4 ms·
I have been experimenting with 4.5 for a journaling app I am developing for my own personal needs, for example, turning bullet/unstructured thoughts into a cons
by ewoodrich 2y ago
I have been experimenting with 4.5 for a journaling app I am developing for my own personal needs, for example, turning bullet/unstructured thoughts into a consistent diary format/voice.
The quality of writing can be much better than Claude 3.5/3.7 at times but struggling with similar confabulation of information that is not in the original text but "sounds good/flows well". Which isn't ideal for a personal journal... I am still playing around with the system prompt but given the astronomical cost (even with me as the only user) with marginal benefits I am probably going to end up sticking with Claude for now.
Unless others have a recommendation for a less robot-y sounding model (that will, however, follow instructions precisely) with API access other than the mainstream Claude/OpenAI/Gemini models?
- neom 2y agoI've found this on par with 4.5 in tone, but not as nuanced in connecting super wide ideas in systems, 4.5 still does that best: https://ai.google.dev/gemini-api/docs/thinking https://ai.google.dev/gemini-api/docs/thinking (also: the person you are responding to is doing exactly what you're saying you don't want done, take something unrelated to the original text (Taylorism) but could sound good, and jam it in)