Podcast Summary: Lenny’s Reads
Episode: "How I Built LennyRPG"
Host: Lenny Rachitsky (narrating for Ben Shi)
Date: March 18, 2026

Episode Overview

This episode is a special audio edition of Lenny’s Newsletter, focusing on the creative process behind "LennyRPG"—a pixel art, Pokémon-style role playing game built by Ben Shi, a non-technical product designer at Miro, using transcripts and data from over 300 of Lenny’s podcast episodes. The episode provides a detailed step-by-step walkthrough of Ben’s approach, tools, lessons learned, and notable challenges and victories from ideation to launch. Lenny also challenges listeners to build something new using his freshly released podcast and newsletter dataset.

Key Discussion Points & Insights

1. Genesis of LennyRPG and Data Release

• Background: Lenny previously shared over 300 podcast transcripts publicly, prompting a creative explosion in the community:
  • Parenting wisdom from PM advice, user research scripts, "anti-memes", bots, and more.
  • Favorite project: "LennyRPG" by Ben Shi—a video game where players encounter podcast guests and battle them with product questions, capturing them Pokémon-style.
• Lenny challenges listeners to create something innovative using his expanded, AI-friendly data dump, offering a chance to win a free year of the newsletter.
• Call to action: "Build something and let me know about it. I'll pick my favorite and give you a free one-year subscription to the newsletter. Just post a link to your project..." [01:41]

2. Ideation and Workflow

Speaker: Ben Shi (narrated by Lenny Rachitsky)

• Starting Point:
  • Inspired by product-moment timing principles ("If you get the timing right, you’ll find a real window of opportunity.")
  • Early idea: An interview-practice app—but dismissed as too stressful.
  • Pivoted to something fun: A Pokémon-style RPG with podcast guests as characters.

Step-by-Step Build Process

• "When I build apps with AI, I usually follow a very simple flow." [04:11]
  1. Define the Core Idea—Articulate the concept, sketch visuals for clarity.
  2. Create a Product Requirement Document (PRD)—Let the AI interview you for clarity instead of writing from scratch.
  3. Build a Proof of Concept (POC)
  4. Add Remaining Features
  5. Polish
  6. Ship

Tools Used

• Planning/Ideation: Miro, ChatGPT
• Coding: Claude Code, Codex, Cursor
• Image Generation: GPT ImageGen
• Quiz Generation: GPT-4o
• Game Engine: Phaser 3
• Database: Supabase
• Deployment: Vercel

3. From Idea to PRD

• Visualization: Used Pokémon screenshots and simple sketches to communicate the look/feel to the AI.

  "For highly visual products like this game, spending some time on visualization can help you get a solid sense of how you want the game to look and feel." [07:38]

• Tested content generation and pixel art workflow using ChatGPT and Lenny’s photo.
• PRD Writing Hack:
  • Used ChatGPT to "interview" him with targeted questions, answered verbally for clarity, then assembled all into a comprehensive PRD.

  "Verbally answering these questions also forced me to think through gaps and assumptions in my idea while giving the AI much better context than a one page description ever could." [11:05]

4. Proof of Concept Development

• AI Collaboration Model:

  • Claude Code: "lead engineer" (architecture, planning, solutions)
  • Codex: Executes detailed instructions and tasks
  • Composer: Handles formatting & simple scripts

• Prototype Tools: Initially started with RPG.js (quick setup) but switched to Phaser for greater flexibility required by quiz-based battles.

  "After a few iterations it became clear that RPG JS was not the right foundation... The more I tried to bend it, the harder it became for the AI to reason about the system cleanly." [15:40]

• Challenge Handling:

  • Used AI to find open-source templates (e.g., reusable Phaser map).
  • Maintained a markdown log of changes for stability and context.

  "One tip for complicated tasks... is to ask Claude Code to create a simple markdown file to log everything it tries so it can keep referring back and updating what works and what doesn't." [19:01]

5. Scaling Up: Processing Data & Assets

• Transcript Processing:
  • Enriched transcript text with metadata from RSS feeds.
  • Automated quiz question generation using OpenAI API.
  • Output: Convenient JSON files for game use.
• Avatar Generation:
  • Automated avatar sourcing by repurposing episode cover art.
  • Refined prompts to ensure consistent art style and quality.

  "That turned a very painful manual task into a one click process." [29:00]

  • Resolved edge cases, like multiple guests in one episode or playful profile image quirks (e.g., Adam Grenier’s rabbit ears).
• Background Music:
  • Used Claude Code to auto-find, wire, and adjust open-source music tracks for different game phases.

  "That part genuinely felt like magic." [35:32]

6. Game Mechanics and Polishing

• Game Loops & Mechanics:

  • Simple, low-stress but competitive: Three questions per opponent for XP; random bonuses and boss unlocks upon reaching XP thresholds.
  • Defeated guests join your collection; leaderboard syncs with Supabase database.

• Polishing:

  • Used Claude Skills to auto-review and QA the codebase.
  • Documented all UI/UX bugs and let AI fix them systematically.
  • Light SEO setup for shareability.

  "Once everything was written down, I let AI pick up the list one by one and fix them. This worked surprisingly well, especially when the issues were clearly described." [43:47]

• Final Step:

  • Deployed on Vercel, shared with Lenny and Slack community.
  • Received an enthusiastic thumbs-up from Lenny.

  "That was the nudge I needed to just ship it." [48:02]

Notable Quotes & Memorable Moments

• On letting AI conduct the PRD interview:

  "Instead of writing the PRD myself from scratch, I let the AI interview me… The prompt I used was: ‘Ask me questions to help you put together a brief PRD for the following web game…’" [10:24]

• On AI as a coding teammate:

  “I treat Claude Code as my lead engineer. It helps draft the implementation plan, think through architecture, and reason about product and design constraints.” [13:31]

• On pivoting frameworks:

  “After a few iterations, it became clear that RPG JS was not the right foundation...After talking it through with Claude code, I decided to stop forcing it and pivot.” [15:54]

• On the magic of AI automation:

  “That turned a very painful manual task into a one click process.” [29:00]
  "That part genuinely felt like magic." [35:32]

• On final encouragement:

  “I honestly wasn't even expecting a direct reply given how busy he is, but to my surprise I got a very kind and encouraging response from him. That was the nudge I needed to just ship it.” [48:02]

Timestamps for Important Segments

• 00:00 – 01:41: Lenny introduces the episode, the context around the data release, and his challenge to listeners
• 01:41 – 07:38: Ben Shi describes the origin of the LennyRPG idea and initial response to the data release
• 07:38 – 13:31: Step-by-step on ideation, visualization, and building a strong PRD with AI
• 13:31 – 22:45: Building the POC, picking and pivoting game frameworks
• 22:45 – 35:32: Scaling up, automated quiz/asset generation, handling edge cases, background music
• 35:32 – 44:24: Designing game mechanics, final polish, QA, UI/UX improvements
• 44:24 – 48:02: Final deployment, community feedback, and shipping

Episode Takeaways

• AI and open data unlock new creative frontiers for product builders—even non-technical ones.
• A clear idea, robust PRD, and methodical process—amplified by AI tools—let small teams or individuals build surprisingly complex and polished products.
• Leverage AI’s strengths for systematic, repetitive, or logic-driven tasks; focus your human effort on vision, creativity, and playtesting.
• Community feedback, even informal, can be the final push needed to launch.

To access the data or try building something yourself, visit lennysdata.com and share back using the links in the show notes.

End of Episode Summary