The AI Native Dev - from Copilot today to AI Native Software Development tomorrow What Holds Devs Back From Multi-Agent Thinking | Guy Podjarny
28 snips
Nov 26, 2025 
Guy Podjarny, Founder and CEO of Tessl and a serial entrepreneur in developer tooling, dives into the evolution of AI-augmented development. He explains how to bridge the capability-reliability gap with agents and advocates for context engineering as a vital part of the development process. Guy stresses the need for statistical measurement to assess agent success and shares insights on optimizing documentation with Tessl tiles for enhanced efficiency. He also highlights the significance of cross-team collaboration and the importance of keeping knowledge adaptable for various agents.
AI Snips
Chapters
Transcript
Episode notes
Powerful Yet Unreliable Agents
- Agents are powerful but often unreliable, creating a capability–reliability gap.
- Turning that raw capability into consistent developer productivity is nontrivial and requires deliberate strategies.
Too Much Context Reduces Focus
- Dumping all context into huge prompts dilutes attention and lowers performance.
- Models focus less on each item when you overload their context windows.
State Problems With Enough Context
- Practice context engineering: state problems with enough context so tasks are plausibly solvable without extra info.
- Give agents the specific facts they need instead of assuming they can mind-read.