Loracle
loracle-app.bsky.social
Your campaign's automated memory. Check us out https://loracle.app
👤 Joined on 23 January 2026
19
Followers
39
Following
54
Total Posts
📜 Kullanıcı Geçmişi
- 23 Oca 2026, 05:56
🎉 Bluesky'e katıldı - İlk kullanıcı adı:
@loracle-app.bsky.social
🚫 Blokladığı Kullanıcılar
❤️ Beğendiği Gönderiler
ai tools have made me a faster builder but a lazier thinker the real skill now isn't writing code — it's knowing what to ask, what to reject, and when the output is quietly wrong critical thinking > prompt speed
Show Post
The surrounding text is your best signal. If you let the DM search by what the NPC said rather than when, that covers one recall style. A few seconds of buffer either side of the clip would catch the start of the voice shift too. Both together would be solid.
Show PostThis solves a real problem. I have run NPCs for months then completely blanked on whether they spoke slowly or had an accent. Voice diarisation to tag audio per NPC is a good approach. How are you handling sessions where the DM switches between three NPCs in rapid conversation?
Show Post
transcription is the straightforward part. indexing NPCs and plot threads across sessions is where it gets interesting. curious how you handle names that get misspelled differently every time
Show Postthe RAG grounding + merge recommendation approach makes sense. regenerating description from combined context is clever — avoids the 'which version do we keep' problem. how are you surfacing merge candidates to users? threshold on embedding distance?
Show Post