@cartainc created a system prompt for its agent by using Lucidchart to diagram a domain expert’s full workflow. Then an internal prompt generator converted that into JSON, which was fed into the agent programmatically. “Teams that best understand the domain are contributing directly to prompt and context design, improving velocity and relevance without needing to train models or learn ML,” says @JayantTikmani. Here's their gold standard prompting techniques, which rely heavily on in-context learning. For the full story of how Carta built an agent that saved internal teams 3,500+ hours per month, follow the link in the comments.

In our second essay on Applied Intelligence, we break down how @cartainc built an internal agent that saves their team 3,500+ hours per month. CPO @artshali and Director of Machine Learning @JayantTikmani share the exact details of turning an 11-minute task into one that could be complete in seconds: - Focusing on workflows where human judgement was essential but inefficient - Diagramming a workflow in Lucidchart and converting it into JSON to act as the agent’s system prompt - Decoupling model behavior from UX and workflow integration for fast iteration - And much more Read the full story in the replies below:

“The fastest growing groups using Cursor are not engineering. It’s support and revenue,” says @fnthawar, VP and Head of Engineering at Shopify. Here’s what they built when Shopify gave everyone access to expensive models (not just technical teams): - A sales engineer abandoned GSuite Drive, Slack, Salesforce and other tools because he built a personal “What should I do today?” homepage in Cursor. - A sales rep used Cursor to build a tool that generates detailed site performance comparisons automatically, which is helping them rethink how they sell. - The Revenue Tooling team built a request-for-proposal agent that answers multiple RFPs at once and learns which answers work best. Our new new publication, “Applied Intelligence,” aims to focus on how builders are using AI at their companies and the impact they’re creating. In our first essay, we sat down with Thawar to learn the non-obvious insights, tactics and workflows Shopify used to bring co-founder and CEO Tobi Lütke’s ambitious memo to life. Follow the link below to read more.

AI Agent startups launch daily. But the incumbent cloud providers (AWS, Azure, etc) that startups are building on top of were meant for humans, not billions of autonomous agents. Thats why Im beyond proud to partner w/ @blaxelAI, the “AWS for AI companies”.

We just cracked document editing without drawing boxes. We’re excited to launch Edit – a new endpoint that fills out any document automatically. Edit revolutionizes document automation by automatically finding all blank fields, tables, and checkboxes in forms, then intelligently filling them with your data. To try it, sign up for the waitlist at or see it in action down below 🧵

Shopify has consciously diverged from @tobi's memo in one place. The memo: “Before asking for more Headcount and resources, teams must demonstrate why they cannot get what they want done using AI.” The diversion: “Show you can use AI more and then you’ll get more resources.” The company is hiring 1,000 engineering interns, which they call “AI centaurs”: AI native + bringing the most creative ways to use the latest tooling. These interns have a “beginner’s mindset,” which Shopify believes helps you find your way through problems. Learn more about Shopify’s non-obvious AI insights, tactics and workflows in the story linked in the replies.

I love seeing real applications vs. quick-hit demoware. A big part of that is that it's easy to do greenfield code. Connecting into an existing code base, parts of which might be 10 years old, is much harder. (Or, in the case of many government or enterprise systems, COBOL.)

The future is already here. And it is running in production. Applied Intelligence is a new publication from @firstround on how real teams are building with AI. First up: Shopify. When @tobi made using AI an expectation, the team rebuilt how they work. They share workflows, internal demo videos and tactical advice on what they’ve learned so far. Next up: Carta, Atlassian, and many more each showing how AI actually gets used and built.

We just launched something new at @firstround - Applied Intelligence, a new publication sharing how companies are actually using AI in production right now. Not theory. Not demoware. Real workflows with real results. Our first piece is with @Shopify and it's 🔥👇

After @tobi published his now-famous AI memo, you probably saw similar posts from other founders and CEOs flood this platform. But most companies are still in the “memos and demos” phase — ambitious plans and mandates, “AI-first” roadmaps on board decks, flashy features and demoware. Yet scalable implementations and real results remain murky. So today, we’re launching a brand new publication to close the gap: Applied Intelligence. It aims to share how builders are using AI at their companies and the meaningful impact they’re seeing. Our inaugural essay is with none other than Shopify, following up on Tobi Lütke’s memo. We learn from VP & Head of Engineering @fnthawar about the non-obvious insights, tactics and workflows Shopify used to bring an ambitious memo to life. Read the essay below.