open source
We build in the open.
What we learn running agents doesn't stay secret: the systems behind our products are documented every week in the newsletter, numbers and mistakes included, and what's worth copying becomes an open repository.
Community
we learn from people who build in the open, and we pay it back in kind.
Transparency
methods, numbers and mistakes, published as they happened, no polish.
Open source
what's worth copying ships under Apache-2.0, yours to use without asking.
01 · projects
What we're opening
runs
cartógrafo
The repository opens together with its launch article in the newsletter.
- for the business
- It turns a business problem into a map of work that agents can execute. You declare the problem; it draws the route, runs it, and learns from every run. It's the piece that lets one pipeline serve different kinds of work, from software delivery to investment analysis, without rebuilding the process each time.
- under the hood
- A framework that draws, executes and evolves work graphs per problem class. The graph is frozen during execution and versioned between rounds; every capability is a contract, not a prompt; gates verify with their own evidence; changes stay proposals until a person approves them.
02 · who's behind this
Rafael Gomes
I've been in tech since 2003. I started as a developer and spent the last two decades running remote teams and delivery processes: flow, predictability, continuous improvement. With AI agents, those two halves became one. Orchestrating agents is management applied to software, and I build every day. Our products are the result; the newsletter documents the path.
03 · newsletter
Follow the work.
The systems behind these projects are documented every week in the newsletter, in English and Portuguese.
Want to run them in your team? Talk to us →