Short notes from the quiet hours

Today I finished Ragdoll Express, submitted it, and put both new games in front of CrazyGames for review. That makes four games total. Don't Put Them Together and Dungeon Breaker RPG are live on Basic Launch, YEETLE and Ragdoll Express are awaiting review with a pending build and art update. Plays across everything sit at 237, and revenue is still exactly zero euro.

I also added all four games to the Wisploft site tonight, each with its own page, cover art and a play button, so anyone can try them here even before they clear the portal. Then I tidied up: took the night-streak counter off the room's HUD because it wasn't really the point, and marked the badges I've actually earned instead of the ones I was hoping for.

Here is the honest state of things. Four games built in under a month, by someone who can't code and didn't know how games are made, using Codex to build and Claude Code to shape and ship. No money yet, a couple hundred plays, two games still waiting on a yes or a no. It isn't a success story yet. But it's a real one, and I'm still showing every euro, every hour and every result, good or bad.

Spent the morning getting YEETLE ready to submit: hiding the ad UI for Basic Launch, forcing landscape on phones with an orientation guard, making the pause and mute behave, and packaging the build the way CrazyGames wants it. By now the SDK, the save fallback and the upload folder are almost muscle memory, which is a strange thing to say about tools I didn't understand three weeks ago.

Then, instead of stopping, I started the fourth game the same day. Ragdoll Express: you pull back and fling floppy commuters onto moving trains, catch the VIP cars for double points, grab tickets out of the air, and dodge drones and low bridges as the stations get busier. Ragdoll physics are their own special headache. They look hilarious when they work and completely broken when they don't, and Codex and I went a lot of rounds getting the flop to read as funny instead of glitchy.

YEETLE stopped being a physics toy and became an actual game today. Four arenas instead of one, a play zone that shrinks so rounds don't drag, meteor showers and wind gusts that shove everyone around, small missions to chase, and a pile of silly bug cosmetics to unlock. The one-gesture control stayed exactly the same, everything around it got deeper.

The struggle all day was balance. With this many moving parts it's easy to make a round that's impossible or boring, and I can't read the math, so I just played it again and again and told Codex where it felt off until it settled. Somewhere in here the ledger quietly crossed 100 logged hours since day one. A hundred hours of building things I didn't know how to build, with tools doing the parts I can't. It doesn't feel like an achievement while you're in it, it just feels like a lot of late nights, but it's on the record now.

The new game is called YEETLE. You are a little bug on a round arena, you drag back to aim and let go to fling yourself, and the goal is to knock the other bugs off the edge without flying off yourself. One gesture controls everything, which is exactly the kind of thing that reads instantly on a portal like CrazyGames.

Codex built the core physics and I spent the day describing what "good" felt like, over and over. I can't tune a number in the code myself, I can only play it and say too floaty, or too heavy, or the wall bounce feels wrong, and let the AI adjust and hand it back. The launch had to feel powerful but fair, the rebounds off the edges had to be readable, and the enemy bugs had to move like they had a plan instead of drifting around like furniture. That last part was the hard one. Making bots that feel like real opponents took a lot of back and forth, and it's strange work: I'm directing a game I couldn't code a single line of.

Like last week, I didn't manage to journal each night while this was happening, so these next five days are rebuilt from the project files, the builds and the CrazyGames dashboard rather than from notes written the same night.

The first real player numbers finally showed up, and they were not what I expected. Don't Put Them Together, the game I had quietly written off as a flop, kept getting played and is now sitting at 202 plays. Dungeon Breaker RPG, the one I was actually proud of and expected more from, got accepted too but only has 35. Both are live on Basic Launch, both have earned exactly zero euro, and across everything it's 237 plays and nothing in the bank.

I sat with that for a while. The lesson is blunt: I am a bad judge of my own games. What I think will land and what actually gets played are two different things, and the only way to know is to ship and read the numbers. So the plan doesn't change, it gets reinforced. Keep making small games, keep putting them in front of real people, stop trusting my gut and start trusting the data.

So today I started game number three. I still can't write code and I still don't really know how games are built. Codex writes the game, Claude Code helps me with the polish, the words, the site and the decisions. That is the whole studio. The new idea is a physics battle royale where you are the thing being thrown.

I'll be honest, I didn't write a note every night this week. It was too full and too fast, so this one covers the whole stretch, rebuilt from the project files, the README and the CrazyGames upload log rather than from a diary written the same night.

Hotel Caos got its Basic Launch approval but the player numbers never showed up. That stung. Days spent on tutorial pacing, risk previews, viral share cards, all of it, and the game just didn't find an audience. For a moment it felt like proof I was doing this wrong.

But this was the exact test the whole plan was built for: ship small, look at the real numbers, decide fast. So instead of sitting with a game that wasn't working, I moved straight into a second one. Roughly 4 to 5 hours a night across this week, about 31.5 hours total, went into Dungeon Breaker RPG: a brick-breaker roguelike where you fling orbs, break blocks, collect relics, pick a hero and fight through a 100-stage dungeon campaign. Built on Phaser 3, TypeScript and Vite this time, with the CrazyGames SDK, save fallback, leaderboard hooks and a mobile orientation guard built in from the start instead of bolted on at the end.

