Coding is Easy, Reviewing it is Hard: Factory, Atlassian, & Entire on the Future of AI Code Review

Coding is Easy, Reviewing it is Hard: Factory, Atlassian, & Entire on the Future of AI Code Review

Oct 9 · 28 min

People

Listen on

Suggest an edit

Madrona Venture Partner Ted Kummert moderates a panel with former @GitHub CEO Thomas Dohmke (Entire), @Atlassian CTO Taroon Mandhana, and Factory co-founder and CTO Eno Reyes on where software engineering goes as coding agents gain autonomy. They cover the code review and validation bottleneck, the role of context in agent performance, what replaces reading code, and how the engineer’s job changes. The panel closes with predictions on open model share of code generation tokens. 00:00 – Introduction 01:53 – How Autocomplete, Tool Calling and Looping Led to Today's Coding Agents 04:04 – How Atlassian Automates Vulnerability Fixes From Jira Ticket to Pull Request 06:46 – Why the Code Review Bottleneck Got 10x Bigger With AI Agents 09:08 – What Six Sigma and Self-Driving Car Safety Thresholds Teach About Validating AI Code 11:53 – Three Requirements for Enterprise Context: Comprehensive, Relevant, Self-Learning 16:49 – Why Entire Stores the Agent Session Log in Git Next to the Code 18:00 – How an Agent Trail and Evals Let You Merge Code Without Reading It 20:20 – What Software Engineers Do When Agents Run the Software Factory 22:14 – Why Every Engineer Now Needs Staff-Level System Thinking 23:29 – Why Creativity and Business Sense Matter When Everyone Can Run Agents 24:49 – Predictions: Open Models at 95% of Code Gen Tokens vs. Double Digits

Atlassian Claude GitHub

Chapters

  1. 0:00 Introduction
  2. 1:53 How Autocomplete, Tool Calling and Looping Led to Today's Coding Agents
  3. 4:04 How Atlassian Automates Vulnerability Fixes From Jira Ticket to Pull Request
  4. 6:46 Why the Code Review Bottleneck Got 10x Bigger With AI Agents
  5. 9:08 What Six Sigma and Self-Driving Car Safety Thresholds Teach About Validating...
  6. 11:53 Three Requirements for Enterprise Context: Comprehensive, Relevant...
  7. 16:49 Why Entire Stores the Agent Session Log in Git Next to the Code
  8. 18:00 How an Agent Trail and Evals Let You Merge Code Without Reading It
  9. 20:20 What Software Engineers Do When Agents Run the Software Factory
  10. 22:14 Why Every Engineer Now Needs Staff-Level System Thinking
  11. 23:29 Why Creativity and Business Sense Matter When Everyone Can Run Agents
  12. 24:49 Predictions: Open Models at 95% of Code Gen Tokens vs. Double Digits