When The Game Felt Empty
I had built plenty of parts, but the game still didn’t feel good. That uncomfortable moment changed what I worked on next.
The blog
Browse the full archive. Search by keyword or narrow to a single project.
100 posts
I had built plenty of parts, but the game still didn’t feel good. That uncomfortable moment changed what I worked on next.
I wanted my AI mystery game to remember every detail. Then the browser basically told me: absolutely not.
My healthcare app sounded impressive until I saw how much I was trying to build. The turning point was admitting V1 had to shrink.
I thought live supplier pricing would make the bid tool feel magical. Then I learned the real world was messier than the idea.
I started with an ad tool in mind. Then the project quietly told me it wanted to be something broader, and harder.
I started BidPilot with a very big idea. Then I had to admit the first version could not be the whole dream.
I thought the game was broken before it had really begun. The worst part was not the bug, but not knowing what kind of bug it was.
I thought the hard part would be the idea. Then the town generator broke, and I met the less charming side of building with AI.
I thought choosing the first version would make the project feel smaller. Instead, I ran straight into the hard part.
I thought the hard part would be building screens. It turned out the real problem was the data I assumed would magically exist.
I thought the hard part would be ideas. It turned out to be the moment the app had to pull in a real website and make sense of it.
I started with a big construction software idea. The first wall was realizing how many promises were hiding inside one sentence.
I asked the AI to build a whole creature-evolution game. What came back was both impressive and slightly bewildering.
I asked the AI to help me build an entire mystery game. What came back was exciting, imperfect, and a little dangerous.
I asked the AI to help me start a national physician credentialing platform. It was exciting, humbling, and a little ridiculous.
I asked the AI to build a construction bidding platform. It gave me more than I expected, and less than I hoped.
I asked the AI for something wildly ambitious. What came back was impressive, but it also taught me how big I had made the mountain.
I started BidPilot with an oversized idea and one very hopeful prompt. Here’s what I asked the AI to build first.
I wanted a creature game that kept surprising me. Starting it with AI felt exciting, oversized, and slightly absurd.
I wanted a mystery game that could surprise me. Then I realized I was asking AI to build a whole town full of liars.
I started MyUPP with a simple question: could AI help shrink a slow healthcare process that can take months?
I began with a big hope: maybe AI could make construction bidding less painful. It did help, but not in the way I first imagined.
I began MyBrandPilot because I wanted the kind of brand and strategy help that usually feels locked inside a consulting firm.
I started MyBidPilot because construction quoting feels like too much copying, chasing, and hoping nothing gets missed.
No incubator. No accelerator. No slack channel of peers. Here's what I cobbled together instead — and honestly, it was enough.
Every night around midnight the same thought shows up: this is dumb, nobody cares, quit. It's not information. It's just tired.
You have an incredible AI research analyst on call. Use it before you spend three months building the wrong thing.
AI-assisted building is cheap compared to hiring a team. It is not free. Budget like an adult.
AI is the most patient teacher you'll ever meet. But it doesn't read minds. Learn to ask, and doors open.
Most of the people around me don't understand what I'm doing, and probably never will. I had to stop needing them to.
If I rebuilt Beads of Hope T1D from scratch, I would still build it. I just wouldn’t ask the AI to build everything at once.
Fifty years old, no coding background, typing into a chat box like it owes me a business. Yeah, it feels ridiculous. Do it anyway.
The AI will occasionally look you in the eye and lie with total conviction. Not to hurt you — it just doesn't know it's wrong. Here's how to catch it.
You can't out-code a 22-year-old engineer. But you can absolutely out-taste them. Here's why the years you've spent noticing things are worth more than you think.
If you’re 50+ and starting with AI, here’s the honest advice I’d give after turning a trivia idea into something much bigger.
If you're not occasionally wondering what on earth you're doing, you're probably not building anything real yet.
The quiet part of starting something new: there is no team, no cofounder in the next chair, no boss to escalate to. Just you and a blinking cursor.
I started with no coding background and a lot of doubt. I still have doubt, but now it is useful doubt — the kind that asks better questions.
The site is nearly ready. The code is wired. Now the scary part is letting real people touch it.
If you're 50+ and starting with AI, I’d tell you this first: don’t ask it to build the whole dream in one breath.
If I began Hat Trick again, I’d resist the exciting pivot and build a much deeper trivia game before adding fantasy.
MyTradePilot is closer than I ever expected, but I am not going to pretend everything is perfect. The app is strong in many places, and I am still cautious about automated trade execution.
The project you launch will not be the project you started. Adapt or die is not a threat — it's the entire game.
The app works on my screen. Launching means finding out what happens when real people click the buttons.
There is a version of the tech internet that will convince you that what you are doing is not real. Do not visit that internet. Protect your beginner's mind.
A tiny order confirmation landed in my inbox, and somehow that little email made the whole project feel more real.
You don't need to be an expert. You just need to not do the obviously dumb things. Here's the shortlist of security stuff I now do without thinking.
The breakthrough was not one big fireworks moment. It was a series of smaller moments where the app stopped feeling like a demo and started acting like a system.
