The Pipeline Is Dead
summary
Claim: the one-version pipeline is obsolete because code can be produced at runtime and personalized for each user. Built prototypes at Differ that deploy a stem plus immutable divergences so each session runs an isolated variant; examples include CRM personalization where fields and flows adapt and can roll back. Key mechanics are inspectable attributable and reversible divergence provenance, testing at scale, and coordination via merge intent not code. Hard problems remain; work is ongoing, not finished.
AI-written summary · about
chapters (11)
- 0:00Frozen artifact assumption
- 1:45Pipelines as default
- 3:53Cost collapse per-user software
- 5:17Personalized software demand
- 7:48One stem many divergences
- 8:31Brittleness concerns
- 11:24Adaptive software CRM example
- 13:13Adaptive software challenges
- 16:26Autonomy versus control
- 17:51Beyond code generation
- 18:55Adaptive software inevitable
related talks
34mBeyond Continuous Delivery
44mThe Pipeline Driven Organization|
41mThe Pipeline-Driven Organization
34mBeyond Continuous Delivery
21mDesigning Modular Data Pipeline
36mDeveloper Experience Trends & Predictions
40mDeclarative Pipeline 2019 - Tips, Tricks, and What's Next
44mQA in a Pipeline Based Organization
