Tag Archives: prompt engineering

Introducing SocialGPT

Posted on by

SocialGPT is a lightweight but powerful Chrome extension that integrates ChatGPT directly into your social media experience. For years, I’ve wanted a tool that would let me write smarter, sharper, more context-aware replies without opening new tabs or juggling windows. Every time I needed to draft a rebuttal or clarify a point, I wished for something embedded – something right there on the page.

Now it exists.

With SocialGPT, you can mark comment threads, automatically pull their content into an in-page editor panel, and generate AI-assisted replies in seconds – no reloads, no switching, no bullshit.

Source code: Bitbucket – GitHub mirror

Key Features

  • Context Marking – Highlight any number of elements in a thread to build a structured conversation context with block indexes (like [1], [2], etc).
  • Floating Reply Panel – A modal-less in-page editor where you can:
    • choose tone (e.g. cynical, friendly, brutally honest)
    • select response length (short, micro, extended)
    • switch models (GPT-4o, GPT-4, o3-mini)
    • input modifiers and custom instructions
  • Facebook-Aware – Automatically detects your profile name and injects it into the prompt for authentic replies.
  • Right-Click Access – Mark content or open the reply interface with a right-click.
  • Mark Mode Toggle – One-click switch to enable or disable GPT reading mode.
  • Response Modification – Use the Modify button to rework or fine-tune previous replies with new tone, instructions or shortened versions. This is especially useful after generation, since context and prompt fields are cleared upon reply.
  • Visual Loader – Subtle spinning loader shows when ChatGPT is generating content.
  • Fact Check Reminder – Prompts include reminders to validate and cross-reference controversial claims or disputed data before producing a final draft. Designed to prevent regurgitation of unchecked social media noise.

Tone Profiles

Organized into four categories:

  • Objective & Informative – neutral and formal, fact-based and concise, academic and precise, analytical and critical
  • Confrontational & Direct – critical and direct, cynical and sharp, aggressive and unapologetic, brutally honest
  • Satirical & Sarcastic – sarcastic and dry, snarky and dismissive, satirical and ironic, witty and clever
  • Approachable & Light – friendly and casual, conversational and soft

Ideal Use Cases

  • Rebuttals in comment sections
  • High-speed debate replies
  • Satirical or snarky thread injections
  • Public moderation with edge
  • Clarifying academic-style posts

Requirements

  • OpenAI API key (GPT-4 or GPT-4o recommended)
  • Chrome browser with extensions enabled

Created by Thomas Tornevall for real-world online interaction. Feedback and pull requests are welcome at either GitHub or Bitbucket.

Stay sharp – speak smart – strike fast.