Morgante Pell

Morgante Pell is an engineer at Cognition and founder of Grit, the software-maintenance company that created GritQL, an open-source language for transforming code. He builds tools that help experienced engineers upgrade complex, established software without manually coordinating changes across sprawling organizations.

At Google Cloud, Pell worked on developer tooling, infrastructure automation, cloud APIs, and Terraform-related projects. Customers typically needed help adapting existing applications to new infrastructure, an experience that shaped his interest in difficult migrations and accumulated technical debt.

He founded Grit in 2022, initially applying early language-model APIs to framework upgrades, code modernization, and other maintenance work. Grit raised $7 million in seed funding and developed agents that could work across repositories and submit pull requests for human review. Its open-source GritQL engine searches, lints, and rewrites code using structural patterns instead of relying on imprecise text matching.

Honeycomb acquired Grit in April 2025, bringing Pell into its engineering organization to connect codebase analysis with production observability. The standalone product was discontinued, while GritQL remained open source. He subsequently joined Cognition, extending his work on autonomous software development.

• Expert-directed coding agents: Pell designs for principal engineers who can specify a migration, dispatch agents across repositories, and inspect the resulting pull requests. His OpenTelemetry example shows how centralized technical judgment can replace extensive cross-team coordination while preserving human review.

• Structural code transformation: GritQL combines syntactic patterns, semantic matching, and dependency awareness to locate the specific calls and imports requiring modification. Targeted replacements avoid the expense of regenerating entire files and the brittleness of model-generated line-numbered diffs.

• Compiler-grounded feedback loops: Agents become more dependable when they receive actionable signals from type checkers, language servers, and tests. Pell uses incremental TypeScript analysis to validate edits without rebuilding an entire application, then checkpoints working states and uses Firecracker snapshots to compare parallel candidate changes. His analysis of AI-era continuous integration extends this concern to testing and deployment bottlenecks.

• Production-aware code agents: Combining repository context with observability data gives automated systems a clearer picture of how software behaves after deployment, helping engineers identify consequential problems and evaluate whether proposed fixes improve real production systems.