Ziv Mizrahi bought a copy of Super Mario Maker 2 in Eilat, played it, loved it — and got annoyed by one missing thing.

You can build anything in Mario Maker. Enemies, platforms, pipes, obstacles, entire levels in Nintendo’s visual language. What you cannot do is write a single line of text. Not one word. No dialogue, no plot, no reason why Mario is jumping on that particular mushroom.

“It lacked the ability to tell a story,” Ziv told me. “It’s more about mechanics. Jump from here, kill that.”

Most people would shrug and keep playing. Ziv, a marketing manager with 20 years of Facebook campaigns, newsletters and landing pages behind him — and zero years of game development — decided to build the missing piece himself. A game engine where the story comes first. No code required.

That was October. Today the thing exists, it works, and a philanthropic fund he never contacted reached out asking him to build a training game for them. He’s still a solo founder. Still bootstrapped. Still asking himself, out loud, whether the world needs what he’s building.

That last part is what makes this story worth telling.

Not a Coder. A Builder.

Let’s get the biography straight, because it explains everything about the product.

Ziv is 40. BA in HR and management from Bar-Ilan. A copywriting diploma from the era when people still dreamed of writing TV commercials. Then two decades of marketing: lead generation, e-commerce, customer journeys, email flows. Twice he detoured into talent sourcing — including at Bank Discount, hunting product managers and compliance officers for one of Israel’s biggest banks. Right now he’s finishing an MA in instructional technologies at HIT, the Holon Institute of Technology.

Notice what’s missing: engineering. He studied some computer science in high school. That’s it.

But there’s a pattern in his career that matters more than any CS degree. He was always the marketing manager who did the work himself — designed the newsletter, built the landing page, drew the illustrations — instead of managing vendors who did it for him. Around 2013 he ran an art project called Pptify: recreating masterpieces like Vermeer’s Girl with a Pearl Earring and Munch’s The Scream using nothing but PowerPoint shapes. It got him a feature in Haaretz’s English edition and a mention in a German blog alongside David Byrne of Talking Heads, apparently a fellow PowerPoint artist. Bank Leumi wanted a customer calendar in that style. In 2022, a lecturer at Northeastern was still showing the works to students.

A man who rebuilds Vermeer in PowerPoint is a man who doesn’t accept that tools define what you’re allowed to make.

Then vibe coding arrived, and the ceiling lifted. “All of a sudden I’m not able only to do A, B and C. I am able to do A, B, C, D, E,” is how he describes discovering Base44 — the Israeli no-code platform founded by Maor Shlomo and acquired by Wix. Ziv compares Base44 to the Velvet Underground: not that many fans, but every fan started their own band. He became one of those fans. Then he started the band.

Three Ingredients, One Engine

Quest IL didn’t come from a single flash. Ziv names three sources, and the combination is the pitch.

First: Mario Maker 2. A creation platform anyone can use — but mute. No story.

Second: Day of the Tentacle, the legendary 90s adventure game. Characters, objects, dialogue, riddles — which Ziv played on the family PC, acquired a few years after his first console. And that first console deserves its own sentence: the Megason, the cheap Nintendo knockoff Israeli kids got when their parents couldn’t afford the real thing. Mario, Worms, Red Alert, The Sims — that was the childhood. When you break an adventure game down, he realized, it’s just a table: characters, objects, interactions between them. Once you see the logic, you can build the machine that builds the game.

Third — and this is the one that turns a hobby into a business — LEGO Serious Play. Yes, that’s a real thing: an official LEGO methodology sold to organizational consultants who use bricks to help companies work through mergers and management problems. Play as a corporate tool, with a price tag and a certification.

Creation platforms. Adventure-game logic. Games in the service of organizations. Stack the three and you get Quest IL: a no-code engine for narrative-driven games, aimed not at gamers but at the people who run training, onboarding and knowledge management — people who currently torture their colleagues with slide decks.

The L&D world runs on presentations, documents and compliance modules — and to be fair, many of its professionals already use AI daily to produce training materials. Ziv’s bet isn’t that they’re behind the curve. It’s that there’s a layer above it they haven’t touched yet.

“Solutions like Quest IL make it easier than ever to create games that turn staff training into a more engaging and enjoyable experience.”

The Box Inside a Box

Here’s the technically honest part, and I respect Ziv for volunteering it before I asked.

