Journaling with an LLM AI: as good as therapy, at least for me

Over the past month or so, I embarked on an experiment with AI-driven journaling using Claude, a Large Language Model. Here's a brief rundown of my journey:

* Choosing Claude: Its impressive context window, around 25 times larger than GPT-4 (handling ~75k words compared to 3k), was a big draw. This larger window allows Claude to recall and reason with content spanning my entire journaling history, offering insights that align with my previous entries — even if they were written a month ago.

* Experiment Set-up: My goal was to gauge the kind of feedback an AI could provide in the realm of personal journaling.

* Routine: Almost daily, I spend around 10-20 minutes crafting a 3-4 paragraph entry.

* Feedback Quality: The insights and feedback I've received from Claude resonate on par with what I've experienced in professional therapy sessions.

* Frequency: The ease and immediacy have led me to journal multiple times a week, which is significantly more frequent than traditional therapy sessions. This regularity, in turn, has provided me with timely and continuous insights into my personal life.

* Price/accessibility: Many people cannot afford quality access to mental healthcare. This is a tool that anyone can access for free.

The idea of an AI substituting for human-led therapy might seem unconventional for many. Yet, my experience has shown that, at least in the context of journaling, there's enormous potential.