This is essentially a thin wrapper around an LLM. Hardcoded prompt templates. The value isnt in the template itself, but more-so the curation of the template. So curate better.
Instead of hardcoding the prompt template, allow people to create/share/vote on arbitrary templates. A prompt library of sorts.
Open-ended prompting seems to go against your prescription of more curation. Also, I don't think most people want to think about LLM prompting. The tool should be prompting for me. I think they are going towards the right direction of how to use LLMs, not away from it.
A "break down task into steps" would be a great feature for every todolist app.
>Open-ended prompting seems to go against your prescription of more curation
Not at all. There is the selection mechanism and the universe of templates. Both can be greater in a system where there is user creation and voting. Otherwise we're saying Goblin.tools has searched and found the best templates possible. Seems unlikely. And even if true now you dont need Goblin.tools anymore because you have just recreate the static templates.
A problem is that the proposed audience are individuals who struggle with options. Giving them more options doesn't help them. Enabling a power user mode with user creation, voting, etc. would be valuable for others. I suspect template sharing in some situations is problematic because getting great results for an audience can have text which some will find objectionable.
This is essentially a thin wrapper around an LLM. Hardcoded prompt templates. The value isnt in the template itself, but more-so the curation of the template. So curate better.
Instead of hardcoding the prompt template, allow people to create/share/vote on arbitrary templates. A prompt library of sorts.