AI CTO Challenges Ideas

Zevi Arnovitz's AI-Powered Product Development Workflow for Non-Technical PMs

Zevi Arnovitz, a non-technical PM at Meta, has developed a sophisticated workflow that allows him to build complete products without coding knowledge. His approach transforms AI tools into collaborative partners that handle technical implementation while he focuses on product vision.

The workflow consists of a structured series of phases, each with dedicated prompts:

Core Workflow Phases

• Issue Creation: Quickly capture feature ideas or bugs in Linear without disrupting current work

  • Uses a slash command that prompts AI to format the issue properly
  • Preserves context and details for later development

• Exploration Phase: Deep understanding of the problem before any coding begins

  • AI analyzes the issue and asks clarifying questions
  • Examines codebase to understand technical implications
  • Identifies key areas that will be affected

• Planning Phase: Creates a detailed implementation strategy

  • Produces a markdown file with clear, minimal, concise steps
  • Includes status trackers for each task
  • Documents critical decisions made during planning
  • Serves as reference for future development in the same area

• Execution Phase: AI implements the plan by writing actual code

  • Different models can be assigned to different parts (e.g., Gemini for UI)
  • Allows real-time collaboration and feedback

• Review Phase: Multi-model code review process

  • Has the primary AI model review its own code
  • Uses multiple AI models to review the same code independently
  • Implements "peer review" where models critique each other's findings
  • Simulates having multiple team leads review the code

• Documentation Update: Ensures knowledge is preserved

  • Updates documentation based on what was learned
  • Improves future code generation by documenting patterns and decisions

Key Principles for Working with AI

• Model Specialization: Use different models for their strengths

  • Claude: Best for communication, planning, and collaborative work
  • Codex (GPT): Excellent at solving difficult technical problems
  • Gemini: Superior for UI/UX design work

• Continuous Improvement: Refine prompts based on AI mistakes

  • When AI makes an error, ask "what in your system prompt made you make this mistake?"
  • Update documentation and tooling to prevent similar errors
  • Treat each failure as an opportunity to improve the system

• Learning Opportunity: Use AI to develop technical understanding

  • Dedicated slash command to explain complex concepts
  • Primes AI to explain using the 80/20 rule for technical concepts
  • Gradually builds technical knowledge through exposure

• Gradual Technical Progression: Start simple and build confidence

  • Begin with ChatGPT projects before moving to code
  • Progress to tools like Bolt/Lovable that abstract complexity
  • Eventually graduate to direct code editing in Cursor
  • Treat code exposure as "exposure therapy" for non-technical people

Broader Implications

• Role Evolution: The line between PM and developer is blurring

  • "Titles are gonna collapse and responsibilities are gonna collapse"
  • Everyone will become builders in some capacity
  • Non-technical people can now build significant products independently

• AI as Amplifier: AI doesn't replace skills but amplifies learning

  • Enables junior PMs to "play at a higher level" than normally possible
  • Provides more "reps" for skill development
  • Shifts focus from "having all the answers" to "delivering solutions quickly"

• Ownership Mindset: Always own your outputs

  • Never blame AI for poor quality work
  • Use AI intentionally with clear guidance
  • Review and refine AI outputs before sharing