Vibe Coded Today

The daily paper for people who build by talking to software.

Front page

Ship real software by talking to it.

Prompts that fill themselves in with your project, guides that take a subject end to end, and fixes for the moment it breaks. Every page dated, checked, and free.

Today’s clipping · No. 32

Pull structured data out of documents

Where models are genuinely more reliable than parsing.

extract-fields.txt
Extract the following from this document: <list the fields and their
types>

Rules:
- Return JSON matching the schema exactly. Use null for anything not
  present — never guess, never infer from context.
- For each field, also return where in the document you found it, so
  I can verify.
- If a value appears more than once with different content, return
  both and flag the conflict rather than picking one.
- Dates normalised to YYYY-MM-DD. Amounts as numbers with the currency
  separate.

Document: <paste>

Asking where each value came from turns an unverifiable answer into a checkable one, and it costs almost nothing.

A different prompt every day. More from the Building with models desk

The one prompt to start every session

Paste this before you ask for any code. It stops the model inventing a stack you didn't ask for.

context-primer.txt
You are helping me build: <one sentence, what the thing does>

Stack I am already using: <languages, frameworks, hosting>
Files that already exist: <paste `ls` output or a short list>
What works right now: <the last thing you got working>
What I want next: <one change, not five>

Rules:
- Change the fewest files possible. Show me full file contents for anything you edit.
- Do not add new dependencies unless you say why and I agree first.
- If my request is ambiguous, ask one question before writing code.
- End with the exact command I run to see it working.

Why it works: most bad AI output comes from missing context, not a weak model. Nine lines of setup beats three rounds of "no, not like that".