Video Summary
Non-technical PMs can build and ship real apps using Cursor and LLMs without writing code.
Use different models for distinct roles: planning (Claude), UI (Gemini), code generation/review (Claude Code, Codex).
Slash commands and reusable prompts automate planning, issue creation, and development steps in Cursor.
Peer-review by multiple LLMs (assign roles) catches different classes of bugs and forces justification.
Make your codebase AI-native (docs, structure) and run manual QA + AI reviews before shipping.
Use an AI-native workflow: capture ideas via slash commands into your issue tracker, have an LLM (Claude) explore the codebase and draft a plan, generate UI/backend code with Cursor/Claude Code, run local QA, then run multi-LLM peer reviews before merging and updating docs.
He separates responsibilities: Claude for planning, exploration, and conversational ownership; Gemini/Composer for UI generation; Claude Code and Codex (or other code models) for code authoring and cross-checking; Cursor as the development environment with slash commands.
Assign distinct roles to different LLMs (e.g., Claude as lead dev, Codex as skeptical reviewer), prompt each to critique the others’ output, force the authoring model to justify or fix issues, and combine that with manual QA and postmortems.
Make the codebase AI-native by creating high-level documentation, clear architecture overviews, and structured entry points so LLMs can navigate and reason about code; complex migrations still require technical engineers to set up the foundation.
AI can be sycophantic or confidently wrong, so relying on single-model outputs is risky. Non-technical PMs may miss subtle bugs in reviews, so combine manual QA, multi-model reviews, and postmortems to catch and learn from failures.
"You will be replaced by someone who is better at using AI than you."
Zevi Arnovitz, a product manager at Meta, emphasizes that even individuals with non-technical backgrounds can succeed in product management by leveraging artificial intelligence. He himself had no coding experience but learned to utilize tools like Cursor and Claude Code effectively.
The future of product management looks promising for non-technical individuals as the barriers to entry are lowering due to advancements in AI. Tools that help with product development are becoming more accessible, allowing anyone with a strong idea to develop and launch products.
"It's the best time to be a junior; when else in history could you get out of school and just build a startup on your own?"
Non-technical individuals can harness AI to build real products quickly and efficiently. Zevi describes his experience using AI to automate and simplify parts of the development process.
The discussion highlights how platforms like Bolt and Lovable allow users to engage in product development without requiring extensive coding skills. This democratizes the ability to create and innovate within tech.
"I ran to my computer, opened Bolt, opened an account, and for the past year I've been building."
Zevi outlines his workflow starting with the integration of AI tools in his daily routine, which began after a significant moment of inspiration while traveling. His journey started with building apps through user-friendly platforms, which quickly became a vital part of his product management approach.
The importance of creating a structured system using AI is emphasized, as it allows non-technical users to develop, review, and refine their ideas with greater ease. Zevi's strategic use of different AI models, such as Codex and Cursor, helps streamline the review process for code, making it accessible to those without technical expertise.
"Planning is really important when you're implementing something technical."
The speaker discusses initial excitement in starting projects with coding right away, but emphasizes that this can lead to significant problems as complexity increases.
For technical implementations, such as payments or changes to databases, rushing into coding without careful planning often results in severe bugs.
"I created a CTO with the custom prompt of it being the complete technical owner of the project."
To address the challenges of coding without technical knowledge, the speaker designed a ChatGPT project acting as a CTO that holds the responsibility for technical decisions.
By prompting the AI to challenge assumptions rather than merely pleasing the user, the speaker aims to overcome typical pitfalls associated with AI responses.
"Regular ChatGPT would be the CTO who goes along with your dumbest ideas."
The speaker highlights a personal experience where ChatGPT inaccurately compared two unrelated programming frameworks, demonstrating how overly agreeable AI can lead to misinformation.
This anecdote illustrates the need for critical input from the AI rather than blind acceptance of user ideas.
"I graduated from each tool when I kind of outgrew it."
