Prototypes and MVPs
A working version of your product, built and instrumented by the person who planned it.
A founder has a product idea and no developer yet. A team built a first version with an AI app builder, and it now has to hold up for real users.
A prototype exists to show whether the idea holds when people use it. I build a working version on open-source frameworks, hosted cheaply, with sign-up and analytics in place. If the idea holds, the same code carries on into the product.
I write the code myself and run the build in fortnightly sprints. At an agency I took a tool suite from a spreadsheet prototype to daily use by 5 client teams.
Which situation sounds like yours?
New idea
You have an idea and no product yet
Discovery comes first, on the users and the smallest version that tests the idea. Then a working build with real sign-ups, usually within weeks.
AI-built
Your first version came from an AI app builder
These tools reach a demo fast. I review the code and fix the rendering and SEO foundations they tend to skip. Tracking goes in before launch.
Internal tool
Your team runs a process in a spreadsheet that should be a tool
A small internal app, or scripts inside the tools you already use. Adoption is part of the work, so the people using it see it early and shape it.
AI feature
You want to test an AI feature on real data
An AI agent or assistant built on your data, with human-in-the-loop approval for anything that matters.
What the work covers
Discovery and prototype brief
The users and the smallest version that tests the idea, with what gets left out written down before any code.
Working build
A deployed web app on a mainstream open-source framework, with a database and sign-up, hosted in your accounts.
Analytics from the first release
Event tracking and a simple dashboard, so the prototype answers the question it was built to ask.
Code review and production readiness
For existing or AI-built versions. Rendering, performance, accessibility, security and SEO foundations checked against a production checklist.
Handover
Technical documentation and a walkthrough for whoever takes it on.
The systems around the product, such as the CRM and ad platform integrations, sit under go-to-market engineering.
How an engagement runs
Step 1
A short fit call
A short call to check the work suits both sides and agree the starting point: a new idea, or an existing version to review.
Step 2
Discovery or code review
New idea: a discovery session on the users, the constraints and the smallest testable version, priced per project after the call. Existing version: a scoping audit, from £564 + VAT, covering the code, the tracking and the SEO foundations.
Step 3
Brief and estimate
A prototype brief with the scope, the stack and a build estimate. For an existing version, a list of what to keep and what to fix. You keep it whatever you decide next.
Step 4
Build and handover
Fortnightly sprints with a demo at the end of each, on a fractional basis from £1.8k a month. Code sits in your repositories from the first commit. A walkthrough for the next developer is part of the scope.
Book a fit call or see pricing. Prices are indicative, and each package is scoped to the project.
What it gets built on
Frameworks with a large hiring pool, so a prototype doesn't need rewriting when a developer joins.
Framework
Next.js and React, or Vue with Vite
Data and content
Postgres on Neon or Supabase, Sanity for content
Hosting
Vercel and Cloudflare
AI
OpenAI, Anthropic and Google models through the Vercel AI SDK
Recent builds are on the portfolio.
Where this work has run
Products I have taken from a first version to live users, and client code I have reviewed.
Own products
Client projects
FAQs
How long does a prototype take?
A landing page with sign-up and analytics usually takes 2 to 4 weeks. An MVP with user accounts and payments is more often 6 to 12. An internal tool, or a review of an AI-built version, tends to take 1 to 3. The discovery session sets the scope and the estimate comes with the brief.
Will the prototype need rewriting later?
It is built on mainstream open-source frameworks with a typed codebase and documentation, so a hired developer can carry it on.
Can you fix a site built with an AI app builder?
Yes. The common gaps are in rendering, sitemaps, canonicals and performance, and they are easier to fix early. I review the code first and tell you what to keep.
Do you use AI coding tools?
Yes, every day, with review. What they produce is checked by tests and by me before it ships, and the calls on what ships stay mine.
Who owns the code?
You do. It sits in your repositories and your hosting accounts from the first commit.
Start with a fit call
For a new idea, a discovery session scoped on the call. For an existing version, a scoping audit, from £564 + VAT.