By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.

The Calum Johnson Show

Vibe Coding Expert: How to Build & Deploy a $1M App in the New Era of AI (Step-By-Step Guide)

Check out Notion, the best AI tool for work, right now at

Matt PalmerguestCalum Johnsonhost

Jul 21, 20251h 43mWatch on YouTube

Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.

Great conversations run long. Their best ideas don't have to.

uListen is free on iPhone, iPad, Android, Chrome, and Claude.

CHAPTERS

Jump to publisher timestamps

Chapter summaries

Why Now Is a Golden Era for Non-Technical Builders
Calum introduces Matt Palmer and frames the episode around building software without being a coder. Matt says viewers will learn how to turn ideas into something tangible that can be shared and might become a business.
- •Calum was inspired by people building profitable products in a weekend with AI tools like Replit
- •Replit's founder said building has been democratized so anyone with an idea can create software
- •Matt says you no longer need to know how to code
- •The episode promises a step-by-step path from idea to first product
Matt's Finance Job and the Fear of Being an NPC
Matt describes the repetitive Excel work at his finance job, including the daily Bloomberg terminal routine, and the discomfort of having his idea to automate it shut down. He explains his fear of being the same as everyone else and what pushed him to buy Python and SQL textbooks.
- •Matt studied physics and economics and went into finance after college
- •He manually copied a rate from a Bloomberg terminal into Excel every day at 4 PM
- •His bosses shut down his idea to automate the work
- •His deepest fear is being the same as everybody else, an NPC
- •He bought Python and SQL textbooks to teach himself automation
From Learning Python to Vibe Coding on Replit
Matt explains how grunt work used to need a person, and how AI tools now automate it. He describes joining Replit a year and a half ago and how adding AI to the workspace lets people go from idea to application without writing code.
- •Matt joined Replit about a year and a half ago, when you still had to know how to code
- •Replit has only been doing its AI-driven approach for about a year
- •New AI models made it possible to go from idea to application without code
Replit's Explosive Growth and Real-World Examples
Calum cites Replit's growth from $10 million to over $100 million in ARR in five and a half months. Matt explains why a working prototype in a morning is so powerful, and gives examples like Billy Howell and a two-person team that passed $10K MRR. He argues software has shifted from capital-intensive to costing mainly time and good ideas.
- •ARR went from $10 million at the end of 2024 to over $100 million in five and a half months after Replit Agent launched
- •Anyone with an idea can see a working prototype in 15 to 30 minutes
- •Billy Howell pitches app ideas to clients, builds them quickly on Replit and sells them
- •A two-person team posted that they passed 10K MRR using Replit
- •Building an MVP now costs mostly time and good ideas, though Uber-scale apps still need engineers
Step Zero: Finding Ideas by Observing Your Life
Matt says ideas come from being observant about inefficiencies in everyday life and writing them down to prototype later. Calum adds a notebook time-tracking exercise and Greg Isenberg's export-button thesis, and Matt says any software is a potential $100 million idea.
- •Notice inefficiencies in daily life, like workout tracking or to-do apps, and treat them as product ideas
- •Write ideas down and prototype them on the weekend in Replit
- •Calum describes logging how you spend your time in a notebook to expose friction
- •Greg Isenberg's thesis is that an export button signals a $100 million AI business
- •Matt says normal people built all existing software, so you can too
Picking an Idea: Bias to Action and Consistency
Calum asks how Matt chooses one idea from many. Matt says to pick the idea that excites you most, bias toward action, and value consistency over perfection. Calum shares a friend's story about perfectionism keeping work private.
- •Matt is a perfectionist but says consistency trumps everything
- •An hour a day for ten years beats one burnout day
- •Pick the idea that excites you most, build it, share it, then see if it can be a business
- •Calum says perfectionism kills many people's dreams
The Four-Step Framework and Planning the Zen Timer
Matt lays out the four steps of plan, prompt, perfect and publish, stressing that AI is imperfect and failure is part of building. He then plans a meditation app and breaks it into components such as a database, login, sounds and mobile friendliness. Calum asks how much planning is needed, and Matt recommends a simple MVP.
- •Four steps: plan, prompt, perfect, publish
- •Vague prompts like 'build me a workout app' tend to give broken results
- •Break the idea into pieces like database, login, sounds and mobile-friendliness
- •Understand concepts at a high level rather than implementation details
- •Start with a simple MVP that gives a quick win
Prompt Building 101 and Knowing When to Start