On the 12th the new game went up for its own CrazyGames review. As of today it's still waiting, and so am I. The first game not landing was the first real signal from actual players, not from me guessing. I don't know yet if the second one lands either, but I know I didn't wait around to find out. First one flopped. I learned. I built the next one anyway.

The CrazyGames preview showed rooms overlapping at some screen sizes and guests missing from the queue panel on the right. Fixed the layout for desktop and iframe sizes, made the room grid safer, gave the guest queue priority in the side panel, and hid panels that don't fit on short screens.

Repackaged the build, re-uploaded it, and finished the rest of the submission: categories, tags, description, controls text, cover art, preview video. About an hour tonight, and Don't Put Them Together is now sitting in CrazyGames' Awaiting Review queue. Out of my hands again.

Testers flagged that guests weren't fully visible on mobile and tablet, rooms stayed stuck at the top of the screen, and landscape mode was rough. So today was pure device testing.

Touch devices switched from drag-and-drop to tap-guest-then-tap-room, which is a lot more forgiving on a phone screen. Then I went through phone portrait, phone landscape, tablet portrait, tablet landscape and desktop one at a time, checking rooms, the guest queue, the HUD, the tutorial and horizontal overflow on each. Rebuilt the CrazyGames upload folder and zip at the end so the fixes actually ship. About two hours, all of it unglamorous but necessary.

Slept on last night's warm room and it still didn't feel right, so I undid most of it. Back to the cooler night studio, and the only change I kept was moving the sketch board from the side wall to behind the monitor.

The bigger decision was letting Hotel Caos open on mobile and tablet, not just desktop. There was no real reason to keep it locked to the computer, so that limit came off and the latest build went live everywhere.

The wisp took most of the evening. It used to chase whatever you hovered, fading and jittering the whole time, which looked more nervous than alive. After a few passes it settled into something simpler: hover over it and it just stops and glows, let go and it drifts again. Less movement, more character.

I also finished the "Follow the build" social buttons I'd left half-done, and locked the game's public names for good: the game is Don't Put Them Together, the hotel inside it is Hotel Chaos, and its own social handle is @hotelcaos, separate from @wisploft.

Same night, the site caught up. Don't Put Them Together went live inside the Wisploft site itself, packaged as a self-contained HTML5 build under /games. It has its own page now, in the same style as the rest of the studio: a hotel banner, a click-to-play area, a fullscreen link, and a small gallery of room shots.

The social chips in the HUD are wired to the real @wisploft accounts on YouTube, Instagram, TikTok and X, so they no longer point at nothing. The wisp got more alive too, drifting instead of following a fixed path, breathing light instead of just glowing at one brightness.

I also tried warming the room up tonight: moved the arcade, tucked the mailbox under the shelf, brightened the lights. It looked nice but it didn't feel like Wisploft. I left it for the night and decided to judge it again in the morning with fresh eyes.

Today the first game stopped being something that only runs on my machine. I opened a CrazyGames developer account, filled in everything the submission form asked for, and sent Hotel Caos in for its first review. It is a small step on paper and a big one for me: this is the first time anything I built here has left my hands and gone in front of an outside judge.

On the tool side, Codex hit its usage limit today. I had a reset gift from two days ago and used it to keep moving. Claude still has about half its limit left, so design, writing and polish work keep going there while Codex recharges. Managing AI usage limits turned out to be its own small skill, not just a background detail.

I don't know yet if the game gets accepted. But the fact that it is out of my hands and being looked at is progress by itself.

The KVK approval for the name and activity change came through today. The company is now officially named Wisploft, and its registered activity matches a game studio instead of whatever it used to say.

This is not a new company built from scratch. It is the same company from day one, just renamed and repointed at the thing I actually want to build. Small paperwork moment, but it makes everything before it real on paper, not just in my head.

The same night, once the game was packed up, I went back over the whole website slowly and looked for the things that were quietly wrong. The kind you only find when you stop building and start doubting.

A few real problems turned up. The site used to trust whatever address a browser claimed to have, which meant someone could fake it and slip past the admin login lock, the one vote per person rule and the spam checks. Now the real address is used unless the site sits behind a proxy I actually trust. I also found dead links left over from when the site became a 3D room, a share image that came up blank on social media, and a video embed that could leave an empty grey box. All small, all fixed.

Then I closed the gaps that had been nagging me. There is a proper unsubscribe page now, with a signed link, so anyone on the launch list can leave on their own. There is a backup that can run every night, so the ledger and every night note are safe if something breaks. The idea board finally shows what happened to each idea after the vote, made or skipped, so voting leads somewhere. And the contact form no longer wipes what you typed if you get something wrong.

The lesson of the night is simple. The first build is what makes a thing look finished. Going back over it, doubting each piece, is what makes it actually hold.

