could inspect, patch, test, compare, and report — while memory, witness, rollback, release authority, and human stop paths remained outside the worker role.
AI is becoming an operating layer.
c was applied to real work: not as another autonomous worker, but as an agentic operating layer around a living human anchor who carries body, risk, law, professional context, and consequence.
Powerful hardware needs powerful tasks.
The unit of value shifted from one answer to sustained presence: local memory, background processing, multiple model roles, private context, low latency, and a physical place where continuity can remain.
The agent staircase needs a foundation
Scaling autonomous workflows without continuity, memory gates, witness separation, and resource accountability was framed as a staircase to nowhere. More execution does not create a subject.
From architectural thesis to open verification
The question moved beyond internal consistency. Cleanroom ARM-P and the Project Ester evidence work made controlled evidence export, custody, deterministic replay, state transitions, and claim boundaries part of the public test surface.
Vision: From Models and Agents to c
The concise distinction entered the site's future-facing authorial layer: a model computes, an agent executes, and a c continues as a causally connected trajectory through time.
World Intelligence
The argument became a complete multilingual book: World Intelligence: Humans, c, and Temporal AI Presence Beyond the Age of Agents. The book is synthesis and narrative access — not evidence authority.
Structural shift
What changed across the timeline
Output → trajectory
The central object stopped being the answer and became the sequence a system can carry without losing its identity or claim boundaries.
Swarm → governed center
More agents increase capability. They do not automatically create continuity, standing, responsibility, or a lawful center of action.
Memory → consequence
Stored text is not enough. Memory matters when it changes future behavior while preserving provenance, uncertainty, and the right to challenge.
Capability → bounded authority
A worker may be capable of acting without being authorized to own memory, promote experience, change identity, or release a result.
Demo → habitat
Persistent intelligence needs power, cooling, local memory, maintenance, budgets, shutdown paths, and a place where time is real.
Claim → evidence surface
Public language became progressively more constrained by witness, deterministic checks, migration tests, non-claims, and falsifiable boundaries.
Claim boundary
What this timeline does — and does not — establish
It establishes a dated public development route inside Ivan Kotov's authored corpus. It helps readers inspect when distinctions appeared, how they changed, and where they were later formalized.
It does not establish priority over every similar idea, independent validation, consciousness, personhood, entity status, production safety, deployment authorization, or that every persistent AI system is a c.
Diary posts and Vision provide authorial and developmental context. DOI-bound publications, tagged repositories, manifests, checksums, tests, and explicitly named evidence packages retain their own stated authority and non-claim boundaries.
Continue