A field guide for the agentic era

Ship faster.
Prompt smarter.

The best engineers of this decade won't just write code — they'll direct it. These are the skills that separate people who use AI from people who ship with it.

~/skills — zsh
claude "refactor the billing module, keep the tests green"
· reading 14 files … planning … editing … running tests
42 tests passed · 0 regressions · 9 minutes saved before lunch
01

Six skills that compound

/01

Context curation

The model is only as good as what it can see. Feed it the right files, docs, and constraints — ruthlessly prune the rest.

/02

Spec-first prompting

Vague in, vague out. State the outcome, the constraints, and what "done" looks like before a single token is generated.

/03

Agentic delegation

Know when to hand the model the whole task versus the next step. Scope work like you'd brief a sharp new teammate.

/04

Trust, then verify

AI writes it, you own it. Review diffs like a staff engineer: tests, edge cases, and the question "why this way?"

/05

Debugging together

Paste the error, sure — but also the hypothesis. The loop tightens when you reason with the model, not at it.

/06

Taste

The rarest skill. Knowing which of five working solutions is the right one — that's still all yours.

Paulius, admiring this page
🤔
photo of Paulius
loading admiration…
● live admiration feed
Hmm. Yes. This is exactly what peak AI-assisted engineering looks like.
Paulius Chief Admiration Officer · pictured mid-admire, moments after this page deployed itself
02

The tools got smarter. Your move.

Every skill above is learnable in an afternoon and compounds for a career. Open a terminal and start.

Start coding with AI →