Almeida Press is the public press desk and blog for my software work. The mission is simple: tools that keep AI work honest, inspectable, and continuable. This site brings the projects together without pretending they are all the same product, or that every experiment is finished.
A place for the work
A repository is a good place for code. It is not always the best first page for someone trying to understand why that code exists. Almeida Press provides an editorial route through those surfaces. The repositories remain the place to inspect implementation and current instructions.
Three questions connect the projects
First: did the work actually pass its test? fakegreen looks for changes that make a build appear green by removing or weakening the evidence. The lit/dark projects make status legible rather than burying it in a confident summary.
Second: what is an agent allowed to do? Super exposes system-administration tools through MCP with permissions and guardrails. Mirror Bot makes a human approval step part of a command’s path to execution. Lumen explores explicit goals, critique, simulation, and withholding.
Third: can someone else continue the job? The handoff language carries verbatim instructions, files to open, and a stop rule. Its second version adds a required repository and commit pin. Continuation needs material another reader can examine.
A curated desk, not a repository dump
The projects page groups a selected set by purpose. Other public work remains on GitHub. The Amália SDK project concerns Python tooling for Portugal’s open European Portuguese language model ecosystem. It is an active alpha, not a completed production platform. I work from Massachusetts.
What this publication will do
The blog will explain decisions, language rules, and the limits of tools. No customer list, benchmark, or endorsement is needed to make the work readable. Almeida Press is an index and a place for explanation: make the work easier to find, inspect, and continue.
Sources: Joshua Almeida on GitHub.
