Nebuly

User analytics and insights from LLM conversations.

[Nebuly @nebuly_ai]

Keyword search matches words. Conversational AI understands intent.

Better prompts = better answers.

But you can’t improve prompts if you don’t know how people actually ask.

Our latest blog explains this shift. Read at https://t.co/yywwHLmM9c

[Nebuly @nebuly_ai]

One global company rolled out a GenAI copilot across its dealer network.

5 minutes with Nebuly showed:
• One group had high adoption
• Others barely used it
• Errors were linked to language

Read here: https://t.co/16oO4D8Gy4

[Nebuly @nebuly_ai]

You ask “What’s going on in Georgia?”

You mean the US election. The assistant gives updates on protests in Tbilisi.

Not wrong. Just the wrong context.

LLMs don’t know your intent unless you track the signals.

Read more: https://t.co/LpqFmJAbPG

#GenAIAnalytics

[Nebuly @nebuly_ai]

“Usage” isn’t ROI.

Token counts and login events don’t show if AI tools are useful, trusted, or effective.

The real signals come from user behavior.
Retention. Drop-off. Satisfaction. Completion rates.

Read more (5-min read):
https://t.co/nN4lhfkvZO

[Nebuly @nebuly_ai]

Introducing a new, user-centric framework for evaluating and developing LLM-based products.

https://t.co/xTkl7jCxbV

[Nebuly @nebuly_ai]

LLMs are changing Conversational Analytics from both ends:

1. We are having conversations with LLMs. 💬
2. Analyzing conversations is best done by LLMs specifically built for this task. 🤖

More on our latest post:
https://t.co/WiKmU6M4AF

[Nebuly @nebuly_ai]

📈 Trends for LLM User Intentions and User Feedback.

You can now see if a particular type of user feedback or user intent is trending up or down.

Quickly understand changes in user experience over a period of your choice.

[Nebuly @nebuly_ai]

🆕 Blog Post: LLM Evaluation Methods: 🤠 Human-as-a-Judge vs. 🤖 LLM-as-a-Judge! Dive into these two perspectives in LLM evaluation. Learn about user feedback, implicit evaluations, and automated assessments. Read more: https://t.co/G3LT6WS3va

[Nebuly @nebuly_ai]

🚀 Our latest product update: Easily track custom User Intent and User Warnings in your LLM conversations! Stay on top of what matters most to you and your customers.

https://t.co/12GdL0bhKf

[Nebuly @nebuly_ai]

Launch early & collect feedback. 💯

[Mike Knoop @mikeknoop]

You cannot reliably say an AI product with 100% accuracy over 10 users will be even 50% accurate over 1000. This is why it’s critical to launch AI products early. To start tuning and improving based on real input diversity.

[Nebuly @nebuly_ai]

👉 Phi-3 is already available on @Azure and @huggingface where the integration with Nebuly is simple https://t.co/yaod5xhQWn

We are excited to see what will be built!

[Nebuly @nebuly_ai]

Another week, another AI model! 🔥 Microsoft unveils Phi-3-mini, a compact language model that runs on smartphones. 📱

[Nebuly @nebuly_ai]

📱 Going mobile: Phi-3-mini will empower developers to create smarter and more effective AI-powered experiences especially within mobile apps.

[Nebuly @nebuly_ai]

3. It's multilingual. All use cases where an LLM is used to process data from different languages will greatly benefit from the LLaMA update.

More here https://t.co/YGcusBzAVi

[Nebuly @nebuly_ai]

@AIatMeta launched their latest LLM last week, called LLAMA-3. @nebuly_ai CTO, Diego, explains what makes it interesting and how it stands out from other models.

[Nebuly @nebuly_ai]

2. 7B-sized models are useful, can be fine-tuned on a single A100 GPU for downstream applications.