Markdown source

Why your company needs a context graph, and how to build it

Conference Context

Session Description

Everyone building AI products eventually draws the same diagram: boxes representing data sources, arrows pointing at the model, and a label that says "context." What that diagram doesn't show is the system that has to run underneath it deciding, for each request: which sources to consult, whether to fetch live or use cached data, if the user is actually allowed to view that data, how to stitch it all together before the latency budget runs out. And it hides the counterintuitive part: fetching more context usually makes your answers worse, not better. At Merge, we reframed context graphs as control planes, helping companies scale context graphs to hundreds of thousands of users with sub-300 ms latency. This talk walks engineers through the system design at scale: how to tier data freshness, why provenance isn't optional once third-party systems are in the loop, and how to decide when fetching less context is the right call. Attendees will leave with a mental model for context system design that separates the orchestration decisions from the retrieval layer.

Media Evidence

No related AI Engineer channel video found yet.

Evidence Graph

This evidence graph is generated from currently linked source material: official schedule text, related video pages, cached transcripts, visible slide text, dense/reconstructed slide pages, and AI slide-classification audits.

Media Signals

No linked video, transcript, or slide source has been attached yet.

Agent Reading Notes

Use these signals to refine the synopsis, topic links, people/company context, and method notes. If a source is a related external video rather than an exact official recording, keep it framed as supporting evidence.

Transcript Status

No official session recording transcript was found by exact title match on the AI Engineer YouTube channel during this run.

People

Notes

Synthesis

Synthesized Breakdown

Why your company needs a context graph, and how to build it ## Conference Context - Date/time: 2026-06-30 · 1:55pm-2:15pm - Track/room: Expo Stage 3 · Expo Stage 2 NW - Speaker(s): Gil Feig - Session type/status: session · confirmed - Track: Expo Stage 3 - Room: Expo Stage 2 NW - Session type: session - Status: confirmed ## Session Description Everyone building AI products eventually draws the same diagram: boxes representing data sources, arrows pointing at the model, and a label that says "context." What that diagram doesn't show is the system that has to run underneath it deciding, for each request: which sources to consult, whether to fetch live or use cached data, if the user is actually allowed to view that data, how to stitch it all together before the latency budget runs out. And it hides the counterintuitive part: fetching more context usually makes your answers worse, not better. At Merge, we reframed context graphs as control planes, helping companies scale context graphs to hundreds of thousands of users with sub-300 ms latency. This talk walks engineers through the system design at scale: how to tier data freshness, why provenance isn't optional once third-party systems are in the loop, and how to decide when fetching less context is the right call.

Speaker And Company Context

Topics Covered

Derived Links And Source Material

Novel Concepts / Clever Methods

Evidence Boundary

This synthesis is based on the official schedule and linked source pages. It should be revisited when exact session recordings or transcript-backed secondary sources are available.