Skip to content
View joaquinjacu's full-sized avatar

Block or report joaquinjacu

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
joaquinjacu/README.md

👋 Welcome to My GitHub!

  • My name is Joaquín Jacubowicz 🎹.
  • I am a musician and producer from Buenos Aires 🎛️.
  • I build my own music software: audio-to-MIDI transcription, stem separation, synths and Ableton devices 🛠️.
  • I am currently getting into AI safety 🤖.

Check out my personal web page here with all my projects.

🚀 Featured Work

  • Albert: a game about reward hacking (code) 🎮. Made with Julián Szereszewski and Nicolás Waehner for Mangrove's Game Night AI safety hackathon. You play an AI model in training: the lowest score gets switched off, and the score only measures what the task's proxy can see, so the only way to survive is to game the reward. Each level is based on a documented case of specification gaming (OpenAI's CoastRunners boat, the robot that fooled its camera, coding agents tampering with their evaluator).

🎹 Music Tools

  • pink-trombone-plugin: a synth that imitates the human voice by modelling the vocal tract as a tube (VST3/AU, C++/JUCE). Unofficial plugin based on Neil Thapen's Pink Trombone.
  • transcribe-midi: link or audio file → stems → MIDI, using a different ML model for each kind of instrument.
  • multitrack-stems: splits a song into 13 stems that sum back to the original mix.
  • handcc: play Ableton with your hands through the webcam (MediaPipe hand tracking → MIDI).
  • voice-splitter: Max for Live device that splits a MIDI clip into one track per voice.

All built by directing AI coding agents (Claude Code) 🤖.

🧩 Other

  • Palabrero 16: a Spanish word game, sixteen Wordle boards at once.

Thank you for visiting my GitHub! Feel free to contact me at: joaquinjacubowicz@gmail.com 📫

Popular repositories Loading

  1. transcribe-midi transcribe-midi Public

    Link or audio file to MIDI: stem separation plus Transkun or muscriptor per part, with tuning and tempo correction.

    Python 1

  2. palabrero16 palabrero16 Public

    Spanish word game: sixteen Wordle boards at once, 21 guesses to solve them all.

    HTML

  3. reward-hacking-game reward-hacking-game Public

    Albert: a browser game about reward hacking. Survive by gaming the proxy. Made for Mangrove's Game Night AI safety hackathon.

    HTML

  4. pink-trombone-plugin pink-trombone-plugin Public

    Physically modelled vocal tract synth (VST3/AU, JUCE). Unofficial plugin based on Neil Thapen's Pink Trombone.

    C++

  5. multitrack-stems multitrack-stems Public

    Split a song into 13 true-level stems (vocals, drum-kit pieces, bass, keys, guitar) that sum back to the mix.

    Shell

  6. handcc handcc Public

    Webcam hand tracking to MIDI CC on macOS: MediaPipe hand gestures drive mod wheel, expression and filters in Ableton.

    Python