Andrew Osula AO

Research case R5

AI Operations · Browser Extension · Product Design

Research

AI Chat Thread Manager

Tracking AI conversations across providers, with checkpoints and next actions stored locally.

AI chat thread manager / research prototype

Know where the work stopped and what needs to happen next.

Fictional threads demonstrate a unified work queue. Filters and checkpoints run locally in this reconstruction.

AI Chat Thread ManagerUnified workspace / 6 illustrative threads
Local metadata

Workspace

Your work, across conversations.

Find the checkpoint. Review the next action. Continue with context.

6 shown

In progress

2

Waiting

1

Needs review

2

Done

1

Status, checkpoint and next action remain visible across providers. Fictional records only.

The challenge

Complex AI threads become difficult to resume when status, checkpoints, next actions and context health are not explicitly managed.

Long AI conversations are difficult to resume when the last decision, the next task and the state of the context are scattered across providers.

The approach

The browser-extension prototype records thread identity, status, checkpoints and next actions in a local work queue. The focus is continuity and a clear hand-off, rather than copying full conversations into a central service.

System

Select a step to see its role in the work.

  1. 01Provider adapters

    Read the supported providers through explicit, limited adapters.

  2. 02Thread identity

    Keep a stable reference to each conversation across sessions.

  3. 03Status

    Mark whether a thread is active, waiting or ready for review.

  4. 04Checkpoint

    Save a checkpoint so work can resume without reconstructing the full conversation.

  5. 05Next action

    Make the next action visible alongside the thread.

  6. 06Context health

    Flag when the available context may be too weak for a reliable continuation.

  7. 07Handoff

    Prepare a concise transfer of context and outstanding decisions.

  8. 08Unified work queue

    Bring threads together in a local work queue without merging their content.

Current position

A working prototype is available as personal product research. The public examples use fictional conversations and do not represent an employer deployment.

Presented as personal research and product-development evidence, not as an employer project.

Next question

Continue testing whether checkpoints and context-health signals help people resume complex work accurately across providers.

Knowledge graph

Follow the methods and information shared by this work and the cases below.

CurrentAI Chat Thread Manager
AI operating modelGoverned AI Agents & Agentic WorkflowsResearch · ResearchStructured AI workDesign Prompt StudioResearch · Research