Built by someone who runs real businesses, not by a company that has never sold anything
JFKAISLAY exists because I needed these tools for my own businesses and they did not exist yet.
I am JFK Løvlien. Norwegian, based in Brazil. I build AI systems, I build iPhone apps and games, and I write thriller novels. Everything on this site started as something I needed myself, and then turned out to be worth selling.
It started as a catalogue problem. In early 2026 AI assistants began recommending products directly to shoppers, and I realised the product data in my own stores had been written for a person skimming a page, not for a model comparing four options. I spent months rewriting over 25,000 products across my own stores, tested it against real traffic, and proved it works. That process is now the Sales Scanner.
Then I built Nexus, which is the AI sitting in the corner of this page. Not a chatbot. A custom system that knows a business inside out, qualifies enquiries, writes content and watches the market. I built the first one for my own author platform, where it runs under the name The Magus, before I built one for anybody else.
What became obvious fairly quickly is that the catalogue was never the interesting part. Nearly every business is losing money in one of four places: enquiries answered too slowly, the same question answered for the thousandth time, work produced by hand that could be produced to a pattern, and knowledge that exists in exactly one head. A property agency has those. So does a plumber, and so does a corporate department with nine layers of sign-off. The build is different every time. The problem is not.
Most business owners do not have the technical skills to build this, and they should not have to. That is my job.
Everything I sell, I use. The catalogue process was tested on my own stores before I sold a single credit. The Nexus architecture ran on my own platform before I built it for a client. The Scanner scored my own products before it scored yours.
There is also Jane SEO, a multi-tenant SEO app of mine live on the Shopify App Store. Multi-tenant means it runs many merchants' stores at once, with each one's data walled off from the others, which is a harder problem than it sounds and good practice for building isolated systems for clients. Shipping through app review is also a useful thing to have done before telling anyone else how to ship software.
Your Nexus runs on your infrastructure. Your code, your data, your API keys. Cancel and it stays yours. No lock-in.
Every Nexus is built from scratch. Every rewrite is unique. No templates, no auto-generated filler, nothing recycled between clients.
IRONCLAD has ten layers, built from the OWASP LLM Top 10 and real attack data. A Nexus cannot spend money, delete data or execute code, so there is nothing to hijack.
Nothing is called fixed because it looks fixed. If a thing cannot be measured on the real path, it is not proven.
One more thing, and it costs me money to say it: a full build is not the right answer for every business that asks for one. If an hour on a call is the honest first step, I will tell you that. If the honest answer is that you do not need me yet, I will tell you that too. I would rather lose the sale than take four thousand dollars off someone who needed a different conversation.
Anthropic's Claude API for the intelligence. Railway for hosting. GitHub for code. Node.js throughout. ElevenLabs and Kokoro for voice. Unity and Xcode for the game. Everything modern, fast and production-grade. No legacy stacks, no bloated frameworks, no complexity for its own sake.
I chose Claude over the other models for one reason: it follows complex instructions better than anything else I have tested. When you build an AI that speaks for someone's business, instruction-following is the whole job. A model that drifts off-brand or quietly ignores a guardrail is a liability, not a feature.
I am building a stealth game for iPhone, set in 1786, on my own. Unity 6.5 on URP and Metal. Ten scenes, a development build of 2.8 GB that takes four to six minutes to compile. One shell world with districts that stream in and out so a phone can actually hold it: a coast camp, a village of seventeen houses, a forge cut into a mound, four tunnels under the island. Social stealth, so villagers walk in groups and you hide by blending into one. A mission table driven by world flags. Climbable ladders whose rung positions are baked at build time, because reading mesh data at runtime fails on the device. Eighty-seven commits in eleven days.
The method is the part worth stealing. Claude Code executes written briefs inside the project with Unity closed, and every test runs on the player's real path — synthetic touch through the input system, thumb on the stick, real scenes loaded, game time. No direct method calls and no teleported test bots. That rule exists because the first round of tests invoked buttons directly, passed in the harness, and then failed on the phone.
Last week a colour flash in the tunnels survived three days of entirely plausible fixes. So I stopped fixing and measured instead: six hundred frames walked, the mean colour printed for every one, twenty-six flagged. The measurement named the cause — the player's own skin shader exceeding Metal's sixteen-sampler limit, and only ever under lantern light, which is why daylight never showed it. Three days of guessing against twenty minutes of measuring.
That is the same discipline that goes into a Nexus. Guessing is cheap, and it is why most AI projects quietly fail six weeks after the demo.
I put the builds on YouTube as I go. When clients would rather learn to work this way themselves than buy the result from me, I teach that too.
I write the Feeder Series, Luxe Noir thrillers published under my own name. Four are out and the fifth, Nirvana Nights, publishes on 1 November 2026. They are on Amazon and at jfklovlien.com, which runs its own AI assistant, its own audiobook pipeline and its own content engine.
Different outputs, same method. JFKAISLAY is where the AI work lives.
