Launch

Why 'Project Ember'?

I initially chose the name Ember for my personal AI assistant because I wanted something that felt human, without using a common name.

As time went on, ideas kept emerging for how AI agents could assist across different areas of my life, and I realised this might be worth documenting.

I started building out a team of agents structured like a company, each with their own roles and responsibilities. Naturally, every organisation needs a name.

I asked ChatGPT to help brainstorm ideas, and as expected it produced the usual oversaturated "AI startup" suggestions. I kept bouncing ideas around, but I found myself coming back to Ember every time.

That's when I landed on Project Ember.

The name resonated for a few reasons.

Ember is defined as:

A small, glowing fragment of a dying fire that remains hot enough to reignite into flame.

I can relate to that.

Most of us go through cycles of ideas and bursts of energy, only to feel like we've peaked and started to burn out as we push too hard. Yet even then, we're able to reignite if the right idea or inspiration strikes.

The second reason is more personal.

I feel like this is the result of years of experiments and iterations. I've had many different startup ideas and followed different paths in life, only to eventually realise they weren't quite right.

Project Ember (or perhaps Plan E) is my fifth iteration.

Why am I building this?

I spend a surprising amount of time on life admin. Managing calendars, remembering ideas, organising projects and finances, shopping grocery specials, tracking homelab infrastructure, and constantly switching context.

None of these tasks are individually difficult, but together they consume time and can be frustrating to keep track of.

Project Ember isn't about replacing people or building "AGI". I care about reducing my mental overhead by creating systems that give me time back.

My goal isn't a single assistant answering questions, we already have ChatGPT for that. I'm experimenting with a team of specialised agents that work together autonomously.

Rather than one generalist trying to do everything, I've assigned various agents different responsibilities, context and expertise.

Admittedly, my current team is heavily focused on technical operations. I come from an IT background and have been expanding my self-hosted infrastructure recently.

As the environment grows, so does the maintenance overhead. Updates, troubleshooting, monitoring, documentation and dozens of small tasks start to pile up.

I've become increasingly impressed with AI's ability to troubleshoot any problem I throw at it. Often getting to a solution quicker than I could've done myself. So why not hand over the repetitive maintenance work I spend hours doing every month?

What will I be sharing?

Each week I'll share the recent successes and failures with lessons I've learned along the way. It'll be fun to look back on how this project evolves, and perhaps it might inspire you to build your own virtual team.

No doubt these reports will include the experiments I ran and the architecture decisions behind them. Various agent workflows and automation ideas, as well as practical examples of where AI actually saved me time.

Next week, I'll introduce you to the team I've built so far. See you there.

Project Ember is still very much a project, but I'm excited to see if this catches.