AI summary by TwinMind
Rabbit R1: Barely Reviewable
Marques Brownlee
19m 53s
Device Overview & Design
- Rabbit R1: AI assistant in portable box form factor, designed by Teenage Engineering
- Size: approximately stack of Post-its, lightweight plastic cube
- Color: extremely bright orange (single color option), difficult to photograph accurately
- Physical controls: one button (right side for questions), scroll wheel, swiveling camera
- Screen: 2.9" diagonal TFT touchscreen with no auto-brightness
- Charging: USB-C port, SIM card tray for cellular connectivity
- Build quality: all plastic construction, basic camera/speaker, MediaTek chip (same as Moto G8 Power Lite ~$150 phone)
User Interface & Navigation
- Primary interaction: press/hold button to ask questions, shake device like Etch-a-Sketch for settings
- Navigation: scroll wheel + button selection (no touchscreen interaction for menus)
- Touchscreen limitations: only usable for keyboard in terminal mode
- UI problems: no back button, must scroll to top every time, two-handed operation for brightness/volume
- Response speed: consistently faster than Humane AI pin (very low bar)
Core Features & Capabilities
- Multimodal AI: voice questions + vision capabilities via swiveling camera
- Vision testing: successfully identified plants, food items, summarized emails/articles from screen
- Voice responses: average distance Earth-Moon query = 384,000km (example response)
- Basic limitations: can't set alarms/timers, record videos/photos, send emails, no built-in calendar
Battery & Performance Issues
- Battery: 1000mAh capacity, "brutally bad" life (~4 hours idle usage)
- Charging: 45 minutes from 0-100%, requires multiple daily charges
- Connectivity: requires separate SIM card for cellular (additional monthly cost)
- AI accuracy: still hallucinates, confidently provides wrong answers
Pricing & Value Proposition
- Cost: $200 (no subscription vs Humane pin $700 + $24/month)
- Unboxing: minimal - cardboard box, cassette-tape plastic container/stand, no cables/charger/manual
- Hardware quality reflects $200 price point
Large Action Model (LAM) Technology
- Concept: AI processes natural language into actions (vs Large Language Models for words)
- Theory: use apps like human would with mouse/keyboard, not API-dependent
- Current reality: only 4 apps available (Spotify, Uber, DoorDash, MidJourney)
- Access: via "Rabbit Hole" online portal for app authentication
- Performance issues: wrong song selection, DoorDash ordering errors reported
- Future promises: 800 apps trained (no UI built), generative UI capability, "Teach Mode" for custom actions
Industry Critique - Unfinished Product Trend
- Pattern across tech: sell first, deliver later via updates
- Gaming: full-price alpha versions with bugs
- Automotive: features "coming soon" via software updates
- Smartphones: major features delayed post-launch
- AI products: "borderline nonfunctional compared to promises"
- Review challenge: assess current state vs future potential
Market Strategy & Positioning
- Price advantage: $200 gamble vs $700+ Humane investment
- Data collection strategy: similar to Tesla autopilot approach - get devices in hands for real-world training
- Problem: lacks compelling "other reason" to purchase today (unlike Tesla's good EV fundamentals)
- Chicken-and-egg: needs users for training but limited current utility
Final Assessment & Recommendations
- Advice: "buy product based on what it is today, not future promises"
- Current state: limited functionality, poor battery, basic hardware
- Future potential: exciting AI assistant concept, but requires significant development time
- Competition: awaiting Google/Apple responses in AI assistant space
