Clickable British flag for changing the site language to English.
Clickable Finnish flag for changing the site language to Finnish.
Clickable Japanese flag for changing the site language to Japanese.

(click flag to change language)

← Back

Project Management for Media Environments - Hashidoi

06/10/2026Author: Joona
A custom built web platform to create, manage and share different kinds of projects.

The problem

I have a lot of fleeting ideas for different kinds of creative work. The issue is that I don't have a proper way to actually manage them, work on them and to keep track with them. As I come from an engineering background, I tend to think about every problem as structured, and that they can be solved with a proper system. I don't have much experience with project management, but I have done quite a bit of software, and specifically web development. When I couldn't find exactly what I was looking for, I decided to go down the route of vibe coding a custom platform for myself.

In the beginning...

I started with drafting the database layout: what kind of data I would need to store, and how it would relate to each other. I also started to think what software stack to use and how the user interface should look like. I ended up using what I've already become accustomed to: A Next.js frontend with Prisma ORM managing a PostgreSQL database. The platform is monolithic, as opposed to using microservices which I'd like to explore in the future. I think that a monolithic architecture is a good fit for my needs as I'm building a single-/few-user platform, and it's easier to manage. Because I knew that building this myself would've taken forever, I started using AI pretty heavily early on in the project. The exact tool I used was Google's Gemini-CLI (later renamed/upgraded to Antigravity-CLI).

The development process

As a bit of a surprise, using AI in the development process has been a very positive experience. I have to admit that I don't exactly enjoy writing code that much; I'd rather focus on the design and system operation. AI speeds things up dramatically, though it doesn't get everything 100 % correct. And I don't mean this in a way that "it can't read my mind, therefore it's not perfect", I mean that it might do something stupid like entirely neglecting user authentication, straight up showing private website content to anyone who has the URL. This means that a lot of experience with programming is still required to understand what the AI is actually doing, and to keep vigilant for any potential mistakes, even regarding basic stuff.


To be continued...
(I accidentally bundled the draft post with the deployment and I will finish writing this later)