How I One-Shotted an Astro Website With GPT 6 Astra in GitHub Copilot

Key Takeaways

• GPT 6 Astra in GitHub Copilot Chat built a working Astro and shadcn/ui heart rate zones site from one prompt.
• The site went from first prompt at 9:20am to live on Vercel with a custom domain at 9:55am.
• The initial prompt used $3.32 of a GitHub Copilot Pro+ monthly allowance.
• The main weakness was too many unnecessary taglines and subheadings, fixed with one short follow-up prompt.
• Training articles, race plans and Brevo email capture were each added with a single follow-up prompt.

What prompt did I use to one-shot the site?

The prompt that built My Heart Rate Zones was one short paragraph, typos included:

this is a heart rate zones calculator app. it needs to be a basic site with the appropriate options to calculate heart rate zones (mainly for runners), use shadcn ui for components, look incredibly on both desktop AND mobile, and have a sports / fitness aesthetic to the site (i.e. it make sure it doesn’t look like just any other shadcn ui site!)

I ran it inside an empty Astro project. Astro is a web framework that ships static HTML by default and adds JavaScript only where a page needs it, which suits a calculator with one interactive component.

What did GPT 6 Astra build from that prompt?

GPT 6 Astra produced a finished calculator, not a scaffold. It chose sensible running-specific options without being told which ones:

Feature	What it does
Two calculation methods	Percent of max heart rate, or heart rate reserve (Karvonen)
Max HR estimate	Tanaka formula: 208 minus 0.7 times age
Five zones	Recovery, Endurance, Aerobic, Threshold, Maximum, each with run types
Privacy	All calculations stay in the browser
Safety note	A medical disclaimer for adults and medication users

The design request mattered most to me. shadcn/ui sites often look identical, and the model gave this one a bold, sporty layout with numbered sections and a running photo hero. It looked GOOD on mobile on the first pass.

What went wrong?

The only real problem with the GPT 6 Astra output was copy bloat: too many taglines and subheadings. Almost every block had a catchy line above it and a supporting line below it, which made the page feel busier than the actual tool. The fix took one follow-up prompt asking the model to remove the taglines, and it took about two seconds to write.

How long did it take to go live?

The full launch took 35 minutes from first prompt to a public URL. In that window I:

1. Submitted the prompt at 9:20am
2. Bought the domain while the model was working
3. Reviewed the output and deployed it to Vercel
4. Pointed the domain at the deployment and went live at 9:55am

What features did I add after the first prompt?

I added three features after launch, each with one more prompt: a set of articles on heart rate zone training, custom 5K, 10K, half marathon and marathon training plans, and a Brevo integration that captures user emails. None needed a second round of fixes.

Frequently Asked Questions

Can GitHub Copilot build a full website from one prompt?

Yes. As of October 2026, GPT 6 Astra in GitHub Copilot Chat built a deployable Astro site with working logic, responsive layout and custom styling from a single paragraph.

Do typos in the prompt matter?

Not in my experience. My prompt had several typos and GPT 6 Astra still followed every requirement.

How do I stop the AI adding too much copy?

Ask for it directly in the prompt. A line such as “keep headings minimal and avoid taglines” should prevent the clutter, and if it slips through, one follow-up prompt removes it.

Wrapping Up

One clear prompt got me a polished, mobile-ready Astro site for $3.32 and half an hour of my morning. You can try the finished calculator at myheartratezones.com. If you take the same approach, be specific about the look you want, tell the model to go easy on taglines, and ship it the same day.