After over 50 days since the launch of my GPTs, I’ve gathered some key insights worth sharing.
Current Data:
- Official session usage: ~2,000
- Backend API action interactions: ~3,000
Insights and Learnings:
- Challenges in Relying Solely on Prompts for GPTs:
- The primary issue is prompt security. I’ve managed to access instructions of several top 100 GPTs through simple prompt hacking.
- Many popular GPTs have overly simplistic prompts, easily replicated and improved upon by general users.
2. Initial Traffic Surge vs. Sustained Growth:
- Top-ranking GPTs mostly derive traffic from their existing platforms (like Twitter, Reddit, etc.).
- Many GPTs initially gain popularity by capitalizing on trends, but struggle to maintain momentum due to a lack of engaging features. (My GPTs has exactly the same issue)
- Divergent thought: GPTs, as contemporary topic creators, can be effective marketing tools for other products.
3. The Potential of Action API:
- Actions are essentially APIs that can inject any data, controlling the workflow and responses of GPT applications.
- Storing data on private servers through Actions can significantly reduce the risk of data or prompt leaks.
- Future prospects: I foresee the emergence of platforms more tailored for ChatGPT than Zapier or RapidAPI, enabling easy integration for non-developers.
4. Exclusive Data Applications is the future:
- Most current GPTs serve as agents, solving specific scenarios via prompts and are easily replaceable, especially when GPT is evolving.
- Quality data can lead to more engaging user experiences, like structured legal regulations or industry rules.
- Current limitations in GPTs’ data retrieval abilities necessitate updates from Open AI or combining Actions for data processing.
- High-quality data also enables portability across platforms, reducing dependency on a single platform like ChatGPT.
5. The Early Days of ChatGPT-like Applications:
- Comparing to the early days of mobile internet’s App Store, the current user base of GPTs, estimated under 5 million, has huge room for growth, especially when segmented by language and region.
- ChatGPT is one of the platform, GPTs’ content and experiences can be adapted to other platforms or AI startups. I anticipate the rise of more platforms, leading to an explosion in user traffic.
Conclusion:
We are still in the early stages of ChatGPT-like applications, a field ripe with potential and opportunities for innovation and growth. Looking forward to exciting future.