3 ms·
Ask HN: Anyone using LLMs as coach or therapist?
Is anyone else using LLMs as a coach or therapist. I'm currently exploring local LLMs as a way to enhance my journaling practice with a coach or therapist. Would love to get some inspiration and ideas on how others are doing this and which tools you use?
- toomuchtodo 1y agoA friend is using ChatGPT as a kink and poly aware pseudo therapist, keeping in mind the limitations of the system, considering the challenge of finding a human version tuned for such a lifestyle.
- dacke 1y agoInteresting, no worries with privacy when using chatgpt?
- toomuchtodo 1y agoThis is not a concern they have. YTMMV (your threat model may vary, n=1). I offered to wire up ollama or deepseek for them and they shrugged. If you want to host local, consider https://github.com/OpenHealthForAll/open-health https://github.com/OpenHealthForAll/open-health and https://github.com/johnbean393/Sidekick https://github.com/johnbean393/Sidekick as patterns. Please consider open sourcing whatever you build.
- aaronbaugher 1y agoNot exactly on purpose, but I used Grok for some business planning, and now it sometimes loops that into other answers, suggesting how something else I asked about could overlap with the business project, or just asking at the end, "So, would you like to work on [other project] some more?" It's a little reminder that I wanted to get back to that project, which doesn't hurt since I tend to procrastinate or get distracted and forget about projects. I think to be a real "coach" they'd have to be able to push you proactively. You can script that kind of thing, but if you're getting that done, maybe you don't need much pushing.
- dacke 1y agoThe proactive piece would be key for me as well. My challenge right now is to find the right setup for leveraging my journals and notes and a local llm for this (don't want to share such personal data with openai, google, etc.)
- ben_w 1y agoNot for journaling or therapy, but almost every time I draft something in German, I ask a model to "rate my German" so that I keep improving. Before chat models, I did something similar with Google Translate, but GT doesn't tell me why things should be different; and worse, GT tends to auto-correct typos, leading to stuff like this: https://translate.google.com/?sl=de&tl=en&text=Schweinwerfer&op=translate https://translate.google.com/?sl=de&tl=en&text=Schweinwerfer... ("Schweinwerfer", spelled like that, would actually mean "pig thrower": Schein (shine) vs Schwein (pig))