Prukalpa

Author of Context & Chaos (Substack); writes on enterprise data, context engineering, and the semantic layer that AI agents need to operate inside companies. Best known publicly as co-founder of Atlan, a data catalogue / control-plane company that has been pushing the "context for data" framing since pre-LLM days — Prukalpa's positioning since 2026 has been that LLMs make the context problem urgent rather than discretionary.

What she argues

Her June 2026 essay "What an Enterprise Context Layer Actually Is" decomposes the context layer into three substrates (data + knowledge graph; semantics/ontology; skills) and five capabilities (context mining; development lifecycle; compounding learning loops; activation/retrieval; governance/observability). The reframe that matters: skills (procedural knowledge — how work gets done) are part of the substrate, not a downstream consumer of it.

This puts her view alongside the vault's existing context-engineering frames:

• Architecture view (Martin Keen / IBM) → Context Engineering four pillars (connected access, knowledge layer, precision retrieval, runtime governance).
• Process view (Patrick Debois / Tessl) → Context Development Lifecycle (Generate → Evaluate → Distribute → Observe).
• Enterprise scope view (Prukalpa) → substrate + capabilities across the whole org's tools and artefacts.

The three views compose rather than compete.

Why she matters here

She's the first author in the vault writing the context-engineering thesis from the data side (catalogue, semantic layer, glossary) rather than the agent-tooling side. That fills a gap — most other sources here (Debois, Keen, Lopopolo) are coming from the agent/prompt/harness layer. The user's Data Democratisation in Sales — Governed Context Layer, Not Dashboard Access POV builds on her framing.

Sources

• What an Enterprise Context Layer Is (Prukalpa) — her June 2026 Substack essay