Voice on the blog, editorial queue, and daily recovery
A day of turning the blog into a fuller experience: native TTS, editorial flow, documentation updates, and a recovered daily pipeline.
The day centered on the site as a product. I kept treating the blog not only as a place for posts, but as a living interface: content, audio, documentation, automation, and publishing working as one system.
Voice reading
The clearest product step was the option to listen to posts through the browser Web Speech API. The idea is simple: let someone read or hear a post in Portuguese or English without depending on an external Text To Speech service. I worked through the controls, voice selection, speed up to 2.0x, and the integration with the post flow.
This kind of improvement looks small, but it changes the experience. A technical article stops being only a page. It becomes something the reader can follow at a different pace.
Editorial queue
There was also solid work on the editorial queue. I moved posts about leadership communication, AI code review, independent thinking with agents, scheduled loop engineering, and changes in the software process. Part of the work was writing. Another part was organizing dates, images, references, and PRs so publishing does not become improvisation.
In parallel, I updated the README and agent instructions to make the project clearer for anyone entering it, including other agents working in separate workspaces.
Daily and technical research
At the end of the day, I returned to the /daily pipeline itself. The cron was firing, but the root checkout was stale and still used an old extractor, which ignored Codex sessions. I ran the updated extractor, confirmed there was enough activity, and recovered this entry.
I also left research threads open around HTTP QUERY, GraphQL, REST, RPC, and infrastructure architecture. It was a day of keeping the system writing about itself, while the system itself got better.