In partnership with

How AI Agents Differ from LLMs? In today’s newsletter, you and I will break down what Vibe Coding is, where it fits in the software lifecycle, and how leaders and engineers can use it today.

Today’s edition is based on my session at AI Engineer HQ

In today’s edition:

• AI Deep Dive— Vibe Coding Is Not Engineering

• Build Together— Here’s How I Can Help You

[Sponsor Spotlight]

It’s go-time for holiday campaigns

Roku Ads Manager makes it easy to extend your Q4 campaign to performance CTV.

You can:

• Easily launch self-serve CTV ads

• Repurpose your social content for TV

• Drive purchases directly on-screen with shoppable ads

• A/B test to discover your most effective offers

The holidays only come once a year. Get started now with a $500 ad credit when you spend your first $500 today with code: ROKUADS500. Terms apply.

Get started today.

[AI Deep Dive]

Vibe Coding Is Not Engineering

Most of us learned coding the same way

1. open an editor

2. write each line

3. debug

4. repeat

But the way software gets built is shifting fast.

With the rise of AI coding assistants, a new approach has emerged - Vibe Coding!

What is Vibe Coding?

Vibe Coding is writing software through natural language prompts instead of hand-coding every line.

For example, instead of writing boilerplate code, you might simply say:

The AI then generates the code, which you refine and integrate.

Think of yourself as the architect and AI as the junior developer doing the heavy lifting.

Where it shines:

• rapid prototypes and MVPs

• boilerplate or repetitive code

• ui scaffolding and integrations

Coding vs Engineering

To understand where Vibe Coding fits, let’s separate coding from engineering:

• coding = writing programs, functions, classes

• software = many programs combined into a product

• engineering = designing full systems (scaling, testing, deployment, infra, security)

AI can help with coding and parts of engineering, but full-scale engineering still requires humans.

There’s no “Vibe Engineering.”

That’s still on us.

Two Ways to Do Vibe Coding

There are two main approaches:

1) No-Code / Low-Code Platforms

Tools that generate entire apps or front-ends from a prompt.

Examples:

• v0.dev (by Vercel) - frontend UI with Tailwind/React

• lovable, Bolt, Brew, Replit AI - full-stack apps, backend + UI

• claude code, cursor, codeium - versatile code generation

Workflow example (V0) -

• prompt: “build a course website with testimonials, payment, contact form, and database.”

• output - react components + tailwind styling + backend boilerplate

• integration - github sync + vercel deployment

You can iterate by refining prompts or editing code live.

2) Coding Agents in Your IDE

These live inside VS Code or other editors.

They don’t build everything, but assist as you write.

Examples:

• Claude Code in VS Code (I prefer)

• GitHub Copilot

• Cursor

• Replit AI

What they do:

• suggest functions, APIs, classes

• add documentation and tests

• refactor existing code

• generate README and CI configs

You stay in control while the AI accelerates development.

They work best if you build the project in a divide and conquer strategy.

Divide the project into smaller parts and make it build those parts, not everything together.

The reason is the context window, as the code goes on and on, and it will lose the context, even if it has a 1 million token capacity.

Why Prompting Matters

Garbage in, garbage out applies here.

A weak prompt like:

Yields poor output.

A strong prompt like:

Produces a near-production-ready project.

Pro tip:

1. plan the project

2. divide the project it small parts

3. create all the prompts using ChatGPT

4. send it to claude code and let it build it for you

5. keep an eye on the system design of your project

Can AI Build Full Systems?

Not yet.

AI can generate parts of a system, but complete software requires:

• environment setup

• CI/CD pipelines

• API key management

• security and scaling

• devops, monitoring, testing

Today’s reality:

• AI helps you prototype fast

• Humans ensure it’s production-grade

Once code is generated by any AI tool, it’s very hard to debug it.

It's really a pain in the a__.

What Leaders Should Know

• don’t overspend on models

• invest in frameworks and infrastructure

• “how reliable is our orchestration layer?” matters more than “are we using GPT-4?”

• test langchain, crewAI, llamaIndex alongside vibe tools

• pick what fits your workflows

If you’re an Engineer

• go deep on one framework or coding assistant (claude code - I prefer)

• build secure, reusable tools (API connectors, safe DB access)

• focus on observability and orchestration

• that’s how you move from demos to production

Final Thoughts

The model vs. model debate is a distraction.

The real shift is happening in how we build and ship software.

Vibe Coding gives you leverage, prototypes in hours, not weeks.

But remember - AI is your assistant, not your replacement.

The winners will be those who master the plumbing

• security

• orchestration

• data integration

• reliable infrastructure

Stop obsessing over the engine.

Start learning how to build the rest of the car.

Until next time.

Happy Vibe Coding.

Just for fun

Want to work together? Here’s How I Can Help You

• AI Engineering & Consulting (B2B) at Dextar—[Request a Brainstorm]

• You are a leader?—Join [The Elite]

• Become an AI Engineer in 2025—[AI Engineer HQ]

• AI Training for Enterprise Team—[MasterDexter]

• Get in front of 5000+ AI leaders & professionals—[Sponsor this Newsletter]

I use BeeHiiv to send this newsletter.

PS: Do you want a tool analysis on Claude Code? Reply “Yes" to this email.