How to Generate Mergeable Code with a Context Engine — Peter Werry, Unblocked
Unblocked's context engine solves the agent knowledge-reset problem by delivering organizational context continuously
“Agents are like new employees. they reset their knowledge every time you start a new task.”
Peter Werry of Unblocked argues that AI coding agents fail to produce mergeable code because they lack organizational context — effectively re-onboarding from scratch on every task. Unblocked's context engine aims to surface codebase tribal knowledge, architecture decisions, and deployment conventions to agents automatically. The talk frames this as the next frontier on an 'AI maturity curve' that progresses from autocomplete to agentic coding.