Why I build what I build

Four things, one pattern

I build four things at once.

Oase Creative has run since 2020 as the e-commerce branding agency: 200+ brands helped, 200+ reviews. Oase is the software platform for design agencies, live since 2026, built with co-founder Lam Hoang. descout is the discovery layer for blockchain ecosystems, also with Lam, running since 2025. Milo is the AI system that keeps this whole stack working 24/7 on a Mac Mini in my office.

Four projects that, on the surface, share nothing. An agency, a SaaS, an onchain product, an agent system. Different customers, different markets, different disciplines.

One pattern. Each of the four is the tool that should have existed when I needed it myself.

I am writing this down because it helps me decide, every month, which next thing I do build and which next thing I do not. This site is not a portfolio and not a sales funnel. It is my reference. The sharper the why is on paper, the less I second-guess what I put on the table later.

First the friction, then the build

I did not start with a vision. I started with an irritation.

At 23 I ran my first e-commerce brand, Enjoy Inc.. Garment steamers sold on a large Dutch marketplace. Around €125k in revenue in a year, at roughly 40% margin. That went well until it stopped, when the partnership with my business partner broke. What stayed was not the brand. What stayed was what I had learned through that brand: how a listing converts, why packaging has to say something, how to stand out among two hundred lookalikes on a marketplace.

And a network full of founders with the same problem. Good product, weak brand. No visual identity. Listings worth a hundred euros that would never compound into real return. I started helping them for free because I could not let it go. When it stopped being informal, I registered the company. That was 2020.

That is the prototype of the pattern. I do not build from a blank sheet. I build from friction I have already felt for a while, often years. What begins as free help to other founders becomes an agency. What begins as eight disconnected tools on my own laptop becomes a SaaS. What begins as two hours of morning tab-switching becomes an agent system.

The order is almost always the same. First I run something myself, long enough to know where it hurts. Only then do I start on a product that solves the hurt structurally. Not the other way around.

Oase Creative: six years in the seller's seat

Oase Creative is not an agency that delivers e-commerce branding because the market was there. It is an agency that delivers e-commerce branding because I sat in the seller's seat first.

That difference feels soft and works hard. Design a listing from theory and you guess which information has to come first. Stare at two hundred sellers in the same category and then at your own sales numbers, and you stop guessing. You know which thumbnail gets clicked and which does not. You know why packaging is more than a box. You know what the first three seconds have to do.

Six years, 200+ projects, 200+ reviews on Trustpilot. Brands that grew from first product to multi-million revenue. Half of that work is design. The other half is understanding what a marketplace actually measures. Marketplaces are not a content game. They are a conversion game. Branding is not a visual layer on top of the product, it is the engine under the ranking. You only really understand that if you have sold there yourself.

What I had not foreseen in 2020: the agency would grow to a place where eight different tools were needed to move a job from quote to delivery. Slack. ClickUp. Google Drive. WhatsApp. Email. A CRM nobody updated. Briefing forms in Typeform. Time tracking in a separate app. Every new client had to be created in three systems. Every brief forwarded by hand.

That became the next friction. So it became the next build.

Oase software: the thing that should have existed

Oase is the tool I wished existed when I created my tenth client folder in Google Drive four years ago.

A platform for design agencies. Client portal, project management, quotes, briefing forms, design review, a built-in AI agent. Instead of eight separate subscriptions that do not talk to each other. Instead of admin as a tax on top of the real work.

What matters here is not that it exists. Similar tools exist. What matters is that I build it from the work I do every day. Not from a research report. Not from a PM who came out of a SaaS company where the distance between "customer says something" and "it is live" is four quarters. At Oase that distance is a week.

A Squarespace designer in Utah showed me that a few weeks ago. Dylan Bindrup got a 47-minute walkthrough of the dashboard with no briefing beforehand. Halfway through he said: "I'm so annoyed about having 50 tabs open." Four minutes later: "you guys thought of everything." He sat on the other side of the ocean, in a different discipline, and recognized every screen without explanation. That recognition is not luck. That is what happens when you build from your own pain that turns out to be other people's pain.

Building for the work you do every day is a different game from building from market analysis. It asks less marketing at the front and more sharpness at the back. You know what the user will say before they say it, because you are the user.

Milo: the two hours that ate my mornings

Early in 2026 I was not drowning in work. I was drowning in context.

Every morning the same routine. Laptop open. Email. WhatsApp. Slack. ClickUp. LinkedIn notifications. Check the blog. Check the webshop. Check the freelancer. Two hours of input management before anything resembling real work began. Oase had started to build brands. What came out felt more like maintaining spreadsheets.

Milo is what I built after that. A multi-agent AI system that runs the entire agency back-end. Five agents, each with their own job. Research at six in the morning, before I am awake. Content drafted by eight. Inbox triaged and labeled. Publication scheduled and executed. Visual content generated where needed.

It runs on Claude Opus 4.7, on a Mac Mini in my office, 24/7. Running cost: around $100 per month in subscription. It replaces work that would conservatively cost about €5,000 per year in disconnected tools and execution.

