How I use LLMs.

Intro

Although slightly averse to the use of LLM, I have started to appreciate their effectiveness in automating tasks which I would simply not like to solve myself. I want to quickly sketch how I use LLMs & agents these days.

LLM Use Cases

Exploration

Exploration of concepts realted to programming, getting a quick grasp, replacing the initial Google search preliminary to many tasks.

Tasks which are out of my ‘main’ field of interest

I would consider myself a rather curious person. Especially with regards to technical topics, my interests are quite diverse. Embedded systems, Machine/ Deep Learning and Reinforcment Learning, Electronics, Devops topics, ….

However, I am starting to realize (rather late, I know) that you cannot learn & know everything and that trying to do so leads to you not knowing anything at all. Focus is key and this is probably the hardest part of being a part-time phd who is working the rest of the time of the day - keeping the focus and knowing what to focus on.

Anyways, sometimes, creating things in the field you are interested in require you being active in a field you are not that much intersted in. Like this blog post. I would like to have a platform where I can create small writeups of my research and learning but I do not like to fiddle with the configuration / setup of the website and such things. I would simply like to write up my thoughts in a .org file and leverage this to publish it at a website. Without LLMs, this would cost me hours, if not days to understand the ecosystem and how the components interface. With LLMs, it’s a matter of 1-2 hours at most.