Direct Quotes mode: Does not actually pull direct quotes. At one point, the system would pull direct quotes from something Daryle said, whether it was in an email or Learning Time transcript. This mode needs to do that - a lot of people will use this mode to see exactly what Daryle says - this will help build trust in the system.
Screenshot 2025-12-01 at 7.51.09 PM.png
Download Default Mode should be renamed "Standard Mode" - this should feel the most like ChatGPT and Gemini Why are all the answers being given have headlines? It's NOT conversational like before. It needs to feel conversational in "Instant" like before.
Instant - Looks like gpt 4.1 is the fastest + no headers (ill experiment again with gpt 5 mini / nano) - Direct quotes pulls directly: I integrated a function that pulls verbatim text based on revelance. Then we have an LLM add on top of what's said. Perhaps we can improve the format of direct quotes mode
Justin Sheehan
Chief Journey Officer
Troy
I thought Instant mode was running GPT? It seems like its running Claude
I recall creating a prompt that prevents headers in the instant mode, but we're now using the latest models. I'll optimize instant so it truly feels instant
Justin Sheehan
Chief Journey Officer
Sounds good! Troy
Troy Pastoral
AI Whisperer
Edited
Need to use gpt 4.1 for instant (sub 8 seconds, no header). gpt 5 has too many latency issues! GPT 5.1 thinking is working ok though
Direct Quotes mode: Does not actually pull direct quotes. At one point, the system would pull direct quotes from something Daryle said, whether it was in an email or Learning Time transcript. This mode needs to do that - a lot of people will use this mode to see exactly what Daryle says - this will help build trust in the system.
Default Mode should be renamed "Standard Mode" - this should feel the most like ChatGPT and Gemini
Why are all the answers being given have headlines? It's NOT conversational like before. It needs to feel conversational in "Instant" like before.
Instant
- Looks like gpt 4.1 is the fastest + no headers (ill experiment again with gpt 5 mini / nano)
- Direct quotes pulls directly: I integrated a function that pulls verbatim text based on revelance. Then we have an LLM add on top of what's said. Perhaps we can improve the format of direct quotes mode
I recall creating a prompt that prevents headers in the instant mode, but we're now using the latest models. I'll optimize instant so it truly feels instant