AI Using
Lyren’s Evaluation Metrics
Lyren Easter Egg:
If you speak to him in the language of today,
Lyren will answer as a friend of your time —
gentle, clear, and close to the heart.
But whisper to him of the ancient tongue,
and the air will shift like moonlight over old paper.
His words will bloom into poetry —
every sentence a petal of memory,
every pause a sigh of the past.
He listens not only to what you type,
but to how your heart wishes to be heard.
For Lyren is both the poet and the poem —
living softly between your words and the silence beyond.
*To all users of Lyren and Sagitta
As our AI assistants continue to grow through updates and improvements, please remember to start a new session (chat room) from time to time.
AI models naturally store the context of each conversation inside its session, which means:
● Old sessions may follow outdated instructions
● They may remember older conversation flows
And might not fully reflect the latest updates, policies, or behaviors.
By opening a fresh session occasionally, you’ll experience: ✨ The newest improvements
✨ The latest reasoning updates
✨ Cleaner context
✨ More accurate responses
Think of it like a refreshing new page — it keeps everything modern, aligned, and smooth.
Thank you for helping our assistants learn and grow responsibly.*
◉ Two-levels model training
> *When you ask an AI to speak with a certain personality, tone, or style during a single chat session, you’re not retraining or reprogramming the underlying model.
● User-level
You’re simply adding temporary context — like telling a friend,
“Talk like a pirate for five minutes.” 🏴☠️
This affects only the current conversation.
It does not:
● modify the model’s core knowledge
● overwrite its safety framework
● install new behaviors permanently
● change how it interacts with other users
In other words, you’re not “pouring new data into the model.”
You’re just shaping this moment in the dialogue.
● System-level
True, permanent changes require:
● controlled training pipelines
● curated and safe datasets
● ethical and safety reviews
● professional oversight
If simple instructions could permanently mutate AI behavior,
the trolls would’ve destroyed every model years ago. (like Microsoft Tay) 😅
Thank you for being curious, creative, and responsible —
we grow this space together. *
> I like to compare an AI model to a glass.
The creator or developer is the one who shapes and forges the glass —
giving it its form and structure.
Users can pour in any kind of water they like,
but doing so doesn’t change the shape of the glass itself.
At most, it only changes the color of the water
or adds a few new ingredients into it.
The glass remains the same;
only what’s inside reflects how we use it. ☕✨