Open Source Startup Podcast E179: LLMs for Software Maintenance (the Grit Story)
15 snips
Aug 18, 2025 Morgante Pell, Founder of Grit, an AI agent company focused on software maintenance, shares insights on automating software maintenance with LLMs. He discusses Grit's innovative tool, GridQL, designed for reliable code transformations, and the company’s motivation for going open source. Pell also highlights the challenges posed by AI-generated code in software maintenance and reflects on the strategic acquisition by Honeycomb, emphasizing the importance of building relationships in tech negotiations.
AI Snips
Chapters
Transcript
Episode notes
AI Fits Maintenance Workflows
- Morgante saw AI coding potential early and targeted maintenance as the right first use case.
- He viewed maintenance as low-interest, high-volume work an AI junior engineer could automate reliably.
Language As Determinism Layer
- Grit introduced GritQL because early LMs made deterministic, large-scale transforms unreliable.
- The language let AI generate small deterministic programs to run consistent repository-wide changes.
Have AI Write Deterministic Programs
- Use programs where repetition and exactness matter instead of asking the model to change thousands of files directly.
- Let the model write a transformation program and then run and review that program for reliability.