Welcome to the factory
A new home for notes on agents, automation, and the small steps that make a software factory.
There is a lot of work between having an idea and getting useful software into somebody’s hands. Some of it needs a person. Some of it is a repeatable sequence waiting to be written down.
Factorize is built around a simple pattern: a trigger, some context, and an agent. A trigger starts the work. Context tells the agent what matters. The agent takes the next step.
This blog is a place to explore that pattern, share what we learn, and think out loud about building software factories.
Start with one useful loop
A factory does not have to start with a complicated pipeline. Start with one task you understand: what sets it in motion, what information it needs, and what a useful result looks like.
Written as an illustrative JavaScript sketch, that loop might look like this:
// A sketch of the pattern, not a Factorize SDK example.
async function runJob(event) {
const context = await gatherContext(event);
const result = await runAgent({
instructions: 'Propose a focused change and explain how to verify it.',
context,
});
return requestReview(result);
}
The interesting questions live inside those steps. Which context is actually useful? How do you decide when the result is good enough? Where should a person review the work?
What you will find here
We will use this space for notes on agent workflows, practical automation, and the decisions behind Factorize. Expect concrete examples, experiments, and lessons from building.
If you are new to the project, start at factorize.sh. If you are ready to work on a factory of your own, the documentation is the next stop.
Thanks for being here. Let’s make something useful.
Thanks for reading.
More notes from the factory →