Chat now supports file uploads as context, including .txt, .csv, .tsv, .xlsx, .md, .json, and image formats. The context window has been expanded to 1 million tokens, enabling longer reasoning and deeper analysis.
You can now control what information each column agent references. Configure context scope to limit references to selected columns and restrict web lookups to specific URLs.
Enterprise workspaces can now access live financial data through Polygon.io. Stream real-time prices, fundamentals, and market metrics directly into your sheets.
Email columns now use Findymail for improved accuracy and match rates. This integration increases deliverability confidence and ensures cleaner contact data.
Sheets and chat threads now automatically generate titles. Each new object is named using AI to preserve clarity and context as you work.
Chat has been reimagined from the ground up. The new interface enables richer search, organization, and control across all conversations.
Paradigm Chat now runs on Claude 4.5, offering faster, more reliable performance. Markdown and file support in chat allow sharing of outputs in clean, readable formats.
Our new people search tool allows you to find and enrich leads directly from chat. Live web search is also available, enabling you to always fetch the latest results.
You can now define which columns chat and enrichment agents can access. Restrict data sources to improve precision, starting with column-level controls and expanding soon to URL-level configuration.
Cell enrichment has been upgraded to reason beyond raw data. Email lookups now include verification, and LinkedIn enrichment accuracy has been improved for both profiles and companies.
Chat history now persists across sessions. Sheets and threads are automatically named for easier navigation. You can stop active chat sessions at any time to maintain focus and control.
Paradigm 1.0 marks our transition from private access to full public release. The platform is now live for all users, offering a one-month free trial before moving to standard plans starting at $20 per month.
Workflows replace traditional webhook configurations with a structured automation system. Input and Output Workflows enable seamless data movement between Paradigm and external tools.
Team and private workspaces are now fully supported. Sheets created in team spaces are automatically visible to all members, simplifying collaboration.
A new in-product onboarding experience introduces users to enrichment, chat, and sheet creation through a guided interactive tour.
Paradigm Email now supports Outlook alongside Gmail. You can connect and switch between multiple OAuth profiles directly from the interface.
The second generation of Chat introduces global context awareness. Use Chat to configure sheets, generate starter data, and add contextual insights throughout your workflow.
Paradigm 1.0 debuts a new design system and logo that reflect our core principles of clarity, precision, and forward-thinking engineering.
Paradigm now supports the GPT-5 model series. On Business plans, you can select models at the sheet or column level to optimize for accuracy, speed, or cost.
Plans now begin at $20 per month. Generations are faster, higher quality, and more cost-efficient. Enterprise plans include advanced features and expanded data limits.
Agent reasoning has been refined through improved scraping pipelines and context window management. Each enrichment now provides detailed reasoning traces for greater transparency.
Multiple drafts can now be composed within a bottom panel, allowing you to view sheet context while writing. Connected inboxes can be switched dynamically through OAuth.
A new review page displays personalized previews for each recipient before sending, improving quality control for outreach campaigns.
Admins can now access a high-level overview of platform usage, including daily active users, cells generated, and sheet edits. View metrics by day, week, or month, and filter by individual for more granular insights.
We have improved agent data quality across the platform by training and releasing support for the latest model families. On email data specifically, Paradigm agents achieved an 11.4% improvement in evaluation accuracy over the previous month.
Our internal Paradigm Agentic Research Evaluation benchmark shows consistent performance gains across top models, with leading results from Claude 3.5 Sonnet, Gemini 1.5, and GPT-4o. These refinements enhance both enrichment reliability and reasoning precision across all workflows.
You can now filter any column type globally. This allows you to create views with different criteria depending on what you are working on. Additionally, you can add global sorts for any number of relevant columns right from the sheet action menu.
We are gradually rolling out a new form of chat that has increased contextual awareness, increased reasoning ability, and the capability to modify all parts of your sheet. While datasets mode can get you started with initial data, agent mode is your assistant throughout the entire process of finding, modifying, and taking action on your data.
We released product documentation, including information on sheet setup and making the most of your column configurations and prompts. Check out the docs for more advanced features like webhooks or template creation.
We updated our enrichment queue to retry any cells that are lagging. We also refined our job cancellation process to work faster and more consistently. Now, whenever you start a job, a cancel button will appear to the side of the enrichment button.
You can now pull live data into your sheets from any webhook-compatible source. For example, internally, we have set this up to automatically enrich our incoming waitlist from our database and our recruiting candidates from our ATS system. This extracts incoming data to match the columns you set up and enriches any missing information based on the extracted data.
In addition to our built-in waterfall analysis from email data providers and our custom web scraping agents, we added a verification layer for email deliverability. We now only surface emails if their deliverability is verified.
Each column is now automatically prompted based on just the column name you input. If you want to provide more specific details, you can add them in the section below the column name. We generally only recommend this if the column name is not self-explanatory. Click the magic wand to autogenerate prompt details.
We added keyboard shortcuts for enriching data, toggling the sidebar and chat visibility, and switching between the email composer and other views. We have also added a menu in the corner that displays all the keyboard shortcuts.