Matt turns the plan into a specific prompt and suggests using other AI tools like Claude to think through ideas. Calum asks how to know when planning is done, and Matt says that if you can write the idea in about five sentences, you should send it. Calum recalls Riley Brown's point that the process is non-linear.
- •Copy something existing, then do it better, as a way to find ideas
- •Use Claude, ChatGPT or Gemini to think through your plan before building
- •The sample prompt asks for presets, web research on techniques and a minimalist design
- •If you can describe it in about five sentences, send it
- •The process is not linear and you may loop back to earlier steps
AI Randomness, Failure, and the Agent Workspace
After sending the prompt, Matt explains that AI output varies from run to run and compares it to a .300 baseball hitter. He walks through the plan Agent generates and shows that Replit is a full cloud environment with files, shell and access to developer tools, even running on his phone.
- •The same prompt can produce results of different quality
- •Context such as prompts, links and images shapes what the AI produces
- •Agent searches the web and produces a visual preview plan
- •Replit gives the AI a real environment with files, shell and deployment tools
- •The workspace runs in the cloud and can be continued from a phone
Reviewing the Plan: Simplicity Over Complexity
Calum asks what Matt looks for when assessing the plan. Matt mostly sends it after a quick check for anything too complicated, and says to boil an idea down to the simplest thing that works and iterate.
- •Matt usually just sends the plan after a quick glance
- •He looks for things to remove, not add
- •Complex architecture unrelated to your vision is a red flag
- •Simplicity and consistency are the overarching themes
- •Starting with something that works makes it easier to tinker
Debugging with AI and Adding User Login
The first build has a timer that doesn't count down, so Matt describes the problem to Agent, which fixes it. He explains debugging as observation and context-giving, ties it to building momentum in life, then asks Agent to add Replit Auth login and session storage.
- •Describing exactly what is broken gives Agent useful context
- •Agent can inspect consoles and test fixes itself
- •Debugging with AI is the moment you start believing you can build anything
- •Matt ties building momentum from facing problems to life more broadly
- •Login and session data are added with a follow-up prompt using Replit Auth
Publishing the App in One Click
Matt tests login, shows the database, and deploys the app as zen-timer.replit.app. He explains that deployment is a snapshot of the environment and that Agent's checkpoints let you roll back files and data, which makes failure safer.
- •Login worked and the database holds users and meditation sessions
- •Deploying takes a snapshot and puts it on the internet
- •Options include a security scan and a custom domain
- •Checkpoints work like video-game saves and can roll back code and data
- •The deployed app is live and session tracking later proved to work
Design Tweaks with Screenshots, URLs and Figma
Calum asks whether the app can be restyled to look like bigger brands. Matt shows how to give Agent screenshots, URLs, scraped page content and Figma imports. He cautions that screenshots are limited, since fonts and image assets are missing, so detailed instructions and a willingness to start over give better results.
- •Paste screenshots or enter a URL to give design context
- •Scrape a page to pull in assets like an MP3
- •Figma import lets you bring designs in
- •AI struggles to replicate fonts and image assets from a screenshot
- •More detailed prompts produce closer results and starting over with new knowledge is fine
Getting First Users: Consistency, Simplicity and Building in Public
Calum asks what helps builders get their first users. Matt says it is hard, ties marketing to getting attention, and stresses consistency, simplicity and execution. They discuss building in public and authentic personal branding as ways to attract an early audience.
- •Being ignored is the hardest kind of rejection
- •Amjad's early AI idea flopped around 2020-2021 and worked two years later
- •A product must be better enough that people drop their old way of doing things
- •Building in public turns failures into content and builds a base of early users
- •Being authentic matters more than polish
A Challenge for Viewers: Build and Share an App
Calum asks what challenge Matt would give his past self in the finance job. Matt says to build an app on Replit and share it on social media. Calum offers to put the handles in the description and possibly get submissions retweeted.
- •Matt's challenge: plan, build and share an app on Replit
- •Practice the habit of going from idea to something that works to sharing it
- •Viewers can tag Matt and the show on Twitter
- •They hope Replit or Amjad might retweet good apps

Publisher timestamps

Get more out of YouTube videos.

Read along on your phone, search it beside the video, or ask your favorite AI app what was said.