# Archive: Cognitive-Mode Thinking Block

This document preserves the retired thinking block concept that used distinct cognitive modes and a mental-map presentation. It is archived for historical reference and should not guide new design or implementation work.

## Purpose

The thinking block was a temporary inline view of the agent's working state. It replaced a generic "thinking" placeholder with a structured sense of progress.

Its job was to help users understand:
- what the agent is working on now
- which ideas still matter
- whether the agent is gathering information, exploring, connecting, or converging

It was not the final output, and it was not a raw chain-of-thought transcript.

## Core concept

The thinking block was a calm mental map inside the conversation.

- One thought was focal at a time.
- Important thoughts settled into quiet context.
- Short-lived operational thoughts passed through and disappeared.
- Relationships remained only when they helped explain the current step.
- The user observed the block. They did not interact with it.

## Design principles

- Show direction, not internal noise.
- Prefer structure over volume.
- Keep the block calm, restrained, and readable.
- Keep the graph bounded. Preserve useful concepts, not every breadcrumb.
- Make progress feel continuous, not chaotic.
- Let the block leave cleanly once the response begins.

## High-level behavior

1. The block appeared when agent execution started.
2. Exactly one thought was active at any moment.
3. Each new reasoning step resolved the previous active thought.
4. A resolved thought either settled into context or disappeared if it was only temporary.
5. The map stayed small enough to scan quickly. Concepts could merge, connect, branch, or be removed as the agent's understanding changed.
6. The block exited when the response began, or when execution ended due to failure or cancellation.

## Thinking patterns

### Absorb
The agent took in information or formed a new concept.

### Connect
The agent recognized a meaningful relationship between concepts.

### Branch
The agent explored alternatives, options, or child ideas.

### Converge
The agent narrowed, merged, or removed concepts as it got closer to an answer.

### Act
The agent performed an operational step, such as using a tool, in service of the reasoning process.

## Content model

Each step in the block expressed:
- the current thought in plain language
- the concepts involved
- whether that thought should remain visible after the step resolved
- any relationships that should remain visible

Concept labels were short and scannable. One plain-language label appeared per visible thought, ideally around 5-8 words. If a thought remained visible after it resolved, it kept that same label rather than switching to a second expanded sentence.

## Lifecycle

### During execution

The block updated as the agent worked. It felt like one continuing process, not a series of disconnected scenes.

### Handoff

When the final response began, the block resolved its current state and exited. The response became the primary artifact in the conversation.

### Failure or cancel

The block still exited cleanly. It did not remain as dead UI after work stopped.