go podcast() 077: LLMs, with great power comes great responsibility
Mar 19, 2026
A lively conversation about practical LLM uses in engineering workflows. They cover rapid prototyping, debugging production logs, and using LLMs for codebase exploration. The discussion tackles team impacts like review load, apprenticeship risks, and three-way pairing with LLMs. They debate tool preferences, CI integration, testing safeguards, and when to handcraft code versus relying on generated output.
AI Snips
Chapters
Transcript
Episode notes
Use LLMs To Validate Ideas Fast
- Use LLMs for rapid prototyping like POCs, understanding unfamiliar codebases, and planning feature feasibility.
- Ramesh used agentic tools to build prototypes and validate ideas quickly before committing months of engineering time.
Small Team Migrated To Cloud Code
- Ramesh's previous team experimented with Copilot, Cloud Code and ChatGPT; Cloud Code later became the standard and Copilot licenses were canceled.
- The CTO drove adoption to increase feature velocity while experienced devs could quickly spot issues.
Using LLMs To Learn Game Development Fast
- Ramesh used Cloud Code to build a Godot game despite being new to game dev and made steady progress across levels.
- He combined tutorials with targeted LLM prompts to overcome unfamiliar APIs and ship playable levels.