The numbers are not the point. The point is that I now start my morning on the treadmill with a coffee and a briefing in my inbox, instead of with six tabs open. Screen time down in the morning. Room for real work up. Not because AI replaces me. Because the system replaces the tab-management that was holding me back.

This is the same friction-to-build move as Oase software. The problem is just smaller and more personal. Not "agency owners suffocate in tools" but "Louie loses two hours a day." Solving it went exactly the same way. First sit in the pain long enough to understand it. Then build a system that removes it structurally.

descout: same pattern, different world

descout is the outsider in this list. An onchain product, not an agency-or-bureau thing. Set up with Lam in 2025, live on Abstract since 2026.

The friction here is not mine, it is Lam's. He has spent years working on top of blockchain ecosystems and kept seeing the same pattern. Foundations fund projects. Projects ship. And the chain stays invisible. Not on X, where something happens. But on Google, in ChatGPT, in Claude, in Perplexity. There the ecosystem does not exist, because there is no structured layer for search engines to read.

descout is the discovery layer that solves that. A directory of every project on the chain. A live activity stream. A data layer across DEX, NFT, social, on-chain. A surface that search engines index and AI agents can query. Not a content shop, not a marketing agency. An infrastructure layer that discovery itself compounds on.

Live on Abstract after a few months of runtime: 534 breaking news articles, 121 long-form pieces, 5,000 monthly visitors, growing 20% month over month. Two productized tiers, starting at $12k and $30k per month. A separate blog goes deeper.

Why does this fit in my list? Because it is exactly the same move, in a different discipline. Nobody fixes this project-by-project. It belongs to the chain, the same way branding does not belong to each listing in isolation but to the seller, and the same way tooling does not belong to each project in isolation but to the agency. Pick the structural problem at the structural level. Build the layer nobody else builds.

The pattern behind the four

If I crawl under all four and pull out the shared layer, I land on four tests that every new decision about what to build has to pass.

One. The problem has to be one I have felt myself, long enough to know where it really hurts. Not from a report, not from an interview. Own friction, own frustration, own lost hours.

Two. The solution has to be structural. Not "better SEO tip" but "a layer that makes SEO not a separate battle anymore." Not "better project template" but "a platform that templates come out of." What can be solved small does not need me. What only the structural level can solve is where my work belongs.

Three. I have to want to be the first user. With Oase software I am my own customer. With Milo I am the only customer. With Oase Creative I was the seller first, the service provider second. With descout that is Lam. If I cannot be the first user, I do not know if it is worth building.

Four. The system has to keep running when I am not there. People can take ten days off. Systems have to keep going. Milo is the sharpest example, but Oase software, descout, and the playbook behind Oase Creative follow the same rule. Otherwise you build a job, not a business.

What these four do together is draw a line between "yes, I will build that" and "no, that is not mine." A hundred ideas come past per month. Ninety-nine fall on one of these four points.

What I will probably do in 2027

Looking forward always reads as overconfident on the page, but I want to lock this in for myself. Not as a promise. As a direction.

Oase Creative stays the anchor. The agency that delivers the work that carries my name and feeds the other builds with real projects, real customers, real data. Without this anchor, the other three are loose sand.

Oase software grows to 1,000 users in 2026, and from there toward the best design-agency platform in existence. Lam runs the software side with a set of agents. I stay first user, support desk, and voice in the product. The roadmap runs: phased payments, capacity planning, an app store for extensions, marketing intelligence on top of project data. None of it comes from a vision document. All of it comes from a list of frustrations I keep while running my own agency on top of it.

descout scales with the chains it runs on. Two tiers, a 90-day deployment, a compounding surface. The question for 2027 is not whether this works. Abstract has shown that. The question is how fast other foundations understand that discovery belongs to the chain, not to each project in isolation.

Milo gets smarter. Not in the AI sense, in the operational sense. Better handoffs between agents. Stricter validation on output. More context kept between sessions. Next agents on the roadmap: an outreach agent that maintains LinkedIn networks autonomously, a CFO agent that tracks revenue and margin live, brand monitoring for what is said about Oase on Reddit and X.

And if a fifth project shows up, it runs through the same four filters. Own friction. Structural solution. First user. Runs without me. Otherwise no.

Why I am writing this down

Two reasons.

The first is practical. Milo helps me every day. But Milo works sharper when the reference he tests decisions against is also on paper. What lived in my head until today now lives in a post he can read, cite, and use to check every new piece of content or decision against. This site is, literally, source of truth for what I build and why.

The second is for me. As a founder it is tempting to drift. A lead comes by that fits something I have not done before. An idea seems to have just enough money in it. A new discipline looks hot. The pattern above, on paper, reminds me that drift costs money and costs focus.

Building four things at once is already a lot. It works because they feed each other. Oase Creative feeds Oase software with real projects. Oase software feeds Milo with data. Milo feeds Oase Creative with content and operation. descout sits next to them, in a different discipline, but follows the same pattern. Four loops that reinforce each other.

A fifth loop has to add that reinforcement too. Otherwise it is not a fifth loop, it is a distraction.

This is what I build. This is why. And this is where Milo and I fall back to when the next question is what to take on next.

Frequently asked questions