Live Demo: SDD vs Vibe Coding in 10 Minutes
This is the demo that makes the difference tangible. Two approaches, same feature, same AI. The contrast speaks for itself.
Use this for conference talks, team workshops, or to convince a skeptical colleague that the spec-first approach is worth the two minutes it takes.
Setup (before the demo)
- Have a project with Planu installed in Claude Code
- Have a simple feature idea ready: "add dark mode toggle"
- Ideally have two terminal windows side by side
Part 1: The Vibe Coding Approach (3 min)
Say: "This is how most people use AI coding tools today."
- Open Claude Code in a project without Planu
- Type:
"Add a dark mode toggle to the settings page" - Watch Claude write code
- Ask the audience: "How do you know this is correct? How do you know it is done?"
- Point out: There is no checklist. No verification. No way to track that the right thing was built.
Say: "This works for prototypes. For production features, you need something better."
The issue is not that Claude wrote bad code. The issue is that you have no way to know whether it wrote the right code without reading every line yourself.
Part 2: The SDD Approach (5 min)
Say: "This is Spec Driven Development with Planu."
- Open Claude Code with Planu connected
- Type:
"I want to add a dark mode toggle to the settings page" - Planu creates a spec — show the acceptance criteria on screen
- Say: "Notice these acceptance criteria. The AI now knows exactly what 'done' means."
- Say to Claude:
"Approve this spec"— show Planu moving to 'approved' status - Claude implements the feature against the spec
- Run:
validate SPEC-XXX - Say: "This tells us whether the implementation matches the spec. Not vibes — verification."
The key moment: When you show validate passing, the audience sees that "done" is a binary state, not a feeling.
Part 3: Why This Matters (2 min)
Key points to land:
- "When something breaks in production, you look at the spec — not trawl through chat logs"
- "When a new developer joins, the spec documents what was built and why"
- "When you have 50 features in flight, Planu tracks all of them — not just the one you are looking at right now"
Close with:
"Planu does not slow you down. It makes your AI more reliable. That is the difference between vibe coding and spec driven development."
Q&A Preparation
"Isn't this just extra documentation overhead?"
The spec is the contract, not the docs. It takes 2 minutes to approve and saves hours of re-implementation when the AI builds the wrong thing. You would rather find out the spec was wrong before implementation than after.
"What if I am prototyping?"
For throwaway prototypes, skip the spec. Planu does not force anything — it is there when you need production quality. The workflow scales down gracefully.
"Does this work with Cursor, Windsurf, or other tools?"
Yes. Planu works with any MCP-compatible AI tool: Claude Code, Cursor, Windsurf, Zed, Cline, Continue, and more. One installation, all your tools.
"How long does writing a spec actually take?"
About 2 minutes for a typical feature. Planu generates the initial spec from your description — you review and approve it. The overhead is genuinely small compared to the verification it provides.
Adapting This Demo
For a 5-minute slot: Do only Part 2. Skip the Vibe Coding comparison and just show the SDD workflow working. The validate output is the most impressive part.
For a 30-minute workshop: Add a hands-on section where attendees install Planu and run through the workflow themselves. The npx @planu/cli install command gets them started in 30 seconds.
For a skeptical audience: Lead with the question "How do you currently verify that AI-generated code is correct?" Most people do not have a good answer. That question frames the entire demo.
Planu — Spec Driven Development for AI coding agents. Get started
Top comments (0)