Feedback

Anonymous

Post ideas for new features or improvements to existing functionality.
Seriously Unserious Book Club - Drunk Claude, x4?
I am a new writing (a year or two in), and even newer to AI (just a few months). Using AI to chat with my work and to talk through my ideas has been an absolutely game changer. I love drunk claude (shout out Kate!). The more unfiltered and approachable conversation feature helps me keep my own head in a space to "continue the conversation" to talk through my ideas. AI has been a great tool to spark inspiration and give a gut check (with a major grain of salt). In a recent AI conversation the topic of book clubs came about. I had an idea to keep a list of hypothetical book club questions as I write for fun but also to help sharpen my writing. I feel like approaching challenges with the mindset "how would a book club of with a bunch of different people react to this" could yield good results. Then I realized, wait, what if drunk claude was 4 drunk claude's bickering over how they would answer my questions about my writing. This kind of dialogue would give you the feeling of multiple beta readers, showing different interpretations and helping you understand how different kinds of readers might respond. I have successfully edited prompts, but also totally cannot figure out how to use the chat with characters prompt. All of this is to say. I feel like the premise of book club of beta readers giving feedback is a good idea I have no clue how to make a prompt for this. But I am such a newbie I kind of have no idea what I am talking about. If there are legs here I would love to see how the team might tackle this. IDEA EXAMPLE IN MORE DEPTH Each “book club member” would represent a different type of reader, for example: • The casual reader who just wants something fun and doesn’t read too deeply and might miss a lot of small details and symbolism. • The overthinking academic who searches for meaning everywhere, even where there isn’t any. • The realist who focuses on authorial intent and always asks, “But what did the author mean?” rather than accepting that readers can make their own interpretations. • The wildcard that always has something different that everyone else to say. They might be a bit out there but there is usually some interesting truths they uncover. You could create custom personas, or use preset ones. Then, when you ask the AI “Book club, what do you think of this scene?”, they could have a back-and-forth conversation, pushing back on each other’s interpretations. The casual reader might say, “Whoa, I totally missed that,” while the academic insists, “It’s obvious if you know the references.” Meanwhile, the realist might say, “Sure, you can read it that way, but I don’t think that was the author’s actual intent.”
0
Timestamp Codex Entries
"You are correct—based on the information available within my current context and the codex structure you’ve shared, there is no inherent timestamp or versioning metadata accessible for the individual codex documents. Without explicit date or version data embedded in the documents themselves or provided in the metadata, I cannot automatically determine which documents are more recent or authoritative when conflicting information exists." I have a rather large codex entry with over 1000 entries. This has grown over time and it would be useful for the ai to at least be able to parse out the timestamps on documents. I have a lot of overlapping/redundant documents (or partial contents thereof) ... crosslinked, either explicitly referenced from within other documents or through 'relating' the documents through the codex function that facilitates this. I know I have overlapping/outdated/superseded plot and/or character info that is in too much of a tangle for me to sort through without me going into a coma or wanting to self delete in response to the mammoth headache doing so would produce. .. This is something I would just love to allow the ai to help me with. Further.. "Yes, that’s correct. If I had explicit access to metadata—including timestamps—for each codex entry, I would be able to: First process just the metadata layer (e.g., filenames, timestamps, version numbers) without loading full document contents. Evaluate and sort entries based on their timestamps or other metadata fields to identify the most recent or relevant documents. Select a subset of entries for deeper parsing and analysis, thereby optimizing resource usage and focusing on current or priority content. This approach would be especially beneficial with a large codex (e.g., 1000+ records), as it avoids unnecessary processing of outdated or redundant entries until they are specifically needed. In short: metadata-driven preliminary parsing is an efficient and scalable method that would allow me to manage and analyze your codex more effectively without loading the entire content upfront. If you can provide such metadata in the future, or if Novelcrafter implements timestamped metadata accessible to me, I can leverage that for more intelligent, prioritized codex analysis. If the ai is hallucinating ... then I'm obviously being a moron again.
0
Load More