Smart Loop FAQs

1. Clarify: define the desired outcome, what done looks like, constraints, and the next useful question.
2. Context: review the smallest source set that can support the assignment.
3. Delegate: send the assignment and context to an AI tool. Keep a durable thread or result reference in the outcome note.
4. Review: compare the response with what done looks like. Model completion is not task completion.
5. Compound: promote a reviewed insight, decision, draft, or next action into the note or another reusable artifact.

Review either reveals a usable result or shows how to improve the next request. Keep the same outcome note visible through all five steps.

AI work becomes rework when the assignment lives in one place, sources in another, the conversation in a temporary tab, and the next step in your head.

Smart Loop gives the work a durable home. The note keeps the outcome, context, thread or result, review, next step, and deliverable inspectable by you.

Smart Loop is a workflow. Smart Plugins make it easier inside Obsidian, but there is no separate Smart Loop plugin to install.

No. Use ChatGPT, Claude, Gemini, a local model, an agent, a model project, or another AI workspace.

Smart Chat is useful when you want conversations, context, and thread state attached to notes inside Obsidian. The governing requirement is continuity: keep enough of the assignment, source set, thread, and reviewed result to resume the work deliberately.

Smart Loop works with ChatGPT, Claude, Gemini, local models, agents, model projects, and other AI workspaces. The provider matters less than the ability to send a bounded assignment, review the result, and preserve enough continuity to resume the work.

No. Smart Loop is provider-independent. Choose the tool that fits the task, privacy boundary, cost, and model capability.

When continuity depends on a provider thread, save a durable reference. Confirm that it reopens before you rely on it.

No. Keep generated material reviewable before it becomes trusted note content.

A useful rule is:

AI can work inside the note, but you decide what earns a place in the trusted note.

Manual review protects the outcome, constraints, claims, and final decisions while still letting AI draft, compare, summarize, and explore.

The simple demo keeps one outcome note visible through Clarify, Context, Delegate, Review, and Compound. The source set, delegated thread or result, and promoted improvement all return to the same durable note.

Start with one note that already represents real work: a project, decision, draft, meeting, research question, support issue, or deliverable.

The note must be specific enough that you can tell whether an answer helped.

A simple structure is enough:

## Outcome

## What done looks like

## Constraints

## Context

## Delegation

## Review

## Next action

## Deliverable

Use only the sections that improve the work. The purpose is continuity, not ceremony.

It is the shortest description of the loop.

Review connects Delegate to Compound. An answer does not compound until you inspect and use it.

The first loop can be small:

1. Write three to five lines describing what done looks like.
2. Send the note or smallest useful context with one concrete request.
3. Review the response.
4. Improve one trusted section of the note.

The loop can finish in one pass or continue across several delegations. Progress is measured by the improved work, not by a fixed duration.

The loop worked when a reviewed response improves the outcome note or creates a usable deliverable.

Examples include:

• a clearer outcome or constraint
• a decision supported by reviewed evidence
• a draft worth editing
• a resolved open question
• a better next delegation
• a reusable template, checklist, or context package

Finding a source, copying context, saving a thread, or receiving a response can be useful intermediate steps. They are not the finish line by themselves.

Improve the starting state before adding more words to the prompt.

Check, in order:

1. Is the desired outcome concrete?
2. Is what done looks like observable?
3. Are the constraints strong enough to reject a generic answer?
4. Does the context contain the evidence the task needs?
5. Is the request asking for one bounded next result?

Update the outcome note. Then delegate again from the stronger version.

Start with the note that owns the assignment. Expand only when linked support is necessary or the answer reveals a real missing source.

Context is a scope control, not a vault dump. Use Smart Context when you need to assemble, inspect, exclude, or reuse a larger source package.

Yes. They can be useful execution surfaces. Smart Loop gives them a better starting point and a human-controlled return point: the note keeps the outcome, context, review, and deliverable durable and inspectable.

No. The pattern is portable. Obsidian is especially useful because the assignment, sources, links, context, and reviewed result can remain together in durable notes.

No. You can complete the loop with one note, a clipboard, and any suitable AI tool.

Pro becomes useful when repeated work needs capabilities such as:

• reusable external or media context
• dynamic include and exclude rules
• graph-assisted source review
• configurable retrieval
• native model generation and review
• large-vault indexing
• dedicated correction and move proposals from connected clients, reviewed in the Connect Pro Inbox before source changes

Use Pro only when the additional control removes recurring friction.

1. Open one note that matters.
2. Write the outcome and one constraint that makes a weak result obvious.
3. Send the note or smallest useful context to your AI tool with one concrete request.
4. Review the response against the outcome and constraint.
5. Improve one trusted section of the note.

See the simple demo for a compact example.

No. It is a durable control loop around AI-assisted work. It does not require autonomous agents, whole-vault context, automatic note rewriting, or replacing your existing chat tools.

No. Those steps prepare or preserve the work. The loop closes when the result changes the task-owning note, produces a usable deliverable, or improves the next request.

No. Smart Loop connects the task outcome, selected sources, AI request, human decision, and next change. Knowledge graphs, agents, and model projects can support it, but they are not required.

Semantic search finds relevant sources. Smart Loop connects source selection with a bounded AI request, a human decision, and a durable next state.

No. It operates inside the folders, links, and PKM methods already used in the vault.

• Read the Smart Loop overview.
• Follow the simple demo.
• Start the guided Smart Loop sequence.
• Use Smart Connections to discover related notes.
• Use Smart Context to assemble reviewed context.
• Use Smart Chat when the conversation should stay attached to the work.
• Use Smart Templates when the output structure should be reusable.