People imagine development as one long stretch of typing. In reality, code is only the middle act. The work that determines whether a project succeeds happens before the editor opens and after the last feature lands. This is the workflow I've settled into — six movements, in order, no skipping.

1. Requirements: shrink the fog

Every project arrives wrapped in fog: 'I need a website for my business.' The first job is asking questions until the fog lifts. Who is this for? What should they be able to do? What exists already? A project with clear requirements feels shorter before it begins — because the surprises have been moved to the cheap end of the timeline.

2. Structure: decide the skeleton

Pages, data, flows. What models exist — products, orders, appointments — and how do they relate? What does the URL map look like? Ten decisions here prevent a hundred refactors later. I sketch the schema and the sitemap on paper first; paper argues less than code does.

3. Design: the interface is the argument

With structure settled, design becomes fast and honest: real content, real constraints, hierarchy built around what the user came to do. This is also where the design system is born — colors, spacing, type and the first components.

4. Development: build in walking slices

I build in thin, end-to-end slices — a journey that works completely — rather than wide layers of half-finished everything. Each slice is committed, reviewable and demonstrable. Version control isn't just backup; it's the project's memory, and I write commit messages for the future human who'll wonder why.

A slice, not a layer
# Walking slice: "a visitor can buy one product"

[✓] Product page renders from the database
[✓] Cart accepts the product
[✓] Checkout records the order
[✓] Confirmation email triggers

# Four thin layers that WORK —
# not four thick layers that almost work.

5. Testing: be your own unfriendly user

Then I switch sides and try to break what I built: small screens, slow networks, keyboard only, forms filled with chaos. The goal isn't to prove the site works — it's to find where it doesn't, while finding it is still free.

6. Deployment: the beginning, not the end

Shipping moves the project from 'mine' to 'everyone's.' Automated builds, environment variables handled properly, a rollback path that takes one command. And then the most important step of all: watching real people use the thing, and letting their behavior write the next to-do list.

A good workflow doesn't slow you down to be careful. It speeds you up by removing the rework.