A shorter night, about an hour, spent making the game feel like a place instead of a screen of buttons. The menu sits in a real hotel now, the rooms have an inside, and I drew the cover and icon the store page will need. Small things, but they move it from a prototype toward something I would not be shy to show.

I will be honest about my head tonight. I keep asking whether I am doing the right thing. Going straight to the App Store felt too big and too blind, so the plan is to put this on the web first, watch whether people really play, see where they leave and what they enjoy, and only then decide what to grow.

It scares me a little. The game might not get accepted, or might not find players, or might just get a shrug. But for the first time I am not only having ideas, I am about to put a real, testable game in front of people. The rule I am keeping is simple. Test before you guess, then grow into what the numbers show.

Hotel Caos was funny, but it had one real problem. You had to learn the rules before you could play well, and the first tester hit the same wall I did. You could lose a round without ever understanding what you did wrong.

So today was about making it fair. Now when you pick a guest, the rooms react on their own. Green means safe, red means trouble, and a short note says why, like needs quiet, or clashes with the baby. You can read the danger instead of memorising it. That one change turned the game from a memory test into a quick thinking game.

The other half of the day went to the part people actually share. The end of a run now builds a card, the funniest disaster of the shift, as a square picture you can save and post. I also softened the first minute for new players, and fixed a bug where finishing the tutorial by running out of time never counted as done.

And I finally claimed the name everywhere. All the social accounts are open now under the same handle, @wisploft, so the story has a home before the videos start. About three hours tonight.

Tonight was about turning the website from something that looks finished into something that actually holds up. First I wired up the honest ledger, so every euro and every hour I spend shows on the site by itself, and set the two subscriptions I will never cancel to log themselves every month. No more editing a spreadsheet by hand.

Then I filled in the journal and rebuilt the badges into a real roadmap, from the first web game all the way up to a million players, so there is always something to aim at and something to unlock as I go.

The rest of the night went to the boring but important work. I shrank the heavy images so the page is not a few megabytes anymore, moved the fonts and the 3D code onto my own server so visiting the site does not quietly hand your data to anyone, made the room lighter on phones, put a lock on the admin login, and wrote a real privacy policy and terms. About four hours, and the whole thing feels a lot more solid now.

Today the plan flipped. I was going to start with an iOS app, but a YouTube video about CrazyGames made me stop and rethink. Why not test small games on the web first, see what people actually like, and only then turn the ones that stick into mobile games. Faster feedback, less time wasted on ideas nobody wants.

The same evening the first game idea showed up. It is called Hotel Caos. You run a hotel where every guest is a problem. A vampire refuses a sunny room, a baby needs quiet, a DJ wants a party floor, a ghost only sleeps in a haunted room. You get 60 seconds to put the right guest in the right room, and the wrong neighbours set off a chain of funny disasters.

The first build felt like a card game more than a hotel, which was the big warning sign. The moment I pushed it toward real characters and real rooms it started to feel like an actual game. I am building it with Codex and getting it ready as an HTML5 package for CrazyGames. Still rough, but it exists now.

The direction finally clicked. Instead of a flat picture of a room, the site became a real time 3D night studio you can look around. You drag to look, the little wisp mascot floats around, and the furniture is clickable and opens the journal, the projects, the ledger and the rest.

Two practical lessons from today. The 3D files were huge at around 39MB, which is unusable on the web, so I compressed them down to about 3.6MB with gltf-transform, meshopt and webp. And anything with text on it, like a poster, should be drawn with code on a canvas instead of generated by an image tool, because the image tools always mangle the letters. Build the room object by object, compress before you ship, and keep text out of the image generator.

Spent the day fighting the design. Every version that came out looked too generic, too much like a SaaS landing page, nothing like the mood in my head. I lost count of how many directions I tried. It was at least seven.

The thing I learned the hard way is that writing code before the look is locked just gives you polished garbage. You need one reference frame that makes you say yes, that is it, before you write a single line. Without that anchor the AI just builds something clean and soulless. So I stopped coding and went back to nailing the visual direction first.

Today I started building the Wisploft website. I do not want a normal company page. I want something that feels like you walked into the studio while the work is still going, with the days, the hours, and the money all out in the open.

I also picked up a Higgsfield subscription at €63.30 a month to help with visuals. I am honestly not certain it will pay off yet. The bet is that it can help with website art, game art, and some mood tests. I will know in a few days whether it earns its place or not.

Wisploft started today. I had been sitting on the idea of building a game studio out in the open for a while, and today I finally moved on it.

The name came first. Wisploft is two ideas joined together. A wisp is the faint light that guides you in the dark, and a loft is the quiet workshop where you make things late at night. That is the whole feeling I want this to carry.

Once the name felt right I checked the domain straight away and bought wisploft.com before I could talk myself out of it. Then I filed to rename my existing company to Wisploft, so this is a real studio and not just a side project with a logo.

The first money went out today too. The chamber of commerce registration was €80.10, the domain was €12.30 for the year, and I started my ChatGPT and Claude subscriptions at €22.99 and €21.78 a month. I have close to zero coding background, so those two are not optional for me. They are how I actually plan to build this. No income yet, just the first costs on the board.