The Continuity Layer
Start Here

From chatbots to continuity, in seven steps

Seven plain-English steps, each building on the last — from a chatbot that forgets you to a continuity layer that carries a whole project forward. No prior deep-AI knowledge needed.

01

Chatbots

Most chatbots are session-bound: you type, it answers, and when the session ends the conversation is gone — unless a memory feature has been bolted on, which is exactly where this staircase is headed.

Example

You explain your project to an assistant on Monday. Friday you open a fresh chat and explain the whole thing again — it has no idea you ever spoke.

Still missing

A chatbot can answer, but it can’t do much on its own — run tools, take actions, carry out a task across steps. For that you need agents.

Glossary: chatbots

02

Agents

An agent is software that uses tools to carry out tasks and workflows — not just replying, but acting: calling APIs, running steps, making changes.

Example

Instead of telling you how to book your usual work trip, an agent opens the booking tool, fills in the dates, picks your usual seat, and submits it.

Still missing

The moment it acts, an agent needs to know things — your preferences, past decisions, who you are. Without stored facts it starts every task blind. For that you need memory.

Glossary: agents

03

Memory

Memory is the store of facts, preferences, and records a system keeps so it doesn’t start from nothing each time.

Example

The agent remembers you always fly economy and bill travel to the “Client X” project, so it stops asking every trip.

Still missing

Memory holds everything it ever learned — but most of it isn’t relevant to what you’re doing right now. Something has to decide what to put in front of the model this moment. That selection is context.

Glossary: memory

04

Context

Context is the working information a system has in view right now, for this specific request.

Example

For “book my usual flight to the client site,” the context is this trip’s dates, your travel preferences, and the client’s address — pulled from memory into this one call.

Still missing

Context tells you what’s relevant now, not what’s true now — which decisions still stand, which threads are open, what shape the work is in. That current shape is state.

Glossary: context

05

State

State is the current condition of a user, workflow, task, or process — what persists between steps and sessions.

Example

The trip is booked but not yet approved; the expense is submitted, awaiting reconciliation. That standing is the state, and it changes as work moves.

Still missing

State tells you what’s true now, not how it got that way — who changed it, when, and whether you can trust it. For that you need provenance.

Glossary: state

06

Provenance

Provenance is the record of where information came from, when it changed, and whether it can be trusted.

Example

The travel policy that capped this trip was set by Finance on 2026-05-01 and superseded the old one — so when the agent applies it, you can see why, and check it.

Still missing

Memory, context, state, and provenance each solve one piece. What binds them into a single working understanding that survives across chats, tools, agents, and time is the continuity layer.

Glossary: provenance

07

Continuity

The continuity layer is what binds memory, context, state, and provenance into one working understanding that carries forward across sessions, tools, agents, and models.

Example

Switch models, hand off to a new agent, come back a month later — the project still knows what’s true, what changed, why, and where it came from. Nothing re-explained from zero.

Where next

This is the layer this project maps.

Glossary: continuity

The continuity layer — the definition

A continuity layer is the part of an AI system that carries forward governed state across sessions, tools, and models: what is currently true, what changed, why it changed, and which evidence supports it. It sits above memory and retrieval, and below agent execution, turning stored history into usable, auditable continuity.

The route from here — four steps, in order

You now have the plain-language staircase. Follow these four steps in order: understand the model, build a local layer, inspect what the Lab has recorded, then check what supports the framework. You can stop after any step and still leave with something usable.