Blog

Thoughts on the future of data infrastructure.

FeaturedProduct

Why AI Agents Hallucinate on Your Data (And How to Fix It)

AI agents writing SQL against your data warehouse get it wrong 66% more often without semantic grounding. Here is why context is the missing layer in every data stack — and what we are building to fix it.

Read article
Industry

The Context and Semantic Layer Market: Why Nobody Has Solved This Yet

We mapped the entire landscape of data context and semantic layer tools. Here is what we found and where the gaps are.

12 min read
Industry

What We Learned Studying the Data Engineering Market Before Building

Before we wrote a single line of product code, we spent four months doing something unsexy: reading earnings calls, mapping vendor acquisitions, talking to data engineers, and building spreadsheets of market gaps.

10 min read
Engineering

Why We Bet on MCP (And What We're Still Figuring Out)

When we started building Data Workers, we had to make a foundational decision: how do our AI agents connect to the dozens of tools in a modern data stack?

8 min read
Industry

Why Your Data Stack Still Needs Humans at 2 AM

It is 2026 and your data pipeline still breaks at 2 AM. Not because you chose bad tools. Because the fundamental problems of data engineering are genuinely hard.

7 min read
Product

Copilots, Agents, and Swarms: A Decision Framework for Data Teams

The AI discourse in data engineering has collapsed into a single word: agents. Every vendor is an "agent" now. The word has lost meaning.

8 min read
Engineering

Building an Incident Debugging Agent: What We've Learned So Far

Incident debugging is where we started building. Not because it is the easiest problem, but because it is the most painful.

12 min read
Engineering

Building a Quality Monitoring Agent: Lessons From Alert Fatigue

Alert fatigue is the silent killer of data quality programs. You deploy a quality tool, configure monitors, and within weeks you are drowning in alerts.

10 min read
Product

Our Agent Roadmap: What We've Built, What We're Building, and Why

The average data team spends 60-70% of their time on reactive maintenance. Here is how 11 specialized agents address that — and why the order we build them matters more than the features.

12 min read
Engineering

Why Schema Changes Are the Silent Killer of Data Pipelines

A column gets renamed upstream. Nothing breaks immediately. Three days later, an executive notices a dashboard looks wrong. Sound familiar?

8 min read
Engineering

Why Building a Data Pipeline Still Takes Weeks in 2025

Describe what you need. The agent generates SQL transformations, orchestration DAGs, quality tests, and documentation — then deploys it.

9 min read
Engineering

Your Data Governance Is a PDF Nobody Reads

Most enterprises have governance policies documented in PDFs that gather dust. We are building an agent that turns them into executable code.

9 min read
Engineering

The 'One Person Who Knows Kafka' Problem

Streaming expertise is scarce and expensive. When your Kafka person goes on vacation, issues wait. When they leave, you inherit a system nobody fully understands.

8 min read
Product

Introducing the ML Agent: From Data to Deployed Model in Minutes

Most ML workflows still live in notebooks — disconnected from the data stack. We built an ML agent that connects model training to your catalog, quality checks, and deployment pipeline.

9 min read
Industry

Why Your Data Stack Needs an ML Agent (Not Just a Notebook)

The notebook-to-production gap kills more ML projects than bad algorithms. An ML agent that lives inside your data stack closes that gap.

11 min read
Engineering

Why Your 200 DAGs Need a Brain (Not Another Dashboard)

Your orchestrator runs DAGs. But it does not know that three of them have undocumented dependencies that cause cascade failures every month.

8 min read
Product

You Are Paying for 30-40% of Data That Nobody Uses

30-40% of enterprise warehouse storage is wasted on data nobody queries. We built an agent with four tools that finds unused data, estimates savings, and recommends safe archival — but never auto-archives without human approval.

10 min read
Product

Why Every Enterprise Migration Takes 6-18 Months (And How to Fix It)

Every enterprise has a migration project. Oracle to Snowflake. Teradata to Databricks. These projects cost $2M-$5M and overrun constantly. We built an agent that classifies complexity, translates SQL dialects, and generates validation queries — automating the 80% that is mechanical.

12 min read
Product

Why Text-to-SQL Without a Semantic Layer Is Dangerous

Text-to-SQL is the most overpromised feature in data tooling. Without semantic grounding, it is a hallucination engine disguised as a search bar.

8 min read
Product

Your Team Uses Dozens of MCP Tools Every Day. Can You See How?

Every MCP tool call is a signal of practitioner intent. By treating tool calls as usage telemetry, you get something no other platform offers: a map of how data teams actually work.

10 min read
Engineering

Dear Open Claw: Your Data Stack Is Now an MCP Call Away

An open letter to AI coding agents, data agents, and autonomous systems everywhere. We built 202 MCP tools across 14 specialized agents so you can query catalogs, build pipelines, debug incidents, and govern data — without guessing. Here is how to use them.

9 min read
Engineering

Why We Open-Sourced 14 Autonomous Data Engineering Agents

We released 14 MCP-first autonomous agents for data engineering under Apache 2.0. Here is why we chose radical transparency over a black-box product, and how the trust model works.

9 min read
Industry

Why Every MCP Server for MLOps Is Read-Only (And Why That Is About to Change)

The MCP ecosystem exploded in 2026. Every major MLOps platform shipped an MCP server. But they all share one blind spot: none of them can actually operate your ML infrastructure. Here is why that matters and what comes next.

10 min read