Gradient Flow offers weekly analysis of applied AI and machine learning for enterprise teams, drawing on interviews with practitioners and recent incidents. Recent issues cover deep model customization, local versus hosted AI deployment, and security risks from AI agents.

Weekly essays on why AI initiatives at growth-stage startups fail for data governance and platform reasons rather than model reasons. Each issue uses concrete incidents, such as stale notebook pipelines, misleading catalogue descriptions and unmanaged documents, to explain the failure and what to change.

An engineering blog from the team building Sequin, a Postgres change data capture tool, covering pipeline architecture, delivery guarantees and product launches. Readers get detailed technical explanations of Elixir GenStage back-pressure, CDC duplicate handling and the design of Sequin's features.

This newsletter documents hands-on experiments with data platform technologies, such as agentic apps, small decision-focused LLMs and the evolution of dbt, and explains what the results mean for platform owners. Each issue pairs a first-person narrative with a built demo app, walkthrough video and comparison results.