Kata 1
Kata 1 — Startup & Tech Observations

Market Dynamics
Attention economy: Indians spending 5-6+ hours daily on video. Category saturation in music and food delivery apps — but how many new creators are people actually following? The scarce resource in this gold rush isn't content — it's trust and attention commitment.
Model commoditization: No MOAT in building a model anymore. Another one comes slightly better and drops prices. Models are commoditized. The value is in application layer, distribution, and data — not the model itself.
Strategic Principles
| Principle | Application |
|---|---|
| Small parallel bets with high fault tolerance | Don't bet everything on one path. Run multiple experiments, kill losers fast |
| Technology comes everywhere. Pointless fighting it | Adapt or die. Luddism always loses |
| All benefits are in scale | AI-generated avatar videos, content production — margin approaches zero at scale |
| Build units first, then ecosystem | Don't start with platform play. Build atomic value units, then connect them |
| Point solution vs ecosystem | Most people build point solutions (easier to scale). Aevy tried entire video ecosystem. Both paths valid, different risk profiles |
| Lollapalooza Effects | Need 4-5 forces going in your direction simultaneously. One force ≠ breakthrough. Five converging forces = unstoppable |
Predictions
- Text-to-3D getting better. 3D designers are where graphic designers were 3 years ago (pre-disruption)
- Games will be next big thing: upscaling + better hardware + lower cost to make + text-to-3D
- When you run enough experiments, your predictive power increases — you develop intuition for what works
Founder Reality
- Bootstrap is true hard mode. In funded companies, salaries come from VCs. Bootstrappers eat what they kill
- Jetstreams: Very hard to find people going very fast in life. If you find one, stay onboard with them. Their velocity compounds yours
The model commoditization point is central to ServaLabs' thesis. If models are commoditized, the moat is: (1) proprietary data (your private data), (2) distribution (servers in homes), (3) platform (OS + app store). This is why we sell hardware + platform, not models. The model is interchangeable. The infrastructure is sticky.