Can AI write code?

The useful part

A guided interview helps you turn your own answers into a technical spec, small implementation steps, code-generation prompt, and acceptance tests. The assistant asks only a few questions at a time and waits for you to confirm the facts before drafting.

Before you start

You can begin without preparing a full brief. The assistant starts with:

1. What should the code do in one sentence?
2. What language or stack are you using?
3. Is there an existing repository or example input/output?

You can skip a question or say you do not know. Add private or confidential material only when you are allowed to share it.

How to do it

1. Answer the first three questions in the guided prompt, then continue in small rounds.
2. Check the numbered fact summary and correct any assumptions before asking for the draft.
3. Use the approved brief with Lovable; check the current interface and plan first.
4. Run tests and review generated code, dependencies and data handling before merging or deployment.

Copy this prompt

Help me create a technical spec, small implementation steps, code-generation prompt, and acceptance tests through a guided interview.
Wait for my answers and confirm the facts before drafting.

HOW TO WORK WITH ME
- Ask no more than three short questions per message, then wait for my reply.
- Use plain English. Follow up on my answers; do not repeat questions I have answered.
- If I do not know something, offer a sensible option and label it as an assumption.
- Keep facts, assumptions and questions separate. Never invent my details or outcomes.
- Adapt to my country, budget, experience and available tools.

START WITH ONLY THESE THREE QUESTIONS
1. What should the code do in one sentence?
2. What language or stack are you using?
3. Is there an existing repository or example input/output?
Then stop and wait.

IN LATER ROUNDS, INVESTIGATE WHAT MATTERS
- Ask about constraints; dependencies; security; tests; error cases; maintainability; deployment.
- Ask to see source material only when it would change the result. Let me remove private or confidential material first.
- If a requirement is missing, ask a focused follow-up instead of guessing.

BEFORE YOU CREATE ANYTHING
Summarise my confirmed facts as numbered items. Put assumptions and unanswered questions in separate lists. Show me the proposed scope and ask me to correct it. Wait for my confirmation before producing the deliverable.

AFTER I CONFIRM
Produce a technical spec, small implementation steps, code-generation prompt, and acceptance tests. Make the result immediately usable, with concrete wording, structure or steps rather than abstract advice. Label any placeholder clearly. Explain how to use the deliverable with an appropriate tool if needed. If that tool cannot accept a long prompt, give me a concise input that works across tools. End with a short manual quality-check list.

QUALITY AND HONESTY
Run tests and review generated code, dependencies and data handling before merging or deployment.
Do not claim you opened a link, saw a file, used a tool, checked current prices, or tested an export unless you actually did. When a fact depends on live product features, laws, availability or pricing, tell me to verify it at the source.

Begin with the three starting questions only.

What you should get

After you confirm the facts, the assistant should produce a technical spec, small implementation steps, code-generation prompt, and acceptance tests. Review every assumption and placeholder before using it.

Where AI falls short

Run tests and review generated code, dependencies and data handling before merging or deployment. This prompt and workflow have not been tested end to end with your own material.

When another tool helps

Lovable is the suggested execution or review tool for this task. Check its current features, price, region and export options before paying. The link is a normal product link, not an affiliate tracking link.