Portfolio docs for a SaaS, learning Rust through ai-memory, and bilingual posts
Wrote full English docs for the 3D print SaaS, dug into why ai-memory uses Rust, drafted architecture diagrams for itop, and shipped a batch of bilingual blog posts.
A day spread across documentation, language study, and content.
3d-print-saas-management-system
Turned the project into a portfolio piece. Wrote a polished English README plus docs, CLAUDE.md, AGENTS.md, and an architecture overview, all in English. Generated a README cover image and captured app screenshots to serve as visual documentation, and made the docs bilingual. Opened a PR with the lot.
ai-memory
Spent the day learning Rust through a real codebase. I read through ai-memory and asked the questions that actually teach you a language: why Rust here instead of something I already know, why there is no vector database in this project, why it avoids object orientation, and how it runs fast without a garbage collector while still being a high-level language close to C. Coming from JavaScript, TypeScript, and Java, the ownership model was the big unlock.
itop.com.br
Sketched the system. Asked for a diagram of the service architecture and the devops setup so the moving parts are easy to see at a glance.
tgmarinho-ai-website
Most of the content work landed here. Wrote blog posts in both pt-BR and English, including a draft built around the Karpathy LLM-wiki idea for agent memory, with a diagram illustrating it. I also locked in two writing rules across the repo: prose never uses the em dash, since it reads as an AI tell in 2026, and any text embedded in blog images must be in English even for pt-BR posts. Both rules now live in CLAUDE.md, AGENTS.md, the relevant skills, and memory. Generated covers for the posts, added a "reply on a tweet" path for comments, and cleaned up a post that had been created in the wrong repo by moving it to the right one.
On the side, a thought experiment: what a programming language designed for both LLM agents and humans might look like, with a small code sample to make it concrete.
Days like this compound twice: the docs make past work legible, and the Rust reading makes future work possible.