Buildiful Machines: Owning the Computer Again

A new direction from the lab

Over the past few months, much of our attention has been on software — particularly Thingy and the systems around it.

But alongside that work, another thread has been developing inside the lab:

Buildiful Machines.

What began as a small internal toolchain for running experiments is gradually turning into something more structured — a hardware and software environment designed specifically for building and running agent-driven systems.

Why this started

Like many builders today, we rely heavily on modern AI tools.

But we kept running into the same tension:

Most powerful tools assume you are working inside someone else's infrastructure.

Cloud APIs. Hosted platforms. Browser sandboxes.

These are incredibly useful — but they also create friction when you're trying to experiment quickly, automate workflows, or run systems continuously.

Sometimes you simply want a machine that belongs entirely to you.

A machine where the rules are simple:

• It runs your tools
• It stores your data
• It keeps working even when external services change

Buildiful Machines began as our attempt to create exactly that.

From experiment machines to a defined system

Initially, these machines were just development boxes used around the lab.

Small computers running Ubuntu, configured for testing AI agents, automation scripts, and experimental workflows.

Over time, we noticed something interesting.

Once properly configured, these machines became extremely useful — not just for development, but for real-world automation.

Tasks that previously required constant (and often repetitive) busywork could now run in the background.

Examples include:

• Monitoring systems
• Automated research workflows
• Long-running analysis tasks
• Small service bots

The machines themselves weren’t the novelty — at least not all of it.

What mattered was the environment.

Introducing Buildiful OS

To make these setups easier to reproduce, we started assembling a lightweight operating environment around them.

This environment is evolving into what we now call Buildiful OS.

At its core, Buildiful OS is intentionally simple:

• Built on top of Ubuntu
• Focused on stability and reproducibility
• Optimized for running autonomous tools and agents
• Designed to be understandable rather than magical

In many ways, it’s closer to a builder’s workstation than a consumer operating system.

The goal is not to reinvent Linux.

The goal is to provide a clean starting point for experimentation.

Why hardware still matters

Software is flexible. Cloud infrastructure is powerful.

But there’s still something uniquely valuable about physical machines you control.

A dedicated system can:

• Run continuously without interfering with your main computer
• Keep sensitive data local
• Provide predictable performance
• Serve as a stable environment for long-running agents

For many workflows, this turns out to be the missing piece.

Instead of constantly restarting experiments or juggling cloud dashboards, you simply let the machine run.

Where things stand today

Buildiful Machines and Buildiful OS are still early.

Right now, we’re focused on refinements:

• Testing hardware configurations
• Documenting setup procedures
• Running experimental automation systems
• Exploring what a reliable "agent computer" should look like

Some of this work is gradually surfacing as guides, tools, as well as physical machines others can also use.

Our focus (editorial and otherwise) remains the same:

Build carefully. Understand the system. Document the process.

Building in layers

One thing we’ve learned from past projects is that ecosystems grow best in layers.

Thingy connects the physical world to structured information.

Present explores presence and human connection.

Buildiful Machines explores something different:
infrastructure for builders and entrepreneurs themselves.

These projects serve different audiences, but they share a common philosophy:

• Tools should be understandable
• Systems should respect their users
• Technology should extend human capability, not replace it

Continuing the journal

As with our other projects, we’ll continue documenting this work openly.

Some updates will focus on software.

Others may explore hardware experiments, agent workflows, or the challenges of building reliable automation systems.

Not every idea will survive.

But the ones that do will be shared here.

Because this journal exists for one simple reason:

To show the work while it’s being built.