What if the best way to introduce a product was to let people play first?

I'm building aftrglo - a motion tool for designers who want their work to look good online without learning motion

The app is still in dev, so for now I made a coming soon page where people can join the waitlist

And since it's a tool for designers, I knew I didn't want a classic marketing landing page We all know how those work But it doesn't feel like me 🤪

So I made a tiny interactive playground instead: a halftone effect that reacts to cursor movement, with adjustable settings to play with Because designers love interactive little things Not being sold to And it worked!

The waitlist is growing with people waiting for the launch (which is coming very soon) and I've also been getting sooo many messages about the effect, the interaction, and how it actually works

I'm very into openness and creative support, so no gatekeeping here 👩‍🏫 ✍️ Here's the breakdown: 1. In Unicorn Studio, I have a scene with the logo image on the bg (the same one I use as the Instagram avatar) 2. On top of it, there's a God Rays effect that reacts to cursor movement 3. Plus a Bloom effect to make it even more glowy (because it's aftrglo, obv) 4. After that, outside of Unicorn Studio, I added a canvas overlay with a halftone effect

So the background is a live 3D scene, and the dotted pattern on top keeps reading and redrawing it in real time Like old newspaper print but inverted: Where the image is darker, the dots get smaller Where it's lighter, the dots get bigger

That's why the pattern breathes and moves in real time Magic! ✨

Play with it and join the waitlist if you want to see what I'm building 👩‍🍳 aftrglo.app

Goated plugin!

What it actually takes to build a multi-tenant edtech SaaS from scratch

Everyone wants to build a SaaS. Few people talk about what happens when 500+ schools hit your system at the same time, each expecting their data to be completely isolated.

I built EduPilotPro, an AI-powered school operating system serving schools across South Asia, the Middle East, and Africa. Here's what the "multi-tenant SaaS" part actually looked like:

Tenant isolation is the first real decision. Every school needs to feel like they own the platform. Their data, their config, their branding. But you're running one codebase, one database cluster, one deployment. We used PostgreSQL with row-level isolation on self-hosted Supabase. Each school gets its own logical boundary without the cost of spinning up separate infrastructure per tenant.

The 8:45 AM problem. Schools don't use software evenly throughout the day. At 8:45 AM, every teacher in every school marks attendance simultaneously. That's a traffic spike that would embarrass most startups. We used Redis + BullMQ to queue attendance writes and process them asynchronously, so the UI stays responsive even when thousands of classrooms hit the system at once.

AI agents, not AI features. We didn't bolt a chatbot onto a school ERP. We built 6 purpose-built AI agents: attendance, admissions, fees, communication, reporting, and scheduling. Each one handles a specific administrative function end-to-end. A teacher types "All students present except Ali who is sick and Ahmed who is 10 minutes late" and the attendance agent parses it, records it, and notifies parents. No forms. No checkboxes.

The stack: Next.js + React + TypeScript on the frontend. Node.js + self-hosted Supabase on the backend. PostgreSQL + Redis + BullMQ for data and queuing. Railway for cloud infrastructure.

What I'd tell any founder about to build multi-tenant SaaS:

Design your tenant model before you write a single feature. Retrofitting isolation is painful.

Assume your traffic will be spiky, not smooth. Queue everything you can.

If you're adding AI, make it do the whole job, not half of it. A half-automated workflow is worse than a manual one because now nobody knows who's responsible.

→ Full case study: EduPilotPro on Contra

Perfect