A London product studioEst. 2026Three brands
Not a chatbot bolted onto a form. In everything we build, the model does the heavy lifting — reading, drafting, matching, designing — and the person keeps the taste. Three brands so far, each built for a moment people actually care about: planning a wedding, putting a small business online, and finding time to read.
Atlas, the AI that plans your wedding with you — from photoreal concept boards to the seating plan.
Live · early accessPaste your Instagram handle. Get a website that looks designed for your business — because it was.
Launching soonThe books you already own, as a feed you swipe through — with an AI that never reads ahead of you.
Private betaWhat “AI at the heart” means in practice
A wedding planner, a website builder and a reading app have almost nothing in common on the surface. Underneath, they're built on the same four convictions — and the same models.
Every product draws the line in the same place: AI reads, drafts, ranks and designs; you approve, edit and send. Nothing goes out in your name that you haven't seen.
In practice
Atlas drafts your supplier enquiries in your voice — you decide who receives them. OnShow writes every line of your site — you can rewrite any of them before it goes live.
Generic output is the failure mode of AI products. Most of our effort goes into the opposite: results that could only be yours, for your business, your day, your book.
In practice
OnShow reads the same Instagram profile differently for a barber and a florist — seventeen niches, seven site families, each with its own typography. Atlas regenerates a bouquet until it's the right bouquet.
No paid placements, no sponsored rankings, no dark patterns, no quiet data grabs. If the product can't earn the recommendation, it doesn't get to make it.
In practice
On makemywedding.ai, couples pay and vendors never do — rankings cannot be bought. Riffle has no accounts and no backend; your books and your reading never leave your device.
AI lets a tiny team ship like a big one — as long as the discipline is bigger than the team. We document as we build and we don't trust a check we haven't watched fail.
In practice
Every change to how a product behaves ships with its documentation in the same commit. Riffle reached beta with 270 unit tests and 12 end-to-end journeys behind it.
Meet Atlas, a wedding co-pilot for couples planning their own day. Describe the wedding in your head and Atlas turns it into photoreal concept boards — venue, florals, attire, tablescape, cake — then finds the real UK venues and suppliers who can make it, and gives you the tools to run the day itself.
Paste your Instagram handle. A few minutes later, a website that looks like it was designed for your business — because it was. OnShow is for the makeup artists, barbers, trainers, bakers and boutiques whose whole business lives on a grid, and who deserve better than a link in bio.
Seventeen niches map to seven site families, each with its own typography, palette and signature sections — so a barber's site never looks like a florist's, and neither looks like a template.
Riffle turns the books you already own into a swipeable feed — sentence-snapped slices of about forty-six words, one thumb-flick at a time. It's the reading app for people whose attention has been trained by Instagram, and who'd rather point that habit at Dickens.
Alice was beginning to get very tired of sitting by her sister on the bank, and of having nothing to do: once or twice she had peeped into the book her sister was reading, but it had no pictures or conversations in it, “and what is the use of a book,” thought Alice, “without pictures or conversations?”
How we build
We build the way we think products should work. The same rule that governs every brand — the model does the work, the person keeps the decisions — governs the studio itself.
Direct
What gets built, in what order, and what “done” means. Every schema change, every design, every launch is read and approved by a human before it exists.
FounderBuild
Detailed directives, then implementation, then a plain report of what was found along the way — including the things that didn't go to plan.
Claude · Claude CodeVerify
Unit tests, end-to-end journeys and rendering checks on every branded surface — desktop and a 390px phone — before anything reaches production.
Automated · then a human readStanding rule
Every check fails closed. An unadjudicated, errored or half-run verification counts as a failure, never a pass. A check that looks green while doing nothing is worse than no check at all — so we treat it as a defect, and every change to how a product behaves ships with its documentation in the same commit.
The studio
Make My AI was incorporated in London in 2026 to hold the products Daniel O’Neill had been building on nights and weekends: a wedding planner that started with planning his own, a website builder for the kind of small business that lives on Instagram, and a reading app he wanted for himself.
Each brand keeps its own name, look and voice. What they share is the engine underneath and the four convictions above — and a founder who'd rather ship one honest product a season than ten thin ones.
Performance marketer by trade, product builder by conviction. London.