Playbook · Building with AI

Updated

How should you use AI in web development?

Use AI in web development for work you can check quickly: boilerplate, tests, refactors and first drafts of UI. Keep a senior engineer on the parts where a mistake is silent and costly: auth, data access, payments, SEO and page speed.

In Stack Overflow’s 2025 survey, 84% of developers use or plan to use AI tools. The top frustration, named by 66%, is AI answers that are “almost right, but not quite”, and 45.2% say debugging AI-generated code takes more time.
Source: Stack Overflow Developer Survey 2025, fielded 29 May to 23 Jun 2025

From my work

Turing.tn needed its pages found in search. We built a custom server-side rendering app in PHP, which took every Lighthouse check to 100, and then did programmatic SEO after a long round of keyword research.

On The Merchant Engine, getting Lighthouse scores to about 100 took a lot of trial and error. Now it ships Lighthouse tests out of the box, so the scores are checked on every change instead of guessed. If I had to name one speed fix that makes the biggest difference, it is asset size.

None of this shows up in a demo. A slow page still works, so it is easy to miss when the only check is whether the page loads.

Where AI helps in web development

  • Scaffolding components, forms and API routes from a clear spec.
  • Writing tests around code that already works.
  • Repetitive refactors: renames, type fixes, moving code between files.
  • Explaining an unfamiliar codebase before you change it. It has more patience for old code than I do.
  • One-off scripts, like a data migration, as long as someone reads them before they run.

Where it creates rework

The problems are in code that runs fine. AI writes what you asked for and often skips what nobody asked for. Few prompts end with “and check who is allowed to read this”.

  • Auth and data rules that let a logged-in user read more than they should.
  • Pages rendered only in the browser, which makes first load and SEO harder than they need to be.
  • Heavy dependencies and unoptimized images that slow every page.
  • Clickable divs instead of buttons, missing labels, and other accessibility gaps.
  • The same logic written three times in three places, each copy slowly forming its own opinions.

How reliable the numbers are

The Stack Overflow survey had 49,009 responses from 177 countries, recruited mostly through Stack Overflow’s own channels. It tells you what developers say about AI, which is useful for the mood and much less useful for speed. Nobody in it was timed.

METR measured instead of asking: in early 2025, experienced developers took 19% longer with AI while believing they were faster, and its 2026 follow-up called the new data “an unreliable signal”. The guide on AI-assisted software development goes through those studies in detail. For web work I take one thing from it: feeling faster and being faster are different things, so I measure the page. Lighthouse does not score how productive I felt.

How I work with it

I use Claude Code in my work. Once code ships, it is my responsibility, whoever typed it. So I rely on checks that don’t care who wrote the code. On The Merchant Engine, Lighthouse CI runs on every change, next to Jest, Vitest and Playwright tests.

A small tool I built this way: Claude reads Reddit through your own browser

When a website should not use AI at all

  • When a search box, a filter or a plain rule does the job. They are cheaper and always give the same answer. A filter has never invented a product you don’t sell.
  • When a wrong answer is costly and nobody checks it, like prices or contract terms shown to customers.
  • When you can’t carry a running cost for every visitor who uses the feature.

What it costs to build with me

I sell this, so read it with that in mind. I build MVPs on the Full plan, $4,995 a month, starting with one user type and one core flow. The build is split into tasks, and the number of months depends on scope, agreed on the kickoff call.

If you only need one AI feature added to an app you already have, it is done on a monthly plan, one task at a time: Limited at $1,995 a month or Full at $4,995. Both are a flat monthly price, pause or cancel anytime. On the Full plan it usually ships within the first month.

Read next

Sources

Start

Find out what stands between you and launch.

Start the 14-day Trial: your problems ranked in the first 5 business days, then fixes for what fits in the 14 days. Not ready yet? Ask for a free teardown.

Start the 14-day Trial · $995

Secure checkout via Stripe · Refund policy

Founding rate: the first 3 clients get 20% off, in return for a named testimonial and a case study. On a monthly plan, the 20% applies to the first month.

Prefer to talk first? Book 30 min or email badii@badii.studio

Free teardown

48h · max 5 / week

A 5 to 10 min recorded review of your live app's public surface. No access needed, just the URL.

I use these details only to reply and record your teardown, and delete them after 12 months. Privacy