Models For Their Strengths
by Zevi Arnovitz on January 18, 2026
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 from a simple assistant into a complete technical partner through a structured, repeatable process.
The "CTO Project" Workflow
Zevi's workflow follows a clear progression that any non-technical person can adopt:
-
Start with exploration in a chatbot environment before moving to code
- "I would recommend starting slow with a GPT project, beautiful UI, super simple, then maybe graduate to like a Bolt or a Lovable, and then go to Cursor"
- This gradual approach helps overcome the fear of code through "exposure therapy"
-
Create a virtual technical partner with the right personality
- Configure AI to be "the complete technical owner of the project"
- Instruct it to challenge your ideas rather than being a "people pleaser"
- "I told it I own the problem, I own how we want the users to feel, you're the complete owner of how this is gonna be built"
-
Use a structured, repeatable process with slash commands:
- Create Issue - Quickly capture bugs or feature ideas in Linear
- Exploration Phase - Analyze and understand the problem through questions
- Create Plan - Generate a detailed markdown plan with tasks and status trackers
- Execute Plan - Build the actual feature or fix
- Review - Have the AI review its own code
- Peer Review - Have different AI models review each other's work
- Update Documentation - Improve documentation for future development
Multi-Model Strategy for Better Results
-
Match models to their strengths:
- Claude: Best for communication, planning, and collaboration
- Codex (GPT): Excellent at solving difficult coding problems but less communicative
- Gemini: Superior for UI/UX design despite erratic thought processes
- Composer: Extremely fast for simpler tasks
-
Use models to check each other's work:
- "I will have each of them review the code and then what I do is I have a slash command called peer review"
- Position one model as the "dev lead" and others as "team leads" reviewing the work
- This creates productive tension: "Sometimes Claude Code will get really sassy and be like 'this has been raised for the third time and for the third time I'm telling you this is not an issue'"
Continuous Improvement Through Reflection
-
Conduct "postmortems" on AI mistakes:
- "When Claude will fail to do something or I'll see this really bad bug... I'll ask it what in your system prompt or tooling made you make this mistake"
- Update documentation and prompts based on these insights
- "Going back and even when you've succeeded looking and understanding what you did and what you could have done better is critical"
-
Make your codebase "AI-native":
- Include markdown files that explain code structure to AI agents
- Document areas of the code to help AI navigate more effectively
- Update this documentation after each project
Mindset Shifts for AI-Powered Development
-
Become a "10x learner" rather than trying to be a "10x PM"
- Use the slash command "learning opportunity" to understand technical concepts
- "I am a technical PM in the making... I want you to explain what we're currently working on using the eighty-twenty rule"
-
View AI as augmenting rather than replacing your skills
- "There's a misconception with a lot of PMs that the job is always having the right answers and being the smartest person in the room"
- "The role of the PM is... harnessing anything that can get us as quick as possible to delivering the right solution to users"
-
Own your outputs regardless of how they're created
- "If you put anything out there or show something in a product review and you say 'oh sorry that was built by AI,' that's your mistake"
- "I think if you use these intentionally and really take the time to understand how to use AI in the correct way it's one of the biggest game changers"