Lutra AI

Introducing Lutra, an AI Agent that generates and runs workflows for you. Lutra first converses with you to understand your goals, and then generates and runs code that integrates with the apps you use. Lutra has crafted an environment for AI generated code to run in a production-ready setting. By taking a code-first approach, we leverage the best practices from software engineering for: 1. Security 2. Maintenance 3. Governance We're excited to announce $3.8M in funding from Coatue, Hustle Fund, Maven and WVV Capital, together with Angel Investors that include Andrej Karpathy, Jeff Dean, Scott Belsky, Quoc V. Le, Tom Willerer, Brian Shultz, Vijay Karunamurthy, John Ciancutti, Irwan Bello, Keith Peiris, Kah Seng Tay, & more! Sign up today at https://lutra.ai Follow us for on X/Twitter at https://lnkd.in/gBx4Tpzn for updates Join our discord at https://lnkd.in/gZZfJqSy

We're excited to share a new suite of business research playbooks! 📗 One of the most popular uses for Lutra is to perform web research on a large database or spreadsheet of accounts, leads, or potential hires. Having deep insights about them enables you to better qualify, prioritize, and even customize your approach. 👍 We are sharing a suite of data-driven playbooks that enables you to quickly understand what makes a business thrive - its team growth, tech stack, or social media engagements - just by plugging in a company name. These playbooks are designed to be easy to use: they come pre-built and you only need to provide a company name, or a spreadsheet of companies. They can surface the insights you need to guide strategy and win. 🚀 Please like and find the link to the Playbooks in the comments!

Here's my take on explaining AI Agents in a user-centric way ... Traditional Software: We operate it directly. Copilots: AI guides us in operating software. AI agents: AI operates the software under our direction and supervision. For traditional software, we point-and-click to get things done, then, there’s copilots which help us understand where to point/click or do a few of those for us. Coming soon, there’s AI agents that manage the entire use of software/computer, and we just tell it what we want to get done and it will figure out how to do so. I've found a lot of explanations of AI Agents be too technical or about how it works, rather than an intuitive explanation of how users will use it -- does this user-centric version resound with you?

Lutra auto-corrects itself to automatically make progress! Excited to see this in action :)

We made an AI for lead research and outreach strategies. It helps you do deep research on a lead, how your product fits in, and makes the connection on how your product can help them. Lutra automatically orchestrates all the tools needed to accomplish this and you only need to provide the lead's name & company. Try it out - really interesting to see what it finds and pulls up. Link in comments.

Getting accurate, detailed data about leads, businesses, anything on the web is easier than before! This video demonstrates research on coffee beans / equipment coffee shops use -- data that's usually hard to find, but now you can just ask the AI to get it and put it into a spreadsheet. Use this to ask any question across a large list of contacts or companies - takes less than 5 mins to get it done. Behind the scenes, we have improved our web answers backend -- it now returns citations, reads webpages deeply, and you can combine it with other extraction tools to achieve a task.

If you're working with local businesses and need to plan your day around which ones to visit and when -- it can be a pain to figure that out. Here's how you can use AI to convert an Excel sheet of locations into a map with pins -- and even better, get it to figure out an optimal route to take for the day. Super cool to create interactive artifacts from your data!

Markdown output from models often end up in my docs with all the ** and formatting wrong -- so we fixed that with how Lutra interacts with Google docs. When you ask it to add content to a Google doc now, it will translate all the markdown syntax into the proper format for Google doc! Feels good when we get the details right :)

The key insight to getting computer use models to work was teaching the machine to count pixels - if it can identify exactly where an interaction needs to be, you can now get it to click, type, drag-n-drop, scroll with the computer. I think this will have broader implications for vision itself, if the model understand pixel coordinates well now, we can ask it to draw bounding boxes (or even polygons, or pixel level segmentation) on the images. Multimodal AI is getting super exciting. Kudos to Anthropic for the new models! p/s: We're starting to develop with computer use models - if you have a use case you'd like to automate, but had always issues with the lack of APIs, please reach out!

I'm still amazed by what AI can do with data - everyone should be using it as their junior business analyst. We made new abilities for Lutra to work well with Google sheets, Excel - pull data from spreadsheets, merge them, do data analysis, make charts, and make updates back into the sheet. Huge opportunity to give everyone the ability to get insights from their data.