GHOSTED.GG The Digital Morgue for Dead Conversations

Ghosting happens when people don’t know how to reply. Ghosted.gg gives you the message that keeps the conversation alive.

Comment

Inspiration

I’ve personally experienced getting ghosted on dating apps countless times — conversations going cold, messages misunderstood, energy mismatches, or just missing the cultural and generational cues. It’s not just dating: similar dynamics happen with recruiters, work chats, and casual Gen Z-style conversations. I realized people overthink replies and need a tool to diagnose why a conversation died. GHOSTED.GG — The Digital Morgue for Dead Conversations was born from this frustration: a way to analyze ghosted chats with insight, humor, and actionable feedback.

What it does

GHOSTED.GG performs an AI autopsy on dead conversations:

Ghosting Probability Score (0–100%)

Cause of Death – the fatal misstep

Time of Death – the message where the conversation flatlined

Toxicology Report – all detected red flags

Autopsy Summary – dating psychology insights with spooky humor

Paste your chat, click “Perform Autopsy,” and instantly get a forensic breakdown of why the conversation failed.

How we built it

Frontend: Next.js 14 (App Router) + TypeScript + Tailwind CSS

Backend / AI: Kiro AI + GPT‑4 + Anthropic models (Sonnet 4.5) for agent design and conversation analysis

AI Pipeline: Structured prompts analyze tone, energy, patterns, and red flags, then return a JSON “autopsy”

Deployment: Netlify / Vercel for instant hosting

UX: Minimal, playful Halloween-themed UI with tombstone animations for visual storytelling

Everything is optimized for speed and clarity: paste → analyze → get autopsy.

Challenges we ran into

AI prompt design: Crafting prompts to consistently detect conversation energy, engagement patterns, and red flags while avoiding hallucinations from the model.

Tone balance: Combining dark humor with empathetic, actionable insights without coming off as harsh or robotic.

Context adaptation: Handling a wide variety of conversation types — dating apps, professional chats, Gen Z texting styles — and resolving scaffolding/context issues using Microsoft Amplifier agents, agent strands, and Kiro powers.

Model hallucinations: Preventing inconsistent or irrelevant outputs from GPT‑4 and Anthropic models, while maintaining a clear autopsy structure.

UI simplicity vs engagement: Keeping the interface minimal and intuitive while delivering a visually playful, Halloween-themed experience that enhances the autopsy storytelling.

Accomplishments that we're proud of

Fully functional AI conversation autopsy built in under 24 hours for Kiroween 2025 Hackathon

Novel UX combining forensic analysis with a playful, thematic UI

Structured, repeatable AI pipeline capable of analyzing messy human chats across contexts

Delivering actionable insights that help users understand why a conversation died

What we learned

Tone-aware, structured prompts and agent scaffolding dramatically improve AI analysis quality.

A fun, themed interface increases engagement, even for technical or diagnostic tools.

Users value tools that explain why something failed, not just provide generic outputs.

Speed, simplicity, and minimal friction matter more than feature bloat, especially for hackathon projects.

Integrating multi-agent orchestration (Kiro + Amplifier) helps resolve context gaps and enhances output reliability.

What's next for GHOSTED.GG The Digital Morgue for Dead Conversations

Add multi-platform conversation import (WhatsApp, Instagram, iMessage) for real-world chats.

Introduce multi-tone reply generation (flirty, professional, casual, humorous) alongside autopsy insights.

Expand “Conversation CPR” features to suggest ways to revive stalled conversations.

Build a browser extension / API for one-click autopsy and reply suggestions.

Enhance analytics dashboards to track ghosting patterns and trends across users and platforms.

Experiment with AI agent orchestration improvements to further reduce hallucinations and improve cross-context understanding.

Built With

  • agent
  • amplifier
  • and
  • anthropic
  • anthropic-sonnet-4.5
  • api
  • css
  • for
  • ide
  • kiro
  • kiro-ai-agents
  • microsoft
  • netlify
  • next.js-14-(app-router)
  • openai-gpt-4
  • orchestrationopenai-api
  • prompt
  • scaffolding
  • sonnet)git/github
  • strands
  • tailwind
  • typescript
Share this project:

Updates