A few weeks ago I started building a project that became ExDeck.
The original idea was simple:
Why does creating a presentation still take hours when AI can already write, code, summarize, and reason?
Most AI presentation tools I tried had one or more of these problems:
Locked users into proprietary formats
Produced slides that looked good but were difficult to edit
Had limited export options
Generated generic content
Felt slow when generating larger decks
I wanted something different.
I wanted a platform where someone could describe a topic in plain English and receive a complete presentation that could be edited, customized, and exported as a real PowerPoint file.
That idea eventually became ExDeck.
The Goal
The goal wasn't to create another AI chatbot.
The goal was to build an actual productivity tool.
A user should be able to type:
Create a 15-slide presentation about renewable energy trends in 2026
And receive:
Structured slides
Speaker notes
Charts
Professional layouts
Editable content
PPTX export
PDF export
without spending hours moving boxes around PowerPoint.
The Tech Stack
I intentionally chose technologies that allowed rapid iteration.
Frontend
React
TypeScript
Modern component architecture
Responsive UI
React gave me flexibility while keeping the interface maintainable as features grew.
Backend & AI
For AI generation I integrated Groq.
The speed difference was immediately noticeable.
When users generate presentations they expect results quickly.
Waiting 30–60 seconds feels slow.
Groq helped reduce generation latency significantly and made the experience feel much more interactive.
Deployment
The entire platform is deployed on Vercel.
Why Vercel?
Because deployment friction kills momentum.
I wanted:
Fast deployments
Preview environments
Easy scaling
Great developer experience
Vercel delivered exactly that.
The Hardest Problem Wasn't AI
Most people assume AI generation is the hardest part.
It wasn't.
The hardest problem was structure.
AI can generate content.
Generating a coherent presentation is a completely different challenge.
A good presentation needs:
Logical flow
Consistent slide hierarchy
Proper sectioning
Balanced slide lengths
Meaningful conclusions
Without structure, presentations become walls of text spread across slides.
A lot of development time was spent improving presentation organization rather than simply generating more content.
Real PPTX Export
One feature I cared about from the beginning was real PowerPoint export.
Many tools generate presentations but lock users inside their ecosystem.
That creates friction.
People still use:
Microsoft PowerPoint
Google Slides
Keynote
So exports matter.
Users should own their content.
If somebody creates a presentation using ExDeck, they should be able to download it and continue editing anywhere they want.
That became a core design principle.
Building for Real Users
One lesson I learned quickly:
Developers often build features they personally find interesting.
Users usually want something much simpler.
People didn't ask for complex AI workflows.
They asked questions like:
Can I export this?
Can I edit this later?
Can I share it?
Can I translate it?
Can I customize the theme?
The most valuable features weren't always the most technically impressive.
Lessons Learned
- Shipping Beats Perfecting
There were dozens of features I wanted before launch.
Had I waited for everything to be perfect, ExDeck would probably still be unfinished.
Shipping early created feedback.
Feedback created priorities.
Priorities improved the product.
- User Feedback Is More Valuable Than Assumptions
Some features I thought would be popular barely received attention.
Other features users requested repeatedly.
The fastest way to discover what matters is to put something in front of people.
- Performance Matters
People forgive missing features.
They rarely forgive slowness.
A fast application feels better than a feature-rich application that takes forever to respond.
- Simplicity Wins
Every additional button creates a decision.
Every decision creates friction.
The best interfaces often remove features rather than add them.
What's Next?
ExDeck is still early.
There is a lot I want to improve:
Better templates
More export formats
Improved chart generation
Team collaboration
Advanced customization
Better AI workflows
The goal remains the same:
Make creating presentations dramatically faster without sacrificing quality or ownership.
Final Thoughts
Building ExDeck taught me that the challenge isn't making AI generate content.
The challenge is making that content genuinely useful.
AI is becoming a commodity.
User experience is becoming the differentiator.
If I can save someone hours of presentation work while still giving them complete control over the final result, then ExDeck is doing its job.
If you'd like to try it, check it out:
I'd love to hear your feedback, criticism, and feature ideas.