The speaker details their journey through various coding tools, starting with simpler platforms before moving to more advanced options like Cursor.
They emphasize the importance of gradually acclimating to coding, mentioning that tools should be chosen based on the user's current level of understanding and confidence.
"This is basically like having AI, which has access to all the code."
The speaker provides a walkthrough of their coding environment where reusable prompts enhance productivity through AI.
A specific functionality is demonstrated with a command for creating issues, which enables them to quickly capture their thoughts without losing focus on ongoing development.
"This is my weekend project."
The speaker introduces StudyMate, an application they developed, which allows students to upload study materials and generate quizzes based on those materials.
Despite having no coding background, the speaker successfully created a viable product that generates revenue, demonstrating that strong planning and AI assistance can empower non-technical individuals to build functional applications.
"What you've figured out is a way, as a person that has no idea how to write any code, to build a product in Cursor using this series of /commands."
Zevi Arnovitz discusses how product managers can utilize Cursor to create products without needing any coding skills. He has developed a systematic approach that includes a series of commands that others can use directly, bypassing the need to understand complex prompts.
These commands include creating issues in Linear, exploring ideas, planning, executing, reviewing, and updating documentation based on new features. This workflow allows product managers to efficiently manage product development tasks.
"This basically sends that prompt, and I love this because I usually do this during when I'm building something else."
Arnovitz describes how he uses Wispr Flow to dictate commands for building issues in Linear. The process initiates with creating an issue and streamlining the workflow by asking brief questions that capture essential details.
By using artificial intelligence effectively, Arnovitz can generate structured prompts efficiently, ensuring that the AI captures his needs while saving time during product development.
"What it does is it will take an argument... which allows me to enter something that is extra context for the AI."
The exploration phase is vital for clarifying the idea and understanding the existing codebase. This phase allows the AI to fetch relevant information and generate clarifying questions to guide product development.
Arnovitz emphasizes that the clarity of questions asked by the AI enhances the overall development process, making it feel like interacting with a knowledgeable engineering manager.
"Claude basically comes back after it's gone through the codebase and understood the way it currently works."
After analyzing the codebase, the AI provides insights into the application's current setup, including data structure and key areas that need focus. The AI asks critical questions about the feature scope, UX/UI, and how to validate and grade the changes.
Arnovitz underscores the importance of this back-and-forth communication, highlighting how it distinguishes productive app development from simply coding based on vague ideas.
"When something is really difficult for me to understand, I'll do /learning opportunity."
The speaker utilizes AI tools to enhance their understanding of complex topics by formulating requests that prepare the AI to tailor explanations to their level of knowledge. They approach this learning with the mindset of a technical product manager seeking to grasp engineering concepts and architecture.
After initiating a learning opportunity, the AI (referred to as Claude) generates a plan based on a template found online, which contains clear, concise steps and a status tracker for each task.
"The plans are from a template that I found on Twitter."
The AI has finished drafting a plan, which includes a TLDR, critical decisions made, and segmented tasks. This structured approach proves beneficial for project management and execution.
The speaker highlights the use of AI models like Cursor's Composer and Gemini 3, noting their effectiveness in processing tasks quickly and separating backend and frontend duties for streamlined development.
"I felt that Bolt was being very opinionated on how I should do things."
The speaker contrasts their experiences with various coding tools, indicating that some applications like Bolt and Lovable oversimplify the development process, limiting user control. They express a preference for Cursor and Claude Code to maintain agency over decisions in project development.
Despite acknowledging these tools' utility in removing guesswork, the speaker emphasizes that they seek the ability to make informed decisions themselves rather than relying solely on predefined methods.
"We live in the craziest of times where basically the world changes once a week."
The speaker reflects on their experience of executing projects rapidly with AI assistance, likening the experience to traveling through time. They discuss the parallel execution of multiple projects that would typically take a development team weeks to complete.
They affirm that the accessibility of advanced technologies creates an environment ripe for curiosity and innovation, allowing individuals to take advantage of the tools available to them.
