While Johnathon was using greatVibe to build the product, I was using it to build the company around it.
Not in theory. Not as a little internal test we dressed up afterwards. I mean the actual company. The brand, the pricing, the financial model, the website, the legal structure, the go-to-market plan. All the stuff that usually ends up spread across different tools, different documents, and different parts of your head.
Claude and Codex were both part of that. Claude helped me think, shape, challenge, write, and decide. Codex helped turn parts of that thinking into real files, working pages, and things I could actually open, test, and change. greatVibe was the place where it all sat, and that is the bit that mattered most.
It kept the work together.
Where I started
I've done enough startups to know how much work hides behind the word launch.
A proper GTM plan takes time. A financial model that can stand up to questions takes time. A brand you actually believe in takes time. Then there is the website, the legal work, the pricing, the investor material, and the constant switching between all of it.
That switching is where a lot of the cost sits. You finish a pricing session, then open the model and try to remember why you made a certain assumption. You move to the website and have to find the brand voice again. You go back to legal and try to remember what was decided yesterday. You are working all day, but a lot of the work is reloading your own brain.
So I opened greatVibe and started with Claude on the real checklist. Not a vague startup checklist with headings that sound useful but don't tell you what to do next. I wanted market-ready and investor-ready. What needed to exist, what depended on what, and what had to happen first.
The first version was solid. Then I pushed back. I added what I knew from doing this before. We cut things, reordered things, argued with the shape of it, and kept going until the plan felt like mine. Then we broke it into pre-release, investor-ready, and post-release work, with the prerequisites up front.
That sounds obvious, but when you're the only person in the room it rarely happens that cleanly.
The team I never had
The closest way I can describe it is this. It felt like having a room full of sharp people who were ready when I was ready. They didn't take over. They didn't decide the company for me. They waited for direction, and that mattered.
That is how the work felt with Claude and Codex inside greatVibe. Claude was where I did a lot of the thinking work. The questions, the structure, the challenge, the first versions, the rewrites, the decision points. Codex was where the work became more concrete. Files changed. Pages appeared. The website could run locally. Ideas stopped being trapped in documents.
I still had to lead it. That is the important bit. The AI didn't know what company I wanted to build. It didn't know what I would believe in, what I would reject, or what would feel wrong the moment I saw it written down. I brought that. I brought the judgement, the commercial experience, and the sense of what was worth doing.
But I wasn't carrying the whole thing alone.
Starting with who it was for
We started with ICP. Who are we actually building this for? Not a broad segment. Not a vague persona. A real person with a real problem who would feel the value straight away.
Claude was useful here because it didn't let me stay vague. I could put a version of the customer into the work, push on it, argue with it, and keep refining until it felt specific enough to make decisions from. Once that was clear, the rest had somewhere to point.
From there we worked on brand. Voice, positioning, the words we would use, and the words we wouldn't. I didn't write a brief and wait three weeks for something to come back slightly wrong. I worked it through in greatVibe, mostly with Claude, until it sounded like us. Not polished in a dead way. Us.
Then GTM. Who we go after first, what the buying moment feels like, what the page has to make clear, and where the friction is likely to be. Then pricing. What we charge for, what we don't, and how we make the model work without painting ourselves into a corner later.
That is where it started to feel like something different. It stopped feeling like a set of separate jobs and started to feel like a company.
When the work became real
The pricing work fed the model. The ICP shaped the assumptions. The brand shaped the website. The website raised questions that came back into the positioning. The legal work forced a few things to be clearer than they had been in my head.
This is where Codex really mattered as well. It wasn't just me writing strategy documents and hoping they would turn into something later. I could move from brand and copy into actual website files. I could make changes, run the site locally, look at the page, and react to the thing itself.
That changes the work. A page you can see is different from a paragraph in a doc. It fights back. It shows you where the words are too soft, where the section is too long, where the idea sounded good in strategy but doesn't land on the screen.
Claude helped with the thinking and the words. Codex helped with the build and the shape of the thing. greatVibe kept the context between them, so I wasn't starting again each time I changed mode.
The work stayed connected
That is the part I keep coming back to.
I could go from the financial model to the website, from the website to the shareholders agreement, from the legal structure back into pricing, and from pricing back into the GTM plan. In a normal week that would mean opening five different things and trying to remember where my head was yesterday.
Inside greatVibe, the thread was still there. Not because one model magically knew everything. Because the work had been built in a connected way from the start. Claude had the shape of the business work. Codex had the implementation context when we were working on files and pages. The environment held enough of the story that I didn't have to keep rebuilding the frame.
That freed up more mental space than I expected. I wasn't spending half my time reminding the system what Gravient was, who we were building for, or why we had made certain decisions. I could get back to the actual work faster.
What we built
We built the brand, the ICP, the positioning, the GTM plan, the pricing strategy, and a six-year financial model with real formulas and assumptions. We built the website against the brand guide we had already written. We drafted the legal structure and shareholder documents far enough that a lawyer could review from a proper starting point, rather than from a blank page.
Some of that work was Claude-heavy. Some of it was Codex-heavy. Most of it moved between the two. That is what actually happened, and I think it is important to say it that way.
None of it was final because an AI said it was final. That is not how I use AI. I was leading the work, reviewing it, changing it, rejecting things, and making the calls. The point is that I wasn't starting from scratch every time. The work moved.
Working with Johnathon
The other important part was that Johnathon and I weren't building in two separate worlds.
He was building the product. I was building the business. But the work could cross over. I could share pricing and GTM work with him. He could bring product decisions back into the business work. We could flag decisions for each other instead of letting open questions hide in documents.
The 24-hour summaries became useful too. Not as theatre. As a plain daily brief. What changed? What is open? What needs a decision?
That kept us moving without turning every question into a meeting. It also meant the business work and the product work stayed close enough to affect each other, which is how it should be at this stage.
What it felt like
I didn't want to stop working, which is not something I say often.
I would look up and it would be midnight. I thought it was nine. Not because I was grinding through work I didn't want to do, but because the work was moving and I could see it moving.
I know the other version of startup work. The scattered documents. The half-finished thinking. The thing you meant to do last Tuesday. The slow build-up of loose ends sitting just outside your line of sight. I know what it feels like to run hard without being sure you're actually building momentum.
This felt different. I felt bigger than I am. Not smarter. Not replaced. Bigger. More organised, more capable, and more able to hold the whole company in my head without carrying all of it alone.
For a while I had trouble sleeping. Not from stress. From excitement. I hadn't had that feeling at work for a long time.
Why it matters
We didn't build greatVibe and then look around for a proof story. The proof story was the work.
Johnathon used greatVibe to build the product. I used it to build the company around it. Claude and Codex were both part of that, but the bigger point is what happened when they were used inside one connected working environment.
That is why I trust it. Not because Claude gave me neat answers. Not because Codex changed files. Because greatVibe helped me direct real work across brand, pricing, GTM, finance, website, and legal without losing the thread.
That is the thing we are building. AI work with a human in command. Work that has context. Work that actually lands.