AI engineering & evaluationSkylar Payne / Fieldwork

A little less
artificial.
A lot more
useful.

I’m Skylar. I build AI systems, figure out why they break, and write about what actually makes them work.

Explore my writing About me
Fig. 01 / The useful machine
From a messy idea to something usefulA playful illustration of tangled inputs passing through an evaluation machine and becoming ordered outputs. Use the Run experiment button to try another pattern. This is an illustration, not a live AI system. MESSY INPUTUSEFUL OUTPUT BUILD → TEST → LEARN TRY. BREAK. UNDERSTAND. REPEAT.
Experiment 01 · ready to tinker

Less magic. More turning the knobs. ↖

Ten years building AIGoogle LinkedIn Startups
01 /

Selected writing

Essays on agents, evaluation, and engineering.

Featured essay ·

If It Has to Happen, Don’t Put It in the Prompt

When agent instructions turn into all caps rules, the fix is often to move the requirement out of the prompt and into a workflow that can check it.

Read the essay
8 recent pieces

GPT-5.4 in OpenClaw doesn’t suck. Your prompts do.

Essay

If DSPy is So Great, Why Isn't Anyone Using It?

Essay

Evals That Actually Get Used

Essay

Bringing Data Science Back to AI Engineering

Essay

A Practical Way to Evaluate Multi-Step and Multi-Turn Agents

Essay

Patterns for AI Reliability

Essay

Artificial Intelligence for Business

Essay

The Answer is Always Evals

Essay
All writing
02 /

About me

Skylar
PayneEngineer / Writer
Founder, Wicked Data

My work

I spent ten years building AI at Google, LinkedIn, and startups. Now I work with engineering teams through Wicked Data.

I help teams review model outputs, investigate failures, and test changes in their codebase. I write about that work here.

More about me ↗
© 2026 Skylar PayneFieldwork