Waldek Mastykarz

Imagine you're building an eval to test whether an AI coding agent can deploy an application to your cloud platform.

Your coding agent just scored 100% on an eval, and every criterion passed.

"Does the solution follow best practices?"

As models get better, your agent tools should get smaller.

When you build an extension for an AI coding agent, it's tempting to include its invocation in your evaluation criteria.

Using an app isn't an experience.

Using an AI agent to run a command can seem ridiculous.

When you evaluate an AI coding agent, it's tempting to focus on one number: the score.

You run an eval.

When building evals for AI coding agents, string contains is incredibly tempting.

Just because an app is used in an enterprise doesn't make it an enterprise app.

When a model has a knowledge cutoff of June 2025, it's tempting to imagine that its knowledge represents the state of the world in June 2025.

There is a seductive argument making the rounds now that we can use AI to "just" replace "old" enterprise apps with their newer versions.

Is AI getting cheaper or more expensive?

Turning a collection of books, papers, or documentation into an agent skill is increasingly easy.

Whenever I have an idea, I talk to an LLM.

Coding agents make it incredibly easy to skip the learning.

If you built it, so can I.

A paragraph is a thought and its development.

Your product didn't get worse.

We have continuous integration and continuous deployment.

Personally, I don't care whether you wrote what I'm reading.