The AI Native Dev - from Copilot today to AI Native Software Development tomorrow What Developers Need To Know About Agents Before 2026
32 snips
Dec 30, 2025 
Reuven Cohen, Founder of the Agentix Foundation, shares insights on why most agent systems fail and the demands of agentic engineering. Maor Shlomo, Founder of Base44, discusses the boundaries of vibe coding, highlighting its effectiveness for non-developers and the importance of guardrails for maintaining software quality. Maksim Shaposhnikov from Tessl contrasts terminal workflows with IDE agents, focusing on developer autonomy and the design of systems that promote maintainability. Together, they explore the future of software development and the evolving role of agents.
AI Snips
Chapters
Transcript
Episode notes
IDE Vs Terminal Tradeoffs
- IDE-based assistants keep developers in the loop with visual UI and safer onboarding.
- Terminal agents suit advanced users who prioritize speed and can handle less visibility.
Progress From Copilot To Terminal Agents
- Start with simple, familiar tools (Copilot in the IDE) and progress toward terminal CLIs as you gain trust and experience.
- Only delegate larger autonomous scopes after you validate assistant reliability in smaller tasks.
Why Sub-Agents Exist
- Sub-agents are spawned instances used to split complex tasks or explore alternative solution paths in parallel.
- They can speed up work and diversify approaches but may increase cost and require careful instruction design.