Vibe-Coding as We Need to Eat Our Own Dog Food
8 days ago
- #AI
- #VC
- #Productivity
- The author describes their intense summer break spent 'vibe-coding' to transform their VC fund into an AI-native sandbox.
- They experimented with various AI coding tools and techniques, aiming to make their workflow more efficient, personalized, and collaborative.
- The author reflects on past challenges with CRM systems like Pipedrive, highlighting issues with data consistency and user engagement.
- They built a polymorphic knowledge base to aggregate unstructured data from various sources, leveraging LLMs for intelligence extraction.
- The author shares their current tech stack, including Supabase, Orq.ai, Mem0, Koyeb, and Shipfox, among others.
- They discuss the benefits and challenges of AI-assisted coding, including technical debt and the importance of documentation.
- The author emphasizes the need for proper testing and guardrails when building with AI, regardless of whether the code is written by humans or machines.
- They reflect on the cost of using AI tools like Claude, noting the high token consumption and the shift to pay-per-usage models.
- The author acknowledges the rapid evolution of AI coding tools and the competitive landscape, with new entrants like Qoder and Kilo Code.
- They conclude by stating that this is just the beginning of their journey, with plans to migrate data, build a knowledge graph, and refine their platform further.