Afterthought

Your thoughts between prompts.

Field notes

The difficult work starts after the answer arrives.

AI can create a detailed response in seconds. Understanding its claims, preserving your questions, and deciding what deserves trust still takes a deliberate second read.

  1. Three different AI response papers meeting at one human-centered annotation desk
    6 min read

    A tool for annotating ChatGPT, Claude, and Gemini responses

    Afterthought gives long AI responses a reading margin: paste from any assistant, attach Thoughts to Sources, and copy one contextual follow-up.

    Read article
  2. A technical plan showing a broken bridge, risk markers, and a safer alternate route
    9 min read

    How to review an AI-generated technical plan for risks and assumptions

    AI plans often sound complete before they are safe to execute. A Source-by-Source review separates useful direction from unsupported confidence.

    Read article
  3. Scattered annotated paper fragments converging into one ordered follow-up sheet
    7 min read

    How to turn comments on an AI answer into one follow-up prompt

    Your notes become useful when the next AI turn can see what caused each one. A contextual follow-up organizes them without changing your words.

    Read article
  4. A hand connecting individual manuscript passages to precise cobalt margin marks
    7 min read

    How to give passage-by-passage feedback to an AI assistant

    Whole-answer feedback forces the model to guess. Passage-by-passage feedback preserves the connection between each judgment and the Source that earned it.

    Read article
  5. A reader following a long folded response with question markers along the path
    8 min read

    How to digest a long AI response without losing your questions

    Long AI responses create a second problem: remembering every question while you are still learning the answer. A structured second read keeps that complexity manageable.

    Read article
  6. A reader making cobalt margin marks on a long paper response
    7 min read

    How to annotate a ChatGPT response before replying

    A fast answer still creates reading work. Annotating before you reply keeps useful ideas, doubts, and questions attached to the passages that caused them.

    Read article

Worked examples

See what a better second read looks like.

Each example starts with one concrete AI recommendation and ends with the exact contextual prompt you can send back.

  1. 01

    Review a technical response

    Keep implementation questions beside the exact claim that caused them, then return one ordered follow-up.

    Read example
  2. 02

    Annotate a risky migration plan

    Separate what is useful from what needs a rollback plan before an AI recommendation becomes implementation work.

    Read example
  3. 03

    Challenge a premature solution

    Turn “I am not convinced” into a follow-up that asks for evidence, boundaries, and a measurable decision.

    Read example