Dev Interrupted Your agents are starving! Airbyte’s Michel Tricot on the data ingestion crisis
40 snips
Sep 8, 2026 
Michel Tricot, co-founder and CEO of Airbyte, explores the data ingestion crisis behind production AI. He discusses agent access and permission boundaries, resolving identities across siloed systems, search-ready data, and the balance between raw platforms and opinionated tools. He also examines adaptive business context, AI’s cloud-like transformation, and Airbyte’s Hydra project for automating connector maintenance.
AI Snips
Chapters
Transcript
Episode notes
Agents Turn Data Access Into Control
- Agentic systems turn familiar data access into a control problem because they act faster, at greater scale, and with less human judgment.
- Teams must govern identity, permissions, operation types, pace, and escalation while connecting databases to live SaaS and internal systems.
Entity Resolution Prevents Context Sprawl
- Connecting more systems creates a second problem beyond ingestion: agents must resolve whether records refer to the same person, customer, or feature.
- Accurate entity links let agents discover relevant Jira, Zendesk, Stripe, and product records without bloating context.
Make Agents Discover Context Gradually
- Retrieve only records and schemas required for the task instead of loading every available tool or source.
- Let agents discover context progressively, such as identifying a feature before tracing its Jira, Zendesk, and customer associations.