Annie Wang
Confidence 0.70 · 2 sources · last confirmed 2026-09-04
Presenter at Google Cloud on the AI Builder Essentials series, paired with Tilde Thurium in the two-person conversational explainers the wiki tracks. Promoted on the second-source rule.
The pairing is consistent across both videos and is part of what makes them useful: Thurium plays the sceptic — asking what the terms mean, admitting she confused graph engineering with knowledge graphs, closing the second video by asking whether the field is “just reinventing data structures and algorithms for the agentic age” — while Wang supplies the ADK-side definitions. She is the source of most of the vocabulary the corpus quotes from this series.
Her definitions, as the wiki cites them:
- Loop engineering — “you replacing yourself with a system… set a goal and let the system keep retrying until it’s meeting the goal.”
- Harness — “everything around the model including its tools, memory and guardrails.”
- Graph engineering — “the organization chart”: a workflow of agent nodes and function nodes, “basically what goes in, in what order, and then what happens next.”
- Knowledge graph vs graph engineering — “Knowledge Graph emphasises on the data and Graph Engineering emphasises on the behavior.”
She also supplies the corpus’s sharpest one-liner on self-evaluating agents (“it’s like asking a kindergartner to grade its own homework”), and — asked what the next vocabulary trend will be — the reply “abstraction-maxxing.”
Appears in this wiki via
- 2026-08-25-thurium-wang-google-cloud-four-ways-loop-engineering-fails — the four-mode loop failure taxonomy; the loop-engineering definition; the escalation to graph engineering.
- 2026-09-03-thurium-wang-google-cloud-graph-engineering-101 — the harness / loop / graph layering, the fan-out / join / router patterns, and the three disambiguations that gave the corpus its graph-engineering page.
Referenced by
- 2026-09-02-github-podcast-demystifying-ai-terms-loop-engineering-squads-harness — her loop-engineering definition is the inflationary half of the pairing the wiki draws against GitHub’s deflationary one.