Costart ProjectsBook a call
Costart Projects · the site of sites

The sites I build, and the engine that builds them.

I'm Robert Costart. I design, code and run a fleet of AI-native products end to end — the ones sliding past alongside. The same engine now powers marketing teams: landing pages that spawn from your own customer data, served through Cloudflare, and tested until they convert.

5
live products
300k+
students on ClassMaster
20k+
homeowners on NearMeTrades
1M
users reached at Trainline
The offering

Landing pages that build themselves from what your customers ask.

Most marketing teams are stuck on an old WordPress, hand-building pages that go stale. The engine replaces it: an AI-coded site on GitHub and Cloudflare that spawns a page for every customer intent, tests it, and points your ads at it.

The old stack · WordPress, by hand
landing-page-final-v7.phptheme broke in the last updatebuilt by hand · 3 weekslast touched 2023plugin nobody ownssame page for everyoneno data, no testshosting + agency bill
The new stack · one page per intent
search queriessupport ticketsCRM & salesproduct usageAI codebaseGitHub→ Cloudflareyour writersboiler repair · Leedscancel my planA/B ✓pricing for 50 seatsis it GDPR compliant?vs. the competitorA/B ✓export to XeroGoogle Ads · Metaland on the matching page, automatically↑ conversion↓ wasted ad spend
  1. 01

    Your WordPress is a museum.

    Hand-built pages, a theme that breaks on update, nobody sure what the plugins do.

  2. 02

    Every customer question deserves its own page.

    The engine spawns one for each real intent in your search, support and sales data.

  3. 03

    Writers write. The engine lays out.

    Your team drafts like a blog; the codebase on GitHub and Cloudflare does the rest, in minutes.

  4. 04

    Pages that test themselves.

    A and B run on every page; the winner takes over without a meeting.

  5. 05

    Ads that land where they should.

    Campaigns point at the page that matches the intent, so spend stops leaking.

Cheaper than WordPress. Faster than an agency. Every page matched to a real customer intent.

Talk about your stack →
Now — a team of agents

Not just building the product. Building the team that ships it.

I run each product like a small studio of AI agents — they source the work, draft and build it, review each other, and hand off what's done. It's how one operator keeps a whole fleet moving, with the judgement calls staying human.

Sourcing2
Research

Scan the web for new leads worth writing to

RSAI agent
Research

Turn a lecture recording into a study pack

RSAI agent
In progress2
Build

Draft & qualify 40 first-touch contacts

BDAI agent
Build

Ship the quote-flow API endpoint

BDAI agent
In review2
QA

Regression pass across the checkout flow

QAAI agent
Research

First-touch outreach drafted & raised

RSAI agent
Shipped2
Design

Launch assets exported & handed off

DSAI agent
Design

Weekly summary & scale digest sent

DSAI agent
ResearcherBuilderDesignerQA
The work

A fleet of products, each shipped end to end.

Legal, education, home services, voice — each one designed, built and run end to end by one operator, with AI through the whole stack. Swipe through what they do for the people who use them.

Also built & earlier

The operator

I ship the thing, and I work out why it grows.

I've spent most of my career being the person who both ships the thing and works out why it's growing. The thread through all of it is a bias for the “how” over the “what”: get close to the work, model it honestly, and let the numbers decide. Today I run Costart Projects as a solo founder, building products with a small trusted bench of designers and developers — and doing it at a scale that used to need a team. This site is both halves of that: the products I'm building, and the operator behind them.

Trainline
European platform

Built the European booking site from a proof of concept into a platform serving a million users.

Expedia
Vrbo / HomeAway

Led the Vrbo/HomeAway platform migration, turning a projected loss into measurable gains.

Soldo
Growth model

Replaced a cost-per-acquisition model with a lifetime-value one, forecasting from 60 days of activity to ~95% accuracy.

AI at scale

What used to need a team, run by one operator.

AI woven through the whole stack — research, build, content, growth — isn't a feature I bolt on. It's how a single founder now operates at team scale, and the same edge I bring to growth teams.

01

Build end to end

Product, engineering and design in one head — from first commit to live, without hand-off friction.

02

Own the growth maths

Model acquisition and retention honestly, forecast lifetime value, and let the numbers pick the roadmap.

03

AI through the stack

Agents and automation compress research, content and ops from days into hours — across a whole fleet.

Contact

Building something, or scaling a growth team?

I take on a small number of build and advisory engagements. If AI at the core of your product and growth is where you're headed, let's talk.

Robert Costart