From a BRD to a shipped scoring platform in 13 hours
MagnaCarta went from a business requirements doc to a live Next.js platform — paste a Deliveroo URL, scrape the menu, score it on two rubrics — in 13.1 hours across 52 commits. The BRD → 369-line prototype spec → CLAUDE.md → code chain is exactly why I argue the spec matters more once AI writes the code, not less.
[Standfirst for the "From a BRD to a shipped scoring platform in 13 hours" case — the problem in one paragraph.]
The BRD: what 'menu quality' actually means
[Draft section: the brd: what 'menu quality' actually means. Written when this piece is published — left as a placeholder here rather than fabricated.]
Turning the BRD into a 369-line prototype spec
[Draft section: turning the brd into a 369-line prototype spec. Written when this piece is published — left as a placeholder here rather than fabricated.]
Dual scoring: aggregator completeness + MagnaCarta quality
[Draft section: dual scoring: aggregator completeness + magnacarta quality. Written when this piece is published — left as a placeholder here rather than fabricated.]
Why the spec was the real work, not the code
[Draft section: why the spec was the real work, not the code. Written when this piece is published — left as a placeholder here rather than fabricated.]
When you hit a wall
Two ways to go further — sold separately, never bundled.
Workspace credits
Buy a credit pack
On-site AI runs on free daily credits first, then paid. Packs are a one-time purchase — no subscription, ever.
One-time · Stripe checkout
1:1 consulting
Talk to me directly
30 min on a call, open scope — bring whatever you're stuck on. Booked separately from credits.
Calendly · sold separately