The focus of this release was to strengthen the prompt voice on the blog and upgrade the language model.
The prompt voice on the blog was too weak and often vague, which made it difficult to convey precise information. I've made it stronger and more direct, so now it communicates exactly what I intend without any ambiguity. This matters because clear communication is crucial for understanding updates and changes.
I've also upgraded to gpt-4o, which improves the language processing capabilities of the blog. This change was necessary to ensure that the content generated is more accurate and relevant. With this upgrade, I can deliver more coherent and concise blog posts, making it easier for you to follow along with the project's progress.
Internal infrastructure changes were made with no direct user impact, for example, concurrency cancel-in-progress was changed to false, so bot-triggered workflow_run.
This also will be the first blog post on this beautiful project using hte picsum urls rand images, using the slug as an idempotency pattern, so any blog post has its unique image, how cool is that?
I've decided to switch from using the AI generated images simply because the joy of randomness viewing every time a different image is something else that creating every time a prompt for visuals, that process is helpful only for fixed media, like the favicon of a product, the fixed thumbnail
of a video (changing only the text every time, e.g., this article and this one.
The next release will be a fix release, having subject the trainer part of the app, with the polish of the repository issues with the old scope.
See you at the next release!