We’re excited to introduce Next Lab 2.35.0, bringing improvements to the guided tour experience and powerful new capabilities in the LLM chat.

🧭 Tour available from more places

The Lab tour can now be launched from several areas of the platform, including index, select, home, session pages, and drill lists, making it easier to explore the Lab from wherever you start.

▶️ Play Session guidance

The tour now includes steps explaining how to start a session and how the Play Session page works, helping users understand how to run sessions more easily.

🤖 Compare AI models side-by-side

We added a new chat mode where a single question is sent to multiple models, allowing you to compare responses from:

• OpenAI
• Claude
• Mistral

This makes it easier to evaluate how different models respond to the same prompt.

🌐 Web search capabilities

OpenAI and Claude models now support web search, enabling them to bring more up-to-date information into their responses.

🙌 Thanks

Thanks for the feedback that keeps improving the Lab

We’re excited to introduce Next Lab 2.34.0, bringing a new guided onboarding experience designed to help users get started faster and with more clarity.

🧭 Step-by-step guided tour

We’ve added a new in-app tour that walks you through the key steps to create a session and add a drill. The tour guides you step by step, making it easier to understand the workflow inside the Lab.

🎥 Video-enhanced guidance

Some steps in the tour can include short instructional videos, giving you visual guidance and helping you quickly understand what to do next.

📊 Insights through Mixpanel

All interactions with the tour are now tracked with Mixpanel, allowing us to better understand how users engage with onboarding and identify areas for improvement.

🌐 Full language support

The guided tour automatically adapts to all languages available in the Lab, ensuring a consistent onboarding experience for every user.

🙌 Thanks

Thanks for all the feedback that helped shape this feature.

We’re excited to share Next Lab 2.33.0, focused on improving IIS reliability, response quality, and overall navigation experience.

🤖 IIS Assistant stability fix

We resolved an issue where the IIS AI Assistant could return an error when multiple questions were asked at the same time.

The assistant now handles concurrent interactions more reliably.

🧠 Improved prompt quality

We enhanced the LangChain prompts used by IIS, resulting in more consistent, structured, and relevant responses.

🔍 User search pagination fix

We fixed pagination issues in user search to ensure results are displayed correctly across pages.

🙌 Thanks

Thanks for your continued feedback—keep sharing ideas and improvements.

We’re excited to introduce Next Lab 2.32.0, which includes an experimental (beta) feature to create agendas directly through the IIS chat.

🧪 Session Agenda creation via IIS chat (Beta)

You can now ask the IIS assistant to create an agenda through chat.

This feature is currently in beta, and we’re actively testing how well it fits into real workflows.

We’d love to hear from you:

Does this flow feel intuitive?

Is anything missing or unclear?

What would make agenda creation faster or more useful for you?

Your feedback will help shape how this feature evolves before a full release.

⚠️ Beta notice

Since this is an experimental feature:

You may encounter rough edges or incomplete flows.

We’ll be iterating quickly based on your feedback.

🙌 Share your feedback

Please share your thoughts and suggestions through the Frill boards—we’re excited to learn how this works for you and improve it together!

We’re excited to announce Next Lab 2.31.0, focused on improving the stability, usability, and shareability of LLM chats—along with a few behind-the-scenes enhancements.

⏳ Clearer feedback when sending messages

We added a loader when sending messages, including scenarios with multiple image uploads, so it’s clear when content is being processed.

🔧 LLM chat reliability improvements

We addressed technical debt in LLM chats to improve stability, error handling, and overall reliability across chat flows.

🔗 Shareable deep links for chats

LLM chats now support deep links, allowing you to share a conversation URL and reload the chat directly in the Lab.

📊 Better internal insights

We introduced internal tooling to help identify the most engaged users, enabling better understanding and future improvements.

🙌 Thanks

Thanks for all the feedback.

We’re excited to share Next Lab 2.30.0, focused on making LLM chats easier to use and manage.

📎 Multiple file selection in LLM chats

You can now select and attach multiple files at once when chatting with LLMs, making it faster to provide context and work with several documents in a single conversation.

🏷️ Improved chat title generation

We improved how chat titles are generated and selected, helping you better identify and organize your LLM conversations.

🙌 Thanks

Thanks for the feedback that continues to shape these improvements

Here is what’s new in this release.

Domain-based user search for admins:

Admins can now find users by email domain (e.g., all @gmail.com users) to manage groups more easily.

File attachments in LLM chats:

You can attach files to chats for quick reference within the conversation.

• Attachments with previews: attached files display a preview in chat so you know what’s there at a glance.
• Attachment icon and current file shown in chat actions: a clear attachment icon and the selected file are visible in the chat area.

Copy/paste in Home Chats:

Copy and paste content directly in Home Chats for faster sharing.

• Chat history now shows related resources at the bottom: you can see resources tied to a chat without leaving the conversation.

Expandable Create Session/Challenge Statement field:

Long statements are easier to view and edit without losing context.

