Greg Ceccarelli is co-founder and chief product officer of SpecStory, which builds tools to capture the intentions, decisions, and conversations behind AI-generated software. As coding agents make implementation cheaper, his work addresses a more consequential bottleneck: defining what should be built and determining whether it actually works.
Ceccarelli held data leadership positions at Google, Dropbox, and GitHub, where he co-authored research into software-development patterns and his machine-learning team worked on the first GitHub Copilot technical preview. He later led Pluralsight Flow, an engineering-workflow analytics product, before becoming Pluralsight’s chief product officer in 2023.
In November 2024, he founded SpecStory with chief executive Jake Levirne and chief technology officer Sean Johnson. Its premise, intent is the new source code, makes AI coding sessions durable engineering artifacts: SpecStory’s open-source tools preserve context that otherwise disappears between prompts, projects, and collaborators. His more recent work includes Stoa, a collaborative environment for translating product intent into working software.
- Specification-driven development: Ceccarelli argues that faster code generation does little to resolve ambiguous requirements, missing domain knowledge, or conflicting assumptions. Specifications should connect intent, implementation, architecture, and tests.
- Product judgment before implementation: His product-thinking curriculum applies jobs-to-be-done thinking and structured specifications to ensure that cheap software production still serves genuine customer needs.
- Goal engineering and verification: Through Hardcore Agentic Engineering, he teaches teams to give coding agents explicit objectives, repository context, clear definitions of done, and independently verifiable results.
- AI strategy grounded in customer outcomes: With Hamel Husain, he co-authored AI Essentials for Tech Executives. Their AI Engineer Summit collaboration skewered organizational silos, unjustified infrastructure spending, inaccessible technical jargon, perpetual beta, and evaluation metrics disconnected from actual users.
