The argument in one line.
Using five parallel Claude terminals, web agents, Opus 4.5, consistent Claude.md rules, iterative plan mode, custom slash commands, and end-of-session verification enables you to build significantly faster while maintaining code quality and security.
Read if. Skip if.
- A developer with some coding experience who uses Claude Code or wants to adopt it and is looking for a concrete, step-by-step workflow to ship faster.
- Someone building web apps or software projects who's heard Claude Code exists but isn't sure how to structure their development process around it.
- A developer who already uses Claude for coding but hasn't explored parallel terminals, web agents, or advanced Claude features like plan mode and custom commands.
- A technical founder or solo builder who wants to reduce time spent on repetitive coding tasks by parallelizing Claude across multiple workflows.
- You're a complete beginner with zero coding experience — this assumes familiarity with terminals, npm, servers, and basic web development concepts.
- You don't build software or web apps — this workflow is specifically for developers shipping code projects, not writers, marketers, or other fields.
The full version, fast.
Claude Code's creator Boris Czerny runs a parallelized terminal workflow that lets one operator move at the speed of a team. The method stacks seven habits: keep five Claude Code terminals open in a lightweight wrapper like Ghostie instead of VS Code, spin up five to ten Claude web or mobile agents to work tasks while you sleep or travel, default the model to Opus 4.5 for everything, maintain a project-level CLAUDE.md rules file that ships with every prompt, iterate inside plan mode with multiple back-and-forths before any code is written, codify repeat actions as custom slash commands, and close each session by prompting Claude to audit its own work for bugs and security flaws.
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
Title hook, authority borrow (Boris Czerny), promise: beginner to Claude Code master.

02 · Step 1: 5 terminals in parallel
Ghostie terminal demo split-pane Claude left / npm run dev right. 5 windows, one per feature. Memory argument vs VSCode/Cursor.

03 · Step 2: 5-10 Claude web agents
claude.ai Claude Code section — spin up agents before bed. Alex runs 3-4 nightly, wakes up to reviewed code.

04 · Step 3: Opus 4.5 for everything
Dismiss Sonnet cost-savings argument. Opus uses fewer tokens per task, better results. Set via /model Default.

05 · Step 4: Consistent CLAUDE.md
Rules sent with every prompt. Alex shows 6-rule template: think first, check in before major changes, explain steps, keep simple, never speculate on unopened files.

06 · Step 5: Plan mode + multi-iteration
Shift+Tab twice = plan mode. Multiple back-and-forth iterations before finalizing. Boris: most apps one-shot from good planning.

07 · Step 6: Custom slash commands
Ask Claude to write slash commands for repetitive tasks. Live demo /git. Meta-trick: ask Claude what slash commands would save time right now.

08 · Step 7: End-of-session verification
Prompt Claude to verify all work, use best coding practices, maintain good security. Security catch for vibe coders.

