How We Use AI
AI is in every part of how we work, from research and strategy to the websites and assets we build. Here’s exactly how we use it.
Frontier models
Claude, ChatGPT and Gemini handle reasoning, writing and analysis, and Claude and ChatGPT Work take on long, multi-step projects from start to finish, including full website builds.
Open models
We use Ollama to deploy open-weight models like Qwen, Gemma, gpt-oss and DeepSeek on local hardware.
Model access and inference
OpenRouter gives us one API to hundreds of frontier and open models. Hugging Face, Together AI, Fireworks and Modal host open-weight models and handle inference for models too large for local hardware.
Build
Claude Code, Cursor and Devin write, test and ship code across full applications, integrations and data workflows.
Research and market intelligence
Exa powers market and competitive research with web search built for AI. Clay builds company datasets for ICP and TAM analysis with AI research agents, and Databar puts 160+ data providers, AI web scraping and waterfall enrichment behind one API, SDK and MCP server.
Infrastructure
Supabase manages data and authentication, Pinecone stores vector embeddings for retrieval, Vercel deploys the front end, and Cloudflare provides security and delivery at the edge.
Every page of this site was built with AI.
That includes the design and code, plus the files and markup search engines and AI assistants read.
- Page design and front-end code
- Interactive features
- Mobile optimization
- Cross-device visual testing
- Image optimization
- Accessibility markup
- Performance testing
- Favicon and share images
- Title tags and meta descriptions
- Open Graph link previews
- Schema.org structured data
- llms.txt for AI assistants
- robots.txt and XML sitemap
- Security headers