Medium

Ka Ling Wu

Ka Ling Wu

·

1d ago

dbt’s New Metrics Spec: What Changed in MetricFlow and How to Migrate

Measures are gone, semantics move onto columns and type_params flattens. A side-by-side migration guide, plus the gotchas autofix misses.

dbt MetricFlow new spec: legacy semantic_models and measures YAML next to the latest spec with semantics on columns

Ka Ling Wu

·

1d ago

The First Day of a Data Agent Shouldn’t Be Boilerplate

Part 5 of 14 · The Complete Agent Stack · Ontology. How a setup agent writes the first draft of your ontology, so you can spend your time…

Title card for Upsolve Self-Serve Onboarding, Feature 14 of 14 in The Complete Agent Stack

Ka Ling Wu

·

5d ago

Context That Ships Together Stays Together

Prompts, skills, data models and examples drift apart when they live in different places. Why an agent should ship as one release.

Title card for Upsolve Workspaces, Feature 11 of 14 in The Complete Agent Stack

Ka Ling Wu

·

6d ago

What the Research Actually Says About Semantic Layers and Text-to-SQL Accuracy

Seven studies, what each one measured, what none of them tested, and the numbers that keep getting misquoted.

Three text-to-SQL accuracy studies on semantic layers shown as cards with their small sample sizes: 11, 60 and 100 questions

Ka Ling Wu

·

Oct 27

Semantic Layers Were Built for Dashboards. AI Agents Need More.

What a semantic layer for AI agents still does well, and the structure and trust it never had to provide for a person with a dropdown.

Semantic layer for AI agents: a dashboard filter lists valid status values; an agent filters on one that doesn’t exist

Ka Ling Wu

·

Oct 22

Context Engineering Statistics & Insights in 2026

The research behind context engineering this year: what bigger context windows didn’t fix, where agents spend their tokens, and where…

Context Engineering Statistics & Insights in 2026

Ka Ling Wu

·

Oct 20

Your Agent Shouldn’t Do Joins in Its Head

Part 3 of 14 · The Complete Agent Stack · Ontology. Why cross-source analysis belongs in SQL and not in a context window.

Title card for Upsolve Data Plane, Feature 07 of 14 in The Complete Agent Stack

Ka Ling Wu

·

Oct 19

We Tested How Much Context Our Analytics Agent Actually Needs

Same model, same questions, five ways of filling the context window, and what the published research says you’ll find.

Five context windows side by side: sparse, overflowing, curated, curated with highlighted distractors, and a smaller model

Ka Ling Wu

·

Oct 15

Before You Buy an AI Analytics Tool, Ask What the Model Actually Sees

Every demo looks the same. How a vendor answers this question, and its follow-ups, tells you whether the tool works from your definitions…

Side-by-side panels: the churn chart a user sees, and the model’s trace with two risky context lines highlighted

Ka Ling Wu

·

Oct 14

Never Let the Agent Hold the Keys

Part 2 of 14 · The Complete Agent Stack · Ontology. Why data permissions belong outside the agent, where no prompt can reach them.

Title card for Upsolve Row-Level Security, Feature 02 of 14 in The Complete Agent Stack

Ka Ling Wu

Ka Ling Wu

Co-founder & CEO of Upsolve AI (YC W24). Building context infrastructure for trustworthy analytics agents. Ex-Palantir, Brown University.

Following

Help

Status

About

Careers

Press

Blog

Store

Privacy

Rules

Terms

Text to speech