09 · Outro + CTA
Subscribe, notifications, Vibe Coding Academy plug, like ask.
Lines worth screenshotting.
- Boris Czerny (Claude Code's creator) runs five Claude Code terminals in parallel as his standard working state — not as an advanced technique but as baseline practice.
- Ghostie is a terminal wrapper that makes multi-window Claude Code workflows easier for non-developers while using far less memory than any GUI IDE.
- Running five terminals instead of five VS Code windows means the computer stays at normal temperature and speed even with five Claude sessions active.
- Cloud agents at claude.ai can be spun up before bed to work overnight, delivering reviewed code by morning without any active developer involvement.
- Opus 4.5 is Boris's model of choice for everything — his reasoning is that it is smart enough to use fewer tokens, making it more cost-efficient than smaller models.
- A consistent CLAUDE.md file is step four in Boris's workflow — not optional, not advanced — because rules sent with every prompt prevent the most common error patterns.
- Plan mode iteration means going back and forth multiple times before finalizing a plan, not accepting the first draft Claude proposes.
- Asking Claude to ask you questions during plan mode draws out requirements you forgot to specify — the interrogation step improves the plan more than rewriting the prompt.
- Custom slash commands store reusable workflow steps so you can trigger complex multi-step operations without retyping the prompt each session.
- End-of-session verification means asking Claude to confirm what was built matches what was requested — it catches divergence before it compounds.
- Spinning up web agents for mobile-use scenarios lets you direct Claude Code from a phone while away from the desk, keeping builds moving during idle physical time.
- The workflow Boris revealed is not theoretical — it is exactly how the person who built Claude Code uses Claude Code every day, which makes it the highest-signal source available.
Steal the stack. Ship overnight.
The man who built Claude Code runs 5 terminals + 10 web agents + a tight CLAUDE.md — you can copy this setup in an afternoon.
- Install Ghostie free. Split every project: Claude left, server log right. Open 5 windows. This is your new default.
- Set up overnight agents. Before bed, queue 3-4 tasks on claude.ai. Wake up to code. This compounds daily.
- Lock in Opus 4.5 as default via /model. Stop second-guessing the cost — the token efficiency argument is real.
- Write a CLAUDE.md for every project using Alex's 6 rules as a base. Rule 6 (no speculation on unopened files) is the most important bug-preventer.
- Plan mode is Shift+Tab twice. Never accept the first plan. Ask Claude what it needs to know before letting it write a line.
- Ask Claude to write your slash commands. Start with /git and the verification prompt. Then ask Claude what else it would automate.
- End every session with the verification prompt. Make it a slash command so you never skip it.
Terms worth knowing.
- Terminal wrapper
- A user-friendly application that runs on top of a standard terminal, adding visual features like split panes, tabs, and themes while keeping full terminal functionality.
- Ghostie
- A macOS terminal wrapper app designed for ease of use, supporting split-pane views so developers can run multiple processes (e.g., a dev server and an AI agent) side by side.
- Web agents (Claude)
- Cloud-hosted Claude Code instances accessible from claude.ai that execute coding tasks in the background without requiring a local machine to be running.
- Vibe coding
- A development workflow where the user describes features in plain language and an AI agent writes all the code, requiring little to no manual coding.
- Opus 4.5
- Anthropic's highest-capability Claude model tier, optimized for complex reasoning and large coding tasks at the time of this video.
- CLAUDE.md
- A markdown file in the root of a project that provides Claude Code with persistent rules, constraints, and context that apply to every prompt in that session.
- Plan mode
- A Claude Code mode (Shift+Tab+Tab) where the agent generates a step-by-step implementation plan and discusses it with the user before writing any code.
- Custom slash commands
- User-created shortcuts stored in the .claude/commands/ folder that trigger pre-written workflows inside Claude Code using a /command syntax.
- AGI (Artificial General Intelligence)
- A hypothetical AI system capable of performing any intellectual task a human can — often used informally to describe an AI that feels remarkably autonomous and capable.
Things they pointed at.
Lines you could clip.
“Claude code's plan mode is, in my opinion, the best technology in the entire world. It is as close to AGI as it gets.”
“You're not saving any money by using Sonnet at this point. Opus is so smart that you'll get way more out of fewer tokens.”
“So many vibe coders introduce security issues into their app. But by using this verification prompt, you'll make sure every session you walk away using best security practices.”
“Before I go to bed, I always spin up three to four agents, and by the time I wake up, I have code to review.”
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.
When the person who built Claude Code posts his actual personal workflow, you read it. Alex Finn did exactly that — pulled Boris Czerny's X thread, stripped it to seven steps, and proved each one live in the terminal.
Named ideas worth stealing.
Boris Czerny 7-Step Claude Code Workflow
- 5 Claude Code terminals in parallel
- 5-10 Claude web agents
- Claude Opus 4.5 for everything
- Maintain a consistent CLAUDE.md file
- Start in plan mode, go back and forth multiple times
- Use slash commands for common activities
- End every session: prompt Claude to verify its work
The actual personal workflow of the person who built Claude Code, posted on X and distilled into 7 steps.
Alex Finn CLAUDE.md Rules
- First think through the problem, read the codebase for relevant files
- Before you make any major changes, check in with me and I will verify the plan
- Every step: give me a high level explanation of what changes you made
- Make every task and code change as simple as possible
- Maintain a documentation file describing the architecture
- Never speculate about code you have not opened — MUST read the file before answering questions
Six rules for a CLAUDE.md that steer Claude toward minimal, explained, non-speculative changes.
How they asked for the click.
“If you learned anything at all, make sure to subscribe, turn on notifications.”
Standard YouTube trifecta (subscribe + notifications + like). Delivered over the completed canvas doc showing all 7 steps.







































