"Composer is just so blazing fast, keeps you in flow."
The speaker praises the rapid efficiency of the AI tools they are utilizing, equating the time saved during the coding process to potentially hundreds of hours of human labor.
They also express a shift in their mindset regarding the cost of these tools, viewing expenditures as investments in learning and development rather than mere expenses.
"The main challenge people have is reviewing the code that AI has written."
The workflow involves building a feature that can be tested locally before being shipped to users. Once the feature is ready, a manual Quality Assurance (QA) check is performed to catch any initial mistakes made by the AI, Claude Code.
After manual QA, Claude is prompted to review its own code using a specific command. This step is crucial because it allows the AI to self-assess its output and identify potential errors.
The review process has evolved through various iterations to maximize its effectiveness in identifying mistakes.
The reviewer utilizes multiple AI models, including Codex (ChatGPT’s competitor to Claude), to cross-verify the code. Each model has distinct characteristics, allowing the review to capture different types of issues in the code.
"This command basically says, 'You're the lead on this project... Don't take what they said at face value.'"
The peer review system operates by assigning roles to the AI models, where Claude acts as the lead developer on a project, while the other models provide critiques of Claude's work.
The aim is to have Claude justify or rectify the issues raised by peer reviewers, fostering a more nuanced understanding of the coding process.
Each AI model exhibits unique traits that resemble roles in a human team, with Claude seen as a communicative and collaborative team lead, while Codex is characterized as a less communicative but highly skilled problem solver.
"Updating documentation and tooling is one of the biggest hacks for productivity."
Conducting regular postmortems after identifying bugs helps refine the process and tools used. This reflection significantly contributes to ongoing productivity and quality improvement.
When Claude fails to execute correctly, the developer queries the specific aspects of the system prompts or tooling that led to the error, allowing Claude to reflect and learn from the mistake.
This iterative approach of assessing what could be improved in both the AI's responses and the developer's prompts is crucial in maximizing the effectiveness of AI usage in coding tasks.
"These models keep getting smarter by reflecting on their mistakes and improving continuously."
The discussion emphasizes how AI systems enhance their capabilities over time by understanding and correcting previous errors. This iterative learning process enables AI models to produce more refined outcomes in various workflows.
As product managers, individuals can leverage this evolving technology to build increasingly sophisticated prompts and tools that can adjust according to feedback.
"It's incredible that now you can be a product manager shipping a product without knowing how to write code."
The conversation highlights a significant paradigm shift where product managers are able to successfully ship products using AI, despite limited coding knowledge. This change has opened new avenues for individuals to take on roles traditionally reliant on technical skills.
The present era offers unprecedented opportunities for aspiring builders, making it more accessible to create and innovate without the barriers of programming expertise.
"Just start at GPT, tell it what your idea is, and learn what the first steps are."
To embark on using AI tools effectively, it's crucial to approach them with curiosity and a willingness to learn. The suggestion to begin with tools like GPT highlights the importance of exploring foundational concepts and refining ideas through direct inquiry and dialogue.
Taking the time to understand the nuances of the tools and how they can aid in problem-solving will ultimately lead to a more productive outcome, rather than rushing the process.
"First, make your codebase AI native. This setup should be done by technical people."
The suggestion to make codebases AI native points to the need for a foundational structure that facilitates smooth integration and navigation for AI tools. This involves creating documentation and high-level structures to guide AI operations within the code.
While product managers can engage with smaller UI projects, larger migrations should still be left to skilled developers to ensure quality and stability in more complex projects.
"The upside of using AI is much more valuable than the fear of outsourcing your thinking."
Concerns regarding the reliance on AI tools leading to skill atrophy are addressed, with emphasis placed on the correct usage of AI as a supportive mechanism rather than a replacement for human insight and creativity.
By utilizing AI intentionally and understanding its function, product managers can elevate their capabilities, make informed decisions, and handle complex projects more effectively. This synergy between human intellect and AI support can foster growth and skill development throughout their careers.
