Graduating Beyond Opinionated Tools
by Zevi Arnovitz on January 18, 2026
Situation
In 2023, Zevi Arnovitz, a non-technical product manager, discovered AI coding tools after seeing a YouTube video about building apps with Bolt or Lovable. Despite having zero technical background, he immediately recognized the potential of these tools to give him "superpowers" as a builder. Zevi began his journey using simpler, more guided AI coding platforms but eventually encountered limitations as his projects grew more complex.
Actions
- Started with guided platforms: Initially used GPT projects, Bolt, and Lovable which were designed to make coding accessible to non-technical users
- Identified limitations: Found that these platforms were "very opinionated" about implementation approaches and made architectural decisions on his behalf
- Graduated to more advanced tools: Moved to Cursor with Claude Code as his knowledge and confidence grew
- Created a structured workflow: Developed a series of slash commands to guide his development process from ideation through execution
- Implemented multi-model review: Used different AI models (Claude, Codex, Composer) to review each other's code, leveraging their different strengths
Results
- Greater control: Gained the ability to make all technical decisions rather than being constrained by platform opinions
- Access to cutting-edge capabilities: Could immediately leverage the latest model improvements without waiting for platforms to integrate them
- Built production applications: Successfully created StudyMate, a platform for students to upload study materials and generate interactive tests
- Developed technical intuition: Through "exposure therapy" to code, gradually became more comfortable with technical concepts
- Achieved independence: Could build significant features without relying on technical collaborators
Key Lessons
- Graduated approach works best: Start with more guided tools (GPT projects → Lovable/Bolt → Cursor) to gradually build confidence with code
- Different tools serve different stages: "I graduated from each tool when I kind of outgrew it" - simpler tools are perfect for beginners but create constraints for more complex projects
- Model specialization is powerful: Each AI model has distinct strengths - "Claude is the perfect communicative CTO, Codex is the brilliant but uncommunicative coder, Gemini is the artistic but chaotic designer"
- Control vs. convenience tradeoff: More accessible tools remove guesswork but limit flexibility; advanced tools require more decisions but enable greater customization
- Learning opportunity in every challenge: When encountering limitations, use them as chances to deepen understanding rather than reasons to give up
- AI-native development is emerging: The future will likely involve code bases specifically designed to be more accessible to AI tools and non-technical builders
The evolution from guided platforms to more flexible tools represents a natural progression for non-technical builders as they gain confidence and require more control. This journey demonstrates how AI is collapsing traditional boundaries between technical and non-technical roles, potentially leading to a future where "everyone's gonna become a builder, titles are gonna collapse, and responsibilities are gonna collapse."