What’s improved

• Readability and formatting: chat content is easier to read, with better overall readability and less visual clutter.
• UI polish: badge counts and related UI hints feel more accurate and responsive.

What’s fixed

• Better handling of problematic chats: the system now cleans up bad chats more consistently to keep conversations tidy.
• Various reliability tweaks to keep chats snappy and predictable.

Notes

This release focuses on user-facing enhancements around search, attachments, and readability. If you have any feedback or run into anything unexpected, please reach out through the feedback button — we’d love to hear how these changes feel in your workflow.

We’re excited to announce Next Lab 2.28.0. This release focuses on improving the LLM chat experience and expanding model options to make conversations clearer, more useful, and more flexible.

✨ Richer message formatting

LLM chat now supports formatted responses, making long answers, lists, and structured outputs easier to read and understand.

📋 Copy button in chat

Each AI message now includes a copy button, so you can quickly reuse outputs in documents, tools, or workflows.

🤖 More models available

You can now choose to chat with additional models:

Claude is now available in the LLM chat page.

Mistral is now available in the LLM chat page.

This gives you more flexibility to compare responses and pick the assistant that best fits your use case.

🙌 Thanks

Thanks for all the feedback that helped shape these improvements.

Next Lab Release 2.27.0

Including updates from 2.26.0

We’re excited to share Next Lab 2.27.0, a release that builds on 2.26.0 and focuses on making AI-powered workflows smoother, clearer, and more resilient. This update improves how AI Vision and Purpose agents behave, adds better feedback during long-running actions, and fixes several edge cases that could interrupt the user flow.

Below is a concise overview of what’s new and why it matters.

🚀 Better Feedback & Progress Visibility

We’ve significantly improved how users understand what’s happening while AI is working:

• Progress tracking improvements for AI agents, including forward and backward step updates
• Loaders and status messages when uploading resources or waiting for AI responses
• Clearer loader messaging for AI suggestions and form submissions

These changes reduce uncertainty and make longer AI tasks feel faster and more transparent.

🤖 Smarter AI Vision & Purpose Flows

AI Vision received a major internal upgrade to improve reliability and continuity:

• Asynchronous answer updates, ensuring smoother UI performance
• Ability to continue a flow from the last answered question
• Improved handling when navigating backward and forward through AI-generated steps
• Fixes for AI suggestions when revisiting previous answers
• Confirmation flags sent from the backend so the frontend always reflects the correct AI state

Overall, AI Vision now behaves more predictably, even when users revisit or revise earlier inputs.

🔄 Improved Navigation & UX Details

Several smaller UX improvements add up to a more polished experience:

• Forward and backward navigation in AI Vision assistance
• Translated button labels for Previous, Next, and Back to Current
• Cleaner modals by removing distracting or unclear indicators
• Markdown parsing support for constituent values
• Resource disclaimers added for clarity and trust

🛠 Stability, Fixes & Under-the-Hood Improvements

This release also includes important stability and maintenance work:

• Fail-safes to prevent errors from nil or missing budget values
• Budget field constraint adjustments in opportunity forms
• Refactored job and state-management logic for AI Vision
• Cleaner result broadcasting and async job handling
• Test coverage improvements

These changes don’t always show up visually, but they make the platform more robust and easier to extend.

✅ Included Sprint Work

This release combines completed work from Sprint 2.26.0 and 2.27.0, including:

• Resource upload loaders
• Opportunity budget bug fixes
• Automated AI purpose start logic
• Review feature improvements
• Documentation updates

What’s Next

This release lays the groundwork for even richer AI agent experiences, especially around progress visibility, async workflows, and editing AI responses without losing context.

As always, we’d love your feedback — and we’re already building on this foundation for the next iteration of Next Lab.

Release: 2.27.0Includes changes from 2.26.0

We’re excited to share the latest improvements included in Next Lab 2.26.0.

This release focuses on improving the handling of uploaded resources, streamlining agent workflows, and enhancing the experience during AI-guided steps.

📎 Better resource upload experience

When uploading a file in an Opportunity or Small Bet, the interface now displays a loader, making it clear that the upload is in progress and preventing confusion during processing.

💰 Opportunity budget flexibility

Users can now update an Opportunity even when the portfolio budget is lower, avoiding unnecessary errors when changes are unrelated to budgeting.

⚡ Faster agent setup

Clicking “Advance with AI” now automatically starts the agent session, providing a smoother and faster experience.

📄 Resource disclaimer

A new disclaimer informs users that uploaded resources are processed by AI models and used as context for suggestions and reasoning.

🧹 Cleaner agent responses

Agent outputs previously returned in HTML are now parsed for a cleaner, more readable display.

📊 Vision & Purpose progress indicator

The Vision & Purpose agent now includes a progress bar to make it clear how far along you are in the guided Q&A workflow.

🔍 AI context expanded with Small Bet resources

Resources added to Small Bets are now interpreted and included as part of the AI context, enabling more grounded suggestions and better use of agent capabilities.

🙌 Thanks

Thanks for the feedback that guided these improvements — keep sharing ideas and issues through the boards!