Positive pAIrry

Positive Pairry is a gaming sidekick that turns toxic voice chat into good vibes. While you play, Pairry platypus, listens for insults and answers in a friendly voice with a nicety instead.

Comment

Inspiration

We were all sharing stories about PC video-gaming and streaming, and realized that we all had one thing in common, as gamers we have all dealt with internet trolls and very toxic game lobbies at some point. While some people can take it in stride, others end up quitting those games or streaming completely after a bad experience.

We then had the idea: "What if you could have an AI bot that listens to what is coming through, and responds to negative feedback on your behalf?"

What it does

Desktop app that takes in incoming audio, detects negative or toxic words and phrases, and automatically formulates a response on your behalf.

How we built it

We had 5 grokbot agents doing the work:

  1. pAIrry project manager - Arbiter of the project development plan
  2. Windows Native Bot - Expert in nuances on Windows
  3. MacOS Native Bot - Expert in nuances on MacOs
  4. Technical Writing Bot - Drafts our specs and documentation
  5. Pipeline Bot - takes in feature requests and determines how they align with the vision of the product before implementation.
  6. Cursor agents orchestrated by the pAIrry project manager bot handled the heavy lifting of development.

AI honorable mentions:

  • BananaNano (Helping with image editing inside of Google Slides)
  • Gemini (Gathering statistics and copy for presentation)
  • FigmaAI (Interface, Desktop App Icon, Landing Page)
  • AWS Bedrock (Transcription and Scoring, Response Generation)
  • ElevenLabs (Voice Options of Pairry)
  • Deepspace skills within Cursor (Demo App Landing Page)

We had 3 humans in the loop:

  • Michael (Full-Stack Developer)
  • Evan (Front-End UX Developer)
  • Sarah (Product/Project Manager)

Challenges we ran into

  • Lack of a variety of responses from Pairry, or responses that were not as contextual
  • It is hard to gauge intent of a message, as some insults are veiled and microaggressions are hard to spot.
  • Is this a feature or a bug? Running into bugs but finding they were actually features in their own way and could be used later.

Accomplishments that we're proud of

  • Fully Realizing the idea to an MVP level, and not talking ourselves out of it along the way
  • Having the most fun in the past weekend with programming that any of us have experienced in a very long time
  • We discovered real meaning in the idea. What started as a gag idea turned into real social value

What we learned

  • It's very hard to score toxicity (especially because of microaggressions), but there are datasets out there that have already done this we can leverage
  • It's hard to deliver the same user experience on two different operating systems.
  • Using Agents we can accomplish a lot in a short period of time when we converge on a goal.

What's next for Positive pAIrry

  1. Optimizing latency
  2. Monetization
  3. Distribution/GTM
  4. Additional Features:
    • Positive encouragement from Pairry to the user, while muting the negative speaker (we actually got this to happen as a bug/feature when testing)
    • Using more ElevenLabs features to customize Pairry's voice

Extended uses for this outside of video games:

  • Helping elderly folks (detects for scams and responds on their behalf)
  • Assisting inidividuals with disabilities by responding on their behalf on phone calls

Built With

Share this project:

Updates

Submission history