specify the acceptance criteria
The prompt works better when it states what “done” means — inputs, outputs, edge cases — because the model fills gaps with guesses otherwise.
§ 02 — Eras · 2023→
Prompts
natural language as the top of the language stack
Left: era context, features, and the 101 hello. Right: 3 deeper 201 lessons, the factorial specimen, and where Prompts lives today — each column scrolls on its own.
The problem it solves
Syntax was programming’s last gatekeeper. Prompts let anyone describe the result — and receive working code.
The problem it carries
No spec, no types, no determinism: the same prompt, a different program. Verification becomes the job.
Key features
- intent as input
- You describe the outcome; the model writes the code that produces it.
- iteration as the loop
- Refine by talking — each turn rewrites the artifact rather than patching it.
- context as the program
- Files, examples and constraints supplied in the prompt shape the output more than the words do.
- verification debt
- Correctness now depends on checking the result, not on the type system.
101First contact — hello world. 201 — 3 lessons on mechanisms you meet after the factorial specimen: the sharp edges practitioners actually use.
201 — going deeper
The 201 assumes you have read the 101 and the specimen. Each lesson is one mechanism — what it is for, how it looks in code, and the trap or payoff that makes Prompts distinctive at scale.
iterate against tests
Give the model the failing test and let it converge; asking for code that satisfies a known assertion is far easier than asking for correct code.
context windows are the budget
Everything relevant must fit: files, schemas, prior decisions. Curating that context is the real skill behind good prompting.
The specimen — factorial
The intent: say what the page should do, receive working code — natural language as the top of the language stack, every prior language beneath it.
Make a timeline page of shipping
routes, navy aesthetic, vanilla JS.Run it today
The tooling is a text box. Paste the specimen into any chat assistant — or straight into an agent that writes the file.
Where it is now
Already writes first drafts, tests, migrations and whole apps. Its costs are the mirror of its power: no spec, no types, no determinism — the same prompt, a different program. Verification becomes the job.