The argument in one line.
Claude Code builds web applications autonomously by separating planning mode from execution mode, and combining that workflow with a CLAUDE.md rules file and todo.md project manager produces apps faster and more reliably than babysitting Cursor or Windsurf.
Read if. Skip if.
- A non-technical founder or solopreneur who wants to build a working web app prototype in under an hour without hiring a developer.
- Someone with basic terminal comfort (can open a command line) who's heard about Claude Code but hasn't installed or used it yet.
- A developer or coder who uses Cursor or other AI coding tools and wants to understand Claude Code's workflow advantages and specific power-user techniques.
- You need to build production-grade applications with strict security, testing, or compliance requirements — this covers rapid prototyping only.
- You're on a tight budget and the $20/month Claude Pro subscription (required for full Claude Code access) is a barrier to entry.
The full version, fast.
Claude Code is a terminal-based AI coding agent that builds applications autonomously instead of requiring line-by-line supervision, and three undocumented workflow habits separate beginners from power users. Install it inside Cursor so you can watch code change in real time, then anchor every session around Plan Mode (Shift+Tab) to draft an approach before any code is written, and switch to execution only after the plan looks right. Add a CLAUDE.md rules file that forces the agent to research first, write a plan to todo.md, check in before working, and keep every change as small as possible. Bounce ideas back to the agent for feature suggestions, switch to Opus when available, and trust it to execute while you work in parallel.
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.

01 · Hook + promise
Positions Claude Code as best AI coding tool, better than Cursor and Windsurf. Senior engineer metaphor established.

02 · What Claude Code is
Terminal-based autonomous coding inside Cursor, live demo of a command being issued.

03 · Pricing breakdown
Pro $17/mo Sonnet 4, Max 5x $100/mo, Max 20x $200/mo Opus 4. Recommends $17 to start.

04 · Install Claude Code in Cursor
Download Cursor free, npm install command, log in, move terminal to right panel.

05 · Build the calendar app
First prompt builds Next.js calendar; demonstrates todo-list approach vs one-shotting; YOLO mode; code writes itself.

06 · Secret 1 - Plan Mode
Shift+Tab toggles Plan Mode ON. Undocumented. Start in plan mode, get the plan right, then Shift+Tab to execute.

07 · Plan Mode demo
Asks Claude to plan a calendar UI redesign; Claude produces detailed plan including glassmorphism; host approves; auto-switches to execution.

08 · Secret 2 - /model opus
Type /model opus to switch to Claude Opus 4. Only for $100/$200 plan users.

09 · Secret 3 - CLAUDE.md and todo.md
Create CLAUDE.md rules file and todo.md project manager. Rules template shown on screen.

10 · Ask Claude what to build next
Ask Claude for improvement suggestions in plan mode; picks event management; adds drag-and-drop, edit, delete in one prompt.

