I’m not a professional software person but I’ll offer my two cents as a no-LLMer:
I first came to HN in the “todo.txt” era of “productivity hacking” and note-taking -platforms like Evernote. Like many people I had a zettelkasten phase, tried to make a second brain, tried to optimize everything blah blah blah.
Over the ensuing 15 years and several career shifts later, it’s fascinating to see how AI as supplanted so many of these tools. However in my personal case, greater professional success has coincided with discernment, i.e., knowing which information is important to internalize and commit to memory, which can be filed for reference, and which can be allowed to fade away or be forgotten.
In my current work, there is a huge amount of information that I really, truly need to know “by heart” to do my job well. There’s an equal portion that I maintain in traditional reference files with reliable retrieval systems. I do use machine learning for certain field tasks, but over time I have been able to learn to do these tasks myself when an internet connection is unavailable.
No LLM tool thus far appears useful for me. One big reason is that I work in a compliance/regulatory space where hallucination is simply unacceptable. If I have to check the output for errors, I may as well just look at the primary source to start with.
Another reason is that in regulatory settings, people will say in filings/documents that they are obeying XYZ law, but it isn’t true. I need to find out *in the field* whether the assertions are true. LLMs are not useful for that, either.
But I think the largest gap is between LLMs’ product promise and my personal professional goals. I want _wisdom_ and clinical experience as a professional, the type of things that accrue slowly over a lifetime and distinguish the people who are truly good at their jobs.