'Hover or Select to find out More' Feature

ChatGPT Feature requests
1

Problem / Motivation

When reading in Atlas (web pages, ChatGPT responses, equations, docs), I often hit small pieces of text that I don’t understand. I want a quick explanation in-place without:

  • spawning a new chat/branch (which clutters my chat list / file system), and

  • permanently saving the interaction unless I explicitly choose to.

Today the Atlas sidebar “Ask ChatGPT” flow effectively creates new chat artifacts (or branches) and leaves a durable trail. For quick “what does this mean?” moments, that’s too heavy.

Proposal

Add a first-class, lightweight Explain Selection feature:

  • User highlights text (or right-clicks a selection)

  • Chooses Explain selection (context menu) or uses a hotkey

  • Atlas shows an ephemeral popover (or small side panel) containing a concise explanation of that selection

Key principle

Ephemeral by default. This should not create a new chat conversation or branch unless the user explicitly chooses to “Save” or “Open in Chat”.

UX / Interaction

Triggers

  • Right-click context menu: Explain selection

  • Hotkey: e.g. Cmd/Ctrl + Shift + E (configurable)

  • Optional: “Explain” button that appears near a selection (like dictionary popovers)

UI

  • Default: small popover anchored near the selection

  • Must include:

    • Close (X)

    • “More detail” (expands within same popover)

    • “Open in Chat” (explicitly creates a chat/branch)

    • “Save to Explanations” (explicit archival location; see below)

Storage & Organization

  • Default behavior: no saving

  • Provide a dedicated archive location (e.g. “Explanations”) if the user taps Save:

    • Searchable list of saved explanations

    • Optional auto-tagging with source: page title, domain, timestamp

    • Option to export/share later

  • Optional setting:

    • “Auto-save explanations” (OFF by default)

Privacy / Data Handling

  • No user API keys required.

  • Clear indication of what text is being sent (ideally via a small preview when selection is large).

  • Sensible limits:

    • Send the selection + minimal surrounding context (e.g., sentence/paragraph) unless user asks “More detail”.

    • Avoid capturing password fields or sensitive form fields.

Scope

In scope (MVP)

  • Works for selectable text on web pages (including Atlas-rendered content / ChatGPT responses)

  • Works for most browser PDFs where text can be selected

  • Ephemeral UI and no chat creation by default

Out of scope (later)

  • “Lens” mode for images / scanned PDFs (box-select image region)

  • System-wide explain-anywhere (outside browser)

  • Deep note-taking / annotation system beyond a simple “Explanations” archive

Expected Behavior / Acceptance Criteria

  1. Highlight text → “Explain selection” shows a popover with an explanation.

  2. No new chat conversation/branch is created unless the user selects Open in Chat.

  3. Closing the popover leaves no trace by default (no new items in chat list).

  4. “Save to Explanations” stores the explanation in a dedicated archive separate from normal chats.

  5. “More detail” updates the explanation in-place without creating a chat.

  6. Works reliably on:

    • standard web pages

    • Atlas/ChatGPT rendered pages

    • common PDF viewers where text selection exists

Why this matters

This is a high-frequency reading workflow improvement: “micro-clarifications” shouldn’t require durable conversation artifacts. The current flow increases cognitive and organizational overhead for users who read a lot and already manage many chats/tabs.

Suggested Implementation Notes (optional)

  • Treat “Explain Selection” as a lightweight, ephemeral session distinct from chat threads.

  • Allow opt-in persistence (Save/Open-in-Chat) as explicit user actions.

  • Consider a dedicated internal “scratch explainer” channel that is not exposed as a normal chat unless promoted.