11 · Recap and CTA
Three frameworks on sticky-note graphic: Plan vs Execute, CLAUDE Rules, Project Management. Subscribe CTA.
Lines worth screenshotting.
- Claude Code runs inside Cursor's terminal, combining the autonomous agent with a code editor that shows every file change in real time — giving visibility without requiring technical oversight.
- Plan Mode, activated with Shift+Tab, forces Claude Code to produce a complete plan before writing any code — the single most important workflow change for non-technical users.
- A CLAUDE.md rules file tells Claude Code your standards, preferences, and project context once — every subsequent session reads it automatically without re-explanation.
- A todo.md file that Claude Code updates as it completes tasks gives you a live project manager that tracks progress across multiple sessions without losing context.
- Babysitting is the primary cost of using Cursor — Claude Code eliminates it by working independently while remembering full context, which is the practical difference between an intern and a senior engineer.
- The pro plan at $17 per month gives access to Claude Sonnet 4, which is the second-best coding model in the world and sufficient for testing and early projects.
- Running Claude Code in Cursor's right-panel terminal while the file tree is visible on the left is the correct workspace layout for monitoring autonomous builds without interrupting them.
- A specific prompt — name the location, the audience, the visual style, and the conversion goal — produces dramatically better results than 'build me a website.'
- The install command is a single npm line that takes under a minute; the only follow-up is typing 'claude' in the terminal to launch the agent.
- The $200 plan is justified only for developers coding 8-10 hours per day — for everyone else, the $17 plan provides enough tokens to build real projects.
- Claude Code is the current best coding agent because it works at a larger scale with significantly less oversight than Cursor or Windsurf, not because it uses a different model architecture.
- Non-technical founders can build and launch money-making apps without writing code because Claude Code handles all technical decisions autonomously within the constraints the CLAUDE.md specifies.
Plan first, then execute — that is the whole trick
Three undocumented Claude Code workflow features — Plan Mode, a CLAUDE.md rules file, and a todo.md project manager — turn a beginner session into a structured, self-managing build process.
- The senior engineer metaphor sets the right expectation for how to interact with Claude Code — give a clear brief, then trust the tool to execute rather than micromanaging each step.
- Claude Code runs autonomously inside a terminal, reading the codebase, planning changes, and executing them without requiring approval at every step.
- The $17/month Pro plan on Claude Sonnet is sufficient to start; upgrading to a Max plan unlocks Opus, which is meaningfully better for complex multi-file projects.
- Installing Claude Code inside Cursor lets you see the code being written in real time on one side while issuing commands on the other — a more useful setup than a standalone terminal.
- Claude Code's biggest advantage is that it breaks tasks into a to-do list and executes them sequentially, rather than trying to build everything in one shot.
- YOLO mode removes the per-step approval prompt and lets Claude run uninterrupted — useful once you trust the tool and want to step away during a build.
- Plan Mode (Shift+Tab) is not documented but is essential — entering it before any major change forces Claude to produce a detailed plan for review before a single line of code is written.
- Getting the plan right before switching to execution mode is what separates good results from wasted tokens — approve the plan, then let it run.
- Claude's plans are specific enough to include design choices like glassmorphism and hover transitions — reviewing them before execution catches misunderstandings before any code is touched.
- Typing /model opus in the Claude Code terminal switches to Opus mid-session — only available on Max plans, but a useful option when a harder problem needs a more capable model.
- A CLAUDE.md file in the project root is included as context on every prompt — it is the place to set rules that apply for the entire project, like 'make every change as simple as possible'.
- A todo.md file that Claude updates as it works gives a persistent, readable record of what has been completed and what is next — it turns an opaque background process into a trackable one.
- Asking Claude what it would add next — in plan mode — surfaces ideas the user would not have thought of, and the best suggestions can then be immediately executed in the same session.
- The three-framework summary — Plan vs Execute, CLAUDE Rules, Project Management — is the repeatable workflow that applies to any project, not just the calendar app built in this video.
Terms worth knowing.
- Claude Code
- Anthropic's terminal-based AI coding agent that autonomously reads, writes, and executes code across an entire project with minimal human oversight.
- Plan Mode
- A Claude Code feature activated by pressing Shift+Tab that makes the AI propose a step-by-step implementation plan for review before writing any code.
- CLAUDE.md
- A markdown rules file placed at a project root that gives Claude Code persistent instructions about the codebase, coding style, and behavior for every session in that project.
- todo.md
- A markdown file used as a lightweight project manager where Claude Code tracks pending tasks, completed items, and next steps across a multi-session build.
- Next.js
- A React-based web framework for building full-stack web applications, supporting server-side rendering, static generation, and API routes out of the box.
- autonomous agent
- An AI system that can plan, execute multi-step tasks, and make decisions independently without requiring human approval at each step.
- Cursor
- An AI-powered code editor that integrates language models into a VS Code-like interface, often used alongside or instead of Claude Code for interactive coding.
- terminal
- A command-line interface on your computer where you type text commands, used to run Claude Code, install packages, and interact with your project.
- drag-and-drop
- A UI interaction pattern where users click and hold an element, move it to a new position, and release to place it — commonly used in calendar and kanban interfaces.
Lines you could clip.
“You are about to gain AI superpowers.”
“It literally feels like you have a senior engineer working for you.”
“When AIs try to one shot things, that is when they usually screw up.”
“The secret to Claude Code is planning and execution. This actually is not in any of the documentation.”
“I love to take any apps I pay for, like Google Calendar, and just build my own version. AI is actively saving me money.”
Word for word.
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
The bait, then the rug-pull.
In two seconds Alex Finn makes a promise most tutorials spend five minutes building to: you are about to gain AI superpowers. No hedging, no context-setting, just the outcome and then 23 minutes of live proof.
Named ideas worth stealing.
Plan Mode vs. Execution Mode
- Shift+Tab to enter Plan Mode
- Describe what you want Claude to build
- Review the plan Claude generates
- Shift+Tab again to switch to Execution Mode
Two-phase workflow: plan first so Claude thinks holistically, then execute. Avoids one-shot failures.
CLAUDE.md Rules File
- Think through the problem, read codebase for relevant files, write a plan to todo.md
- Plan should have checkable todo items
- Before beginning work, check me and I will verify the plan
- Make every task and code change as simple as possible
System prompt for Claude Code baked into the project. Applied as context on every prompt automatically.
todo.md Project Manager
CLAUDE.md instructs Claude to maintain a todo.md file with a running checklist of tasks. Enables passive oversight without babysitting.
How they asked for the click.
“Make sure to hit the subscribe button and leave a like if you learned anything at all.”
Standard subscribe ask at end plus teaser to advanced workflow video on-screen. No mid-roll CTA.





































































