

#517: Agentic Al Programming with Python
158 snips Aug 22, 2025
This discussion features Matt Makai, VP of Developer Relations at DigitalOcean and creator of Full Stack Python, who dives into the world of agentic AI programming. He breaks down how coding assistants are evolving from simple autocomplete tools to collaborative partners. Topics include managing technical debt, the quirks of AI tools like Cursor and Claude Code, and effective git workflows for AI-driven changes. Matt shares insights on why typing less can lead to more bugs and the future of open-source agents, highlighting the shift towards developers as system editors.
AI Snips
Chapters
Transcript
Episode notes
Agentic AI Multiplies Productivity
- Agentic AI can 10x–100x productivity for specific software tasks when used appropriately.
- Use agents for repeatable, annoying work rather than replacing developer judgment or design decisions.
Use Good Models And Rich Context
- Use higher-quality models and give generous context to get useful results from LLMs.
- Provide multi-page prompts and plan steps like you would for a junior developer to improve outcomes.
Start With Read‑Only Code Analysis
- Start by using models in read-only mode to analyze code before letting them modify files.
- Gain confidence in their understanding of your project before asking for edits or refactors.