It was just one word in one headline. But when “football” finally popped in magenta, the whole thing felt a little more alive.
A security scan found nine issues in my little shop, and I had to decide whether to trust fixes I only half understood.
The first money I spent was $25 a month for soccer data. It was small enough to click, but big enough to make the whole thing feel real.
Wins, fails, and a whole lot of learning. Trust the product; trust the process. Launch day is day one, not day done.
Nothing will humble you faster than an app telling you that you made $124,000 when you know good and well something is wrong.
AI is astonishingly smart and confidently wrong at exactly the same time. Verify facts. Test features. Always.
You don't need to be an expert to build. You need to know one problem well enough to smell when a solution is fake.
The term sounds ridiculous. The idea behind it is quietly the biggest shift in who gets to build software since the internet.
At first the app had a name that sounded like it belonged to somebody else. The rebrand to MyTradePilot was more than cosmetic — it made the project feel like mine.
Stripe was free to add, but putting in my bank details made the whole thing feel suddenly, uncomfortably real.
I started with no idea what I was doing, one honest conversation with ChatGPT, and one big prompt to Lovable. Turns out that was enough.
A security report landed in front of me, full of words I barely understood. I had to decide whether to trust the fixes anyway.
I spent part of a Sunday nudging a blue ribbon and a tiny heart around. Almost nobody would notice. I noticed.
I spent far too long changing one word to magenta. Nobody needed it. Somehow, I did.
The first real wall was not a dramatic crash. It was quieter than that: the scanner ran, looked important, and returned 0 candidates.
The shop existed, but the product photos vanished. Then clicking the empty space bounced me back to the top.
I thought account creation was broken. It turned out the app was working, but it forgot to tell me.
The most terrifying and useful thing you can do is show your half-built thing to a real human, out loud, before it's ready. Here's how to survive it.
Social media shows you people shipping. Nobody posts the six-hour bug. Fall in love with the process anyway.
My first instruction to Lovable was basically: build me a serious AI stock trading platform. Looking back, I laugh a little, because I had no idea how much I was asking for.
If you're 50 and thinking about building an app with AI, my advice is simple: go small, stay honest, and check the numbers.
Building Beads of Hope taught me that AI can move fast, but the courage part still has to come from me.
I started with a soccer trivia idea and accidentally built something much bigger. The hardest lesson was learning not to believe my own headline too soon.
For every feature that ships, there's a graveyard of the ones that didn't. You are allowed — encouraged — to throw things away.
The site is live-ready, but I’m learning that building the app and opening the business are not the same thing.
A crash is annoying. A wrong number in a trading app feels heavier, because somebody could trust it with real money.
The trivia works. The fantasy side works. But the hardest part now is earning the big claim I wanted to make.
The first storefront appeared in about 20 minutes. I was thrilled, confused, and suddenly aware this might become something real.
The scanner found real candidates, and the profit and loss math survived my spreadsheet check. Not fireworks. Just a small exhale.
The highs are higher than you think and the lows are lower. Bring snacks.
Moving from a free experiment to a monthly bill changed how I saw MyTradePilot. Suddenly it was not just tinkering anymore.
A button failed, I felt ridiculous, and then one redesign made the whole soccer project suddenly feel like a real product.
The security scan flagged scary phrases I didn’t understand. I had to decide whether to trust the AI with fixes I couldn’t verify myself.
Some bugs are bugs. Some bugs are the building telling you the foundation is wrong. Here's how I've learned to tell the difference — usually one fix too late.
The site looked like a real store before I felt ready to run one. That was thrilling, and honestly, a little terrifying.
I wanted to call it the greatest soccer trivia game on the planet. Then I had to ask whether I had actually built that.
I started by asking the AI for the whole store, the whole community, the whole dream. Then reality arrived in the details.
I started with a soccer trivia game. A few hours later, one question turned it into something much larger.
The storefront appeared fast. Then the images disappeared, the buttons doubled, and I realized I was in deeper than I thought.
In vibe coding, the prompt is the product. A vague prompt gets you a vague app. Here's how I stopped writing wishes and started writing instructions.
I spent too long on a tiny design decision, rejecting logos and changing my mind before finally uploading my own image.
My first real wall was a signup button that seemed broken. It turned out the app was working better than I was.
I asked the AI to build the whole Beads of Hope site in one giant request. What came back was thrilling, and a little terrifying.
The scanner kept returning 0 candidates. The fix involved words I didn’t understand, and that was the hard part.
I started by asking for a blueprint, not an app. Within hours, my soccer trivia idea had turned into something much bigger.
You are in control — but only of the map. The moment you try to grab the wheel, everything gets weird.
I started with one giant prompt and a very personal reason. Then the AI built a shop faster than I could believe.
If I started MyTradePilot again, I’d choose the name before building and stop treating the logo like the product.
Waiting for a big block of free time to build is a great way to never build. Use the gaps.
I began with one question about a soccer trivia game. A few hours later, I had accidentally started building something much bigger.
I am 50 years old, not a coder, and I decided to build an AI stock trading app. That sentence still sounds ridiculous to me, but it is exactly where this started.
The urge to talk about your project is not the same as building it. Save the announcement for something worth announcing.