Software Engineering Daily Scaling AI in Enterprise Codebases with Guy Gur-Ari
10 snips
Oct 9, 2025 
Guy Gur-Ari, co-founder of Augment and a former Google research scientist, dives into the transformative world of AI coding assistants. He discusses the rise of vibe coding and its impact on software creation, contrasting it with the challenges of maintaining complex legacy codebases. Gur-Ari explains how Augment focuses on contextual understanding and automating code reviews, which are becoming bottlenecks in AI-driven development. He also predicts that developers will increasingly manage fleets of AI agents to enhance efficiency and innovation in engineering.
AI Snips
Chapters
Transcript
Episode notes
Code Lets Models Close The Loop
- Language models reached practical usefulness and coding is the next major reasoning domain after math and science.
- Code excels because we can execute and close the loop between generation and runtime feedback.
Feed Agents All Available Feedback
- Provide agents with available signals like type checks, linter errors, and tests to improve correctness.
- Surface runtime signals (logs, metrics, traces) later to further ground agent decisions.
Autonomy Plus Infinite Context
- Keep agents autonomous but avoid stuffing context unless the agent requests it or you know it's necessary.
- Implement an 'infinite context' engine to surface relevant code without exposing token limits to users.
