Cracking the Furby code: The path to evolve an icon
summary
Reverse engineers the original Furby firmware (a leaked 6502 assembly of ~296 pages) and shows that the “learning” is table lookup not ML, that 40 million units sold on clever tricks, and that a first‑generation bug can lock a Furby “sick” (making affected units worth hundreds of pounds). Rebuilds a working, cable‑driven Furby using Teensy microcontrollers, Pololu Maestro, MP3 trigger, 11Labs voice and 3D‑printed parts; demos blink, mouth and speak but remain fragile and reliability is the main unresolved problem.
AI-written summary · about
chapters (22)
- 0:00Furby's cultural impact
- 2:11Reverse engineering Furby
- 9:08Furby's original source code
- 11:36Furby's fake learning
- 13:24Furby's personality mechanics
- 15:34Understanding Furbish and naming conventions
- 17:22Furby sickness bug
- 18:04Furby Easter eggs
- 20:33Furby's low-cost emotion
- 23:02Furby's family tracking
- 26:32Experimental setup choice
- 27:16Live Furby control demo
- 29:47Furby's analog state machine
- 33:27Custom Furby version 4
- 35:17Furby motion system
- 39:30Future robotics costs
- 42:22Component coordination
- 43:46Serial communication basics
- 46:58Embedded threading solutions
- 48:41Pololu Maestro controller
- 50:53Processing IDE remote control
- 51:34Furby project next steps
related talks
28mLearnfun and Playfun: A Nintendo automation system
48mReprogramming Biology
57mDecoding Dolphin Communication
30mCFU Playground: Model-specific Acceleration on FPGAs
37mHow to Break AI Systems (Before Someone Else Does)
30mCracking the code: neuronal networks in the brain
41mUnlocking Any Door In 21st Century. Immersion In Bio-Security
43mBuilding Qrio, a Bot That Plays Videos for My Toddler