How We're Building the Future of the Startup Economy with AI

Walk into our office on any given day, and you'll see engineers pair-programming with AI. Product managers are iterating on specs in real-time with ChatGPT. Our developers are delegating entire tasks to AI agents while they focus on the big picture.

We integrated our systems with best in class AI because we knew it would super charge our next chapter.

Our AI Stack: Tools That Move the Needle

We're not married to any single AI provider. Instead, we use the best tool for each job.

The Workhorse → OpenAI

• Powers most of our internal tooling
• The reliable choice for retrieval and general tasks

The Specialist → Anthropic (Claude)

• Excels at complex reasoning and natural language
• When we need nuanced understanding, Claude delivers

The Eyes → Gemini

• Handles document parsing and visual tasks
• Perfect balance of cost and accuracy
• Particularly strong with tables and complex layouts

Developer Tools

Crowd Favorite → Cursor

Our most popular AI Integrated Development Environment. It's like having a senior developer sitting next to you, except they never get tired and they've read every Stack Overflow answer.

Task Delegator → Devin

Working with Devin helps get things done faster than ever. Our developers literally delegate work to Devin through Slack:

• "Hey Devin, update the copy on the landing page"
• "Devin, refactor this component to use hooks"
• "Can you update the README with the new API endpoints?"

Devin maintains living wikis for our repos and handles the smaller tasks so our engineers can focus on architecture and innovation.

The Secret Sauce: AI Rules Files

Here's something most companies miss: AI tools are only as good as the context you give them. We maintain .cursorrules and .windsurfrules in every repo.

Think of these as constitutional documents for AI—they understand our coding standards, architectural decisions, and the "why" behind our choices.

Example snippet from our rules:

• // We prefer functional components over class components
• // Always use TypeScript for new files
• // Follow our established patterns for API calls

The result? AI suggestions that actually fit our codebase instead of generic Stack Overflow answers.

AI Across the Organization (beyond code)

Quality Control That Actually Works

With Langsmith, every LLM call gets traced. No more "why did the AI do that?" We can see exactly what happened, debug prompts, and improve our workflows.

Patronus is our hallucination detector that catches AI mistakes before they become problems. It's like having a fact-checker that never sleeps.

Meeting Intelligence

Granola ensures that we never lose a great idea because of poor note taking. Granola is our always-on scribe that captures meeting notes, makes them searchable, and lets us query them later. "What did we decide about the API redesign three weeks ago?" Granola knows the answer every time.

Company Culture & AI

#ai Slack Channel

This isn't just another Slack channel—it's our innovation lab. Team members share discoveries, debate new tools, and collectively figure out what's worth our time. This channel open to the entire organization—you’ll even see our Head of People in the chat testing out new tools to optimize employee onboarding, payroll, and more.

Weekly AI All Hands

Every week, we get together to share what's working, what's not working, and what's on the horizon. It's part show-and-tell, part strategy session, part "did you see this crazy new tool?" A recent example that resulted in significantly better outputs for document parsing, which feeds into our automated financial workflows:

• Document parsing w/ Gemini: An engineer surfaced a blog post about using Gemini Flash for document parsing during an AI All Hands meeting. Following the call, he began actively using and testing Gemini Flash for document parsing. One month later, he released a proof of concept showing that Gemini 2.0 Flash performed better than OpenAI models for table parsing, noting it was "lower latency and significantly cheaper". Two weeks later, we switched our production document parsing system from GPT-4o to Gemini Flash, reducing cost, latency and hallucination rates. All from a blog conversation during a team all hands!

AI in Practice

A typical developer's day:

• Morning: Cursor helps debug a tricky React component
• Midday: Devin handles a batch of copy updates while the team works on architecture
• Afternoon: ChatGPT helps iterate on a new feature spec
• Evening: Granola captures the day's decisions for tomorrow's standup

The result: Our engineers spend more time solving interesting problems and less time on repetitive tasks.

The Bottom Line

We're not using AI to replace human creativity, we're using AI to amplify it. Every tool in our stack serves one purpose: helping our team move faster without sacrificing quality to build tools that get more capital into funds, which drives more capital into startups.

The future isn't about humans versus AI. It's about humans with AI building things that neither could create alone.

We're always experimenting with new tools and sharing what we learn. The AI revolution isn't happening to us—we're actively building it.

Ready to see what AI-powered innovation for private markets looks like? Join our team and help us build the future of startup funding.