Summary

Joe Reis argues that the industry's approach of scaling central data warehouses to handle AI agent workloads starts from the wrong end of the architecture. He proposes an inversion: give each agent its own embedded DuckDB instance and let agents exchange immutable, semantically-identified analytical slices directly peer-to-peer, without routing through a shared analytical platform. The design embeds semantic contracts, provenance lineage, and compatibility checks into the data products themselves, replacing central authority with deterministic artifact-level trust.

Key Insight

The right architecture for agent-native analytics inverts the warehouse-centric model by giving each agent its own embedded analytical engine and encoding semantic meaning, provenance, and compatibility checks directly into immutable data products exchanged peer-to-peer — shifting trust from a central authority into the artifacts themselves.

Spicy Quotes (click to share)

• 7

  Everyone is trying to make the data warehouse ready for agents. I think that starts from the wrong end of the architecture, so I inverted the problem and built the opposite in working code.

• 3

  A human analyst may know that 'revenue' means revenue net of refunds because someone from accounting clarified it during a meeting. An agent may have access to portions of that context, but it cannot be relied upon to recover and apply the same institutional knowledge consistently.

• 4

  An LLM may propose intent and semantics, but trust decisions must be expressed as deterministic artifacts and checked mechanically before other systems depend on them.

• 3

  Derived data cannot promote itself to source authority — a derived slice receives DERIVED tier when it is created and cannot satisfy a request that explicitly requires authoritative source data.

• 4

  I suspect the industry's shift toward agent-native data will make peer-to-peer analytical exchange a necessity rather than an alternative.

Tone

analytical