Airtable

Airtable was founded 13 years ago with a mission to democratize software creation—to empower the millions of people who use applications but can't build them themselves. Howie Liu started as an IC, deeply involved in both backend (real-time data architecture) and frontend UX design. He recognized that for a pure software product where the technology IS the product, the CEO's intimate involvement in design decisions was essential to finding product-market fit.

Howie and his founding team spent 2.5 years building Airtable before launch, following a pattern similar to Figma's trajectory. The product was meticulously crafted as a no-code platform with powerful Lego-like primitives: collaborative CRUD interfaces, multiple view types, a layout engine, and automations. This deliberate, hands-on approach during the founding phase proved critical because every architectural and UX decision directly shaped the product's value proposition.

Airtable became one of the great product-led growth success stories of its era. By removing friction and allowing users to immediately experience the platform's capabilities, the company grew viral adoption without relying heavily on paid acquisition. This PLG motion scaled users rapidly—the company expanded from early adopters to becoming a dominant no-code platform used by SMBs and enterprises alike.

For years, Airtable thrived with a maturing approach: strong teams organized by feature areas, successful sales motion up-market, and steady product development. However, Howie recently confronted an existential question: If you founded Airtable today with the same mission in a fully AI-native world, how would you execute differently? This inquiry led to a major reorg. The company split engineering into two groups: the "fast thinking" team (officially AI Platform) shipping major new AI capabilities nearly weekly, and the "slow thinking" team handling complex infrastructure like HyperDB (supporting multi-hundred-million-record datasets). Howie significantly reduced his one-on-one meetings to focus on hands-on work with AI, spending hundreds of dollars daily in inference costs experimenting with LLM capabilities. He personally became likely the highest inference-cost user of Airtable globally. The company pivoted to make their AI agent Omni the default interaction mode rather than a sidebar feature, fundamentally reconceiving how users would build apps through agentic conversation rather than traditional GUI manipulation. This new AI-centric approach allowed Airtable to apply its existing no-code primitives as a "domain-specific language" that AI agents could manipulate more reliably than generating code from scratch.

Howie has returned to being an IC-CEO, writing code and building features himself. He treats the CEO role with a barbell approach: either urgent, timely meetings around specific topics and insights, or deeper relationship-building over longer stretches (monthly lunches/walks). He spends hours daily playing with competitors (Cursor, Windsurf), testing new AI products (Claude, GPT-4, Runway, Sesame), and running personal "weekend projects" that force genuine product usage beyond surface-level testing. His conviction is that to taste the soup, you must help cook it—especially in an era where every new model release opens novel form factors and UX patterns. Airtable now positions itself as the AI-native alternative to pure vibe-coding platforms: combining the reliability and explainability of structured no-code components with the power of agentic AI generation, specifically targeted at business app creation rather than viral consumer tools.