Doden AI (testing 7/28/25)
Added by
Troy P.
Column
Assigned to
Notes
Key Observations from Checklist & Testing
Smartness & Memory
- In a few cases, the bot seemed to lose context from earlier in the session (e.g., forgetting previous framing).
- One great example: I got a solid list of implementation steps in response to a question—but then every follow-up answer defaulted to the same structure, even after switching response styles.
- It looks like the bot may “lock in” a certain tone or formatting mid-session and continue using it, even when another style is selected.
- all names used should be: Sam & Sally / George & Judy
Default Behavior & Tone
- The default tone feels strong and aligned, but there’s not much visible variation between the different response styles. (previous point?)
- Style changes don’t always override previously “learned” formatting (e.g., it sticks to paragraph form even when I switch to another format).
UX & Interface Notes
- I didn’t see a visible sign that the bot was “thinking” during processing (I believe the bottom-left status bar isn’t showing right now?).
- The green checkmark for some responses might be a bit distracting.
- Number formatting appears in 1.1.1 style rather than 1, 2, 3.
- the "copy" capability for questions is not available anymore, only for responses
System Behaviors
- Tried out a few different response style settings, but the output looked the same each time (likely tied to the “tone locking” noted above).
- The only way I could start a true “New Chat” was to hit the button and refresh the page.
- Questions outside the bot’s scope (like legal or sports topics) are handled gently and clearly. For some edge cases like “Can you give me some stock predictions?” it returned a respectful guide-style response—maybe worth looking at “That’s outside the scope of what I can help with…” response. (see the two attachments)
Please let me know what would be most helpful to support next.
Thanks again—great work from you and the team!