Fabio Franchino

Fabio Franchino

@fabiofranchino.com

Founder @ EloquentOps https://eloquentops.com

72 followers21 following197 posts

One of the things I learned writing on social networks is: asking for answers in the post just doesn't work. People answer/reply because they want to according to the post, not because the author asks for an answer (like "and you, what do you think?") ๐Ÿคท๐Ÿป

Read a book in the traditional way, page by page (on paper, possibly). Then, ask ChatGPT for a summary of the same book. After that, challenge both it and yourself on what you grasp further, to learn more and gain different perspectives, and possibly pursue further reading. It's powerful.

Time for another personal software. WhatsApp allows exporting the full chat with someone. I wanted to gain some insights from a time perspective. I ran @bolt.new, and after a few minutes, I had what I needed to play with. The speed of having an idea, making it, and then moving on is impressive.

At some point, there'll be tools that will easily allow regular people to fine-tune LLM models that will be plugged in somewhere, and that will generate any kind of competition (and markets) you can imagine.

Just finished a quick data research using LLM, its internal data, asking it to find on the web the missing or more refined data in order to crunch it, providing me the final answer as a brief comment supported by charts.

Maybe it's not 100% accurate, but even though it's 80% (and it's more), it's pure gold. We live in an era where, if you have ideas, you can quickly validate them with data, saving you a lot of time

The challenge of the moment while working with LLMs: a solid *automatic* way to "choose" between models, in particular between reasoner and non-reasoner models, in order to have speed for simple task, and quality for more complex one.