// Field notes

Field notes from AI-native engineering.

Short essays on what actually changes when AI is in the loop — and what doesn't. Drafted from real ships. Published as they're written.

Practice
Drafting

Why we stopped writing prompts and started writing specs

The shift from prompt-as-feature-description to spec-as-contract — and what changed when we made it.

—

Practice
Drafting

Context is the program

Same model, same prompt, different context, completely different output. A practical guide to engineering what the model sees.

—

Manifesto
Planned

The unglamorous half of AI engineering

The work AI never adds unless told to: secrets, timeouts, idempotency, runbooks. Why this is the half that matters.

—

Manifesto
Planned

Architectural taste in a regression-to-the-mean machine

LLMs converge on the median codebase. Notes on the bets only humans can make — and why they compound.

—

Process
Planned

The intern test

A diagnostic for telling whether you're prompting an AI or programming with one. If it fails the test, you're not yet senior with AI.

—

Tight specs, hard deadline, real users?

Start a project

// Field notes

Field notes from AI-native engineering.

Short essays on what actually changes when AI is in the loop — and what doesn't. Drafted from real ships. Published as they're written.

Practice
Drafting

Why we stopped writing prompts and started writing specs

The shift from prompt-as-feature-description to spec-as-contract — and what changed when we made it.

—

Practice
Drafting

Context is the program

Same model, same prompt, different context, completely different output. A practical guide to engineering what the model sees.

—

Manifesto
Planned

The unglamorous half of AI engineering

The work AI never adds unless told to: secrets, timeouts, idempotency, runbooks. Why this is the half that matters.

—

Manifesto
Planned

Architectural taste in a regression-to-the-mean machine

LLMs converge on the median codebase. Notes on the bets only humans can make — and why they compound.

—

Process
Planned

The intern test

A diagnostic for telling whether you're prompting an AI or programming with one. If it fails the test, you're not yet senior with AI.

—

Tight specs, hard deadline, real users?

Start a project

// Field notes

Field notes from AI-native engineering.

Short essays on what actually changes when AI is in the loop — and what doesn't. Drafted from real ships. Published as they're written.

Practice
Drafting

Why we stopped writing prompts and started writing specs

The shift from prompt-as-feature-description to spec-as-contract — and what changed when we made it.

—

Practice
Drafting

Context is the program

Same model, same prompt, different context, completely different output. A practical guide to engineering what the model sees.

—

Manifesto
Planned

The unglamorous half of AI engineering

The work AI never adds unless told to: secrets, timeouts, idempotency, runbooks. Why this is the half that matters.

—

Manifesto
Planned

Architectural taste in a regression-to-the-mean machine

LLMs converge on the median codebase. Notes on the bets only humans can make — and why they compound.

—

Process
Planned

The intern test

A diagnostic for telling whether you're prompting an AI or programming with one. If it fails the test, you're not yet senior with AI.

—

Tight specs, hard deadline, real users?

Start a project