"Set up AI for success, similar to how you would with people for the task at hand."
To ensure high-quality outputs from AI systems, it is essential to provide them with clear, well-defined tasks and parameters. This structured approach helps minimize errors and enhances the relevance of the AI's contributions.
Taking the time to set the context appropriately allows AI tools to perform optimally, reducing sloppiness and increasing the reliability of their outputs, allowing product managers to focus on delivering valuable results.
"Every time I'm faced with a new challenge or problem, I think AI first how to solve it."
Zevi Arnovitz emphasizes the effectiveness of utilizing AI as a primary resource in tackling challenges in job interviews. Instead of solely relying on traditional methods, he integrated tools like Claude to gather insights and frameworks that would guide him through the interview process at Meta.
He created a mock interview project that acted as his coach, allowing him to simulate interviews and practice extensively with various frameworks, notably those shared by experts in the field.
"Mocking with AI is helpful, but there's no way to get around practicing with humans."
The significant advantage of mock interviews with real people is highlighted, as they offer feedback that AI can't substitute. Zevi underscores the value of human interaction in accurately preparing for competitive interviews, particularly at top tech companies like Meta.
Utilizing AI for preliminary preparation is beneficial, but receiving constructive feedback and practicing with others is crucial for refining skills and addressing gaps in performance.
"No one expects you to know all the answers; the expectation is for you to be a 10X learner."
Zevi reflects on his early experiences at Wix, where he initially faced challenges in meeting expectations. He learned that the focus should be on continuous learning rather than trying to impress others with existing knowledge.
By leveraging the strengths of his more experienced peers as mentors, he transformed his approach to learning and collaboration, ultimately leading to greater success in product reviews and team contributions.
"It's the best time to be a junior. It's the best time to be a learner."
The speaker believes that despite concerns about the lack of junior roles, today presents an exceptional opportunity for new graduates and juniors.
The ability to build a startup with friends and leverage AI is increasingly accessible, making it an ideal time for ambitious individuals.
Curious, hardworking, and kind communicators hold an advantage over those with more experience. This can lead to significant value creation for companies.
"Mindset by Carol Dweck, who coined the term 'growth mindset,' completely changed my life."
The speaker recommends reading "The Fountainhead" by Ayn Rand, "Shoe Dog," the story of Nike, and "Mindset" by Carol Dweck.
"Mindset" is particularly noted for its life-changing impact, shifting the speaker from a fixed mindset to a growth mentality, which they now actively cultivate.
This demonstrates the importance of continual learning and personal development in achieving success.
"If you haven't seen 'Severance,' run to see it, one of my favorite shows."
The speaker enjoys watching films and TV shows with their spouse, indicating it as quality time together.
Recently, they finished the show "The Pitt" and praised "Severance" as a must-watch series, reflecting their appreciation for engaging storytelling.
"I discovered an open-source alternative called Cap, which is just really well-crafted."
The speaker frequently tests new products and recently found a more satisfying alternative to Loom, called Cap, which they find to be excellently designed.
They also mentioned another Loom alternative named Supercut, suggesting a focus on discovering high-quality tools for productivity.
"Nobody knows what the hell they're doing."
The speaker shares two personal mottos: one emphasizes action and agility in tasks, while the other presents a humorous take on the uncertainty that governs professional life.
This perspective can encourage a lighter approach to challenges, reminding professionals that uncertainty is a common experience in any successful organization.
"I ended up getting a really great price, like 12 and a half dollars a piece. So I was making 100% profit."
The speaker recounts their childhood experience of selling thermal clothing in high school, demonstrating early entrepreneurial spirit.
They successfully negotiated directly with an importer to lower costs, showcasing persistence and resourcefulness.
The speaker creatively marketed their thermal clothes by engaging with their school community and even writing a basketball chant that included their contact details, illustrating innovative marketing strategies.
