Episodes (Page 4)
✨
Classic golden signals are insufficient for GenAI apps.
Marina Petzel
✨
Coding agents pose significant security risks to your machine.
Rowan Christmas
✨
APIs evolved from completions to agents, requiring new interaction models.
✨
Hardware failures are inevitable in large-scale GPU training.
✨
Open LLMs are competitive, but production serving is complex and costly.
Sujee Maniyam
✨
A personal AI agent can run on a Raspberry Pi.
✨
VS Code moved to weekly releases by rebuilding systems around AI agents.
✨
AI-generated code complicates production debugging and troubleshooting.
✨
Frontier LLMs require massive synthetic token generation.
Bogdan Gaza
✨
AI code generation outpaces human review speed.
Gabriel Spencer-Harper
✨
Browser agents struggle with accuracy over many steps.
Derek Meegan
✨
Token budgets cripple developer productivity and increase costs.
Laurent Gil
Zilvinas Urbonas
✨
Models are outdated, missing knowledge of recent code changes.
Jakub Hojsan
✨
MCP protocol is not the issue with agent harnesses.
Jan Curn
✨
AI memory is a critical, often rented, component.
Dylan Couzon
✨
Justin Reock shared five AI trends impacting software development.
Justin Reock
✨
Rania Khalaf discussed the evolving Chief AI Officer role.
✨
Autonomous agents increased code output significantly but not shipping.
Laurie Voss
✨
Connecting agents manually is inefficient; loop engineering helps.
Vlad Luzin
✨
AI agents wrote over 25% of pull requests reviewed by Greptile in April.
Human Code