Posts

I love listening to podcasts, and news in general. For a couple of years, I’ve been subscribed to a somewhat-niche (but really high quality) Italian weekly magazine. I like their Android application, and I use its Text To Speech functionality, but there’s a big problem with that: it’s impossible to increase the audio reproduction speed. I usually listen to articles while doing something else, such as washing the dishes, or commuting to the office....

If you’re a bookworm like me, you can’t resist the fascination of the paper medium. Reading a proper book is a much more relaxing experience than reading on a screen. Some studies also found a correlation between phone usage in bed and increased risk of having poor sleep quality. These are some reasons I’ve always preferred reading from paper books, especially at nighttime. Computer Science books are often expensive, and one could always object that paper production has an undeniable carbon footprint....

Recently, we had a couple of naming-related incidents at work. We respect certain naming conventions, but sometimes errors slip into our Terraform infrastructure’s code, and remediation may be costly. I had to find a way to simplify resource naming policy enforcement: at first, I thought about using Git Hooks (with a custom HCL parser) or local linters. Both of these approaches came with downsides that would make it unpractical to employ them on multiple repositories: first of all, the necessity to set up hooks or linter configurations in each repo....

A friend asked me to create a Telegram channel to cross-post my articles onto (I did! You can find it here). The idea seemed nice, but I didn’t want to bother with writing a small summary for each article every time, and most of all, forgetful as I am, I would absolutely forget to post to the channel. I only had one choice: automate everything using GitHub Actions, Azure Function Apps, and Azure OpenAI, trying to have fun and learn something new in the process....

It’s impossible to keep up with the rapid developments in the field of LLMs. On Feb 21, Google released a new family of models called Gemma. These models promise top performance for their size. According to the Gemma technical report, the 7B model outperforms all other open-weight models of the same or even bigger sizes, such as Llama-2 13B. 💡According to the technical report accompanying Gemma’s announcement, the new models were trained using the same methodologies as Google’s top-tier Gemini models...