Quest IL is built on Base44. A game-creation builder inside a builder. “It’s like building a box inside of a box,” he admits. “People will argue: why do I need your box when I can create a game on my own using Base44?”

His answer, after ten months of development: because most of his target users will never open Base44. The head of L&D at an insurance company is not going to vibe-code a compliance training game. Quest IL gives them a builder with pre-defined parts — scenes, characters, objects, interactions, inventory, dialogue, riddles that double as quizzes — and stitches together a stack of external AI services through APIs so the user never leaves the platform. ElevenLabs for voices. Fal for generated video cutscenes. PixelLab for turning a single character image into an animated sprite sheet. He calls it an orchestration platform, which is accurate and also the first time I’ve heard the term used by someone who learned it by building one.

Then there are the agents, which is where it gets interesting.

Ziv trained two AI agents on Base44. The first is a librarian: it generates backgrounds and characters and files them into Quest IL’s asset libraries. As of our conversation, 865 backgrounds and 990 characters, so users don’t start from a blank page.

The second agent he trained to use Quest IL itself — fed it the building blocks, the interaction logic, the features, and told it to make games. The early results were, in his words, crappy. Then they weren’t. To test it, he prompted the agent with the plot of the Seinfeld episode where Jerry dates a beautiful woman with man hands — she cracks lobsters bare-handed, opens beer bottles without an opener. Five minutes later the agent had produced a playable narrative game based on that plot.

The output still needs human polishing. But a solo founder who can’t hire has effectively recruited two employees who work for API credits. “Now it feels like I’m less alone,” he wrote on LinkedIn. It’s a joke, but it’s also the most concrete illustration I’ve seen of what “AI-native solo founder” actually means in practice.

The Haunted Castle Test

Before we talked, Ziv made me play his game.

Ziv’s Haunted Castle is a 3D horror-comedy browser game he built as a side project — a monster-filled mansion, a murderous hunchback, a sword you’re supposed to find near the castle gate. I found the castle first, spent a while running from a ghost, and only then found the sword. Killed the hunchback. Escaped by taxi. Ziv informed me I’m one of the roughly 10% of players who actually finish — he knows because he built a dashboard tracking the full funnel: who finds the sword, who kills the hunchback, who quits.

Here’s what matters about this silly game. It wasn’t built on Quest IL — it was built directly with Claude Fable 5, Anthropic’s newest model, on Base44, with 3D models from Meshy AI. Total build time: two days, maybe four half-days. Total cost, by Ziv’s estimate: 200 to 250 shekels. Around $70. Over 2,000 unique players, entirely organic — and by the end of the voting, a second-place finish out of 20 apps in Base44’s Launchpad competition, earning the Haunted Castle a spot on the platform’s Wall of Fame. Not bad for a side project built to test a model.

When I was running a hyper-casual games studio, a prototype like that took a Unity developer a week — a few hundred dollars of salary at minimum, and that’s the cheap version. Ziv built it for the price of a decent dinner in Tel Aviv, without writing code.

Inside Quest IL the math is even more aggressive: he estimates a full narrative training game — dozens of scenes, dialogue, voices, quizzes — costs dozens to low hundreds of shekels in AI credits to produce. He’s building a spending dashboard so end users can see exactly what they burned. Compare that to what a custom e-learning module costs from an agency and you understand why a philanthropic fund cold-contacted a solo founder with no marketing budget.

But the Haunted Castle also carries a warning, and Ziv knows it. If Fable 5 can produce a good-looking 3D game in two days from raw prompts — what protects the box inside the box?

“The ground beneath you is always moving. I keep examining, in my inner dialogue: is there a need for Quest IL? For now I think the answer is yes.”

He sees the wave of impressive AI-generated games since Fable 5 launched. The value, he argues, is precisely for the non-technical: the HR professional who needs a recruitment game, not a game engine. If you know Unity, you were never his customer anyway.

I’d add: the Haunted Castle also showed him a second business. Bespoke games. The platform makes compromises — a box that works the same for everyone produces generic results. Fully custom games, built with the skills he’s accumulated, don’t. He’s now openly planning to offer both. The business model, he says with a candor you don’t hear at demo days, “is not 100% solved.”

Slide Four of Twenty-Seven

About demo days.

Ziv’s first contact with the outside world came when he answered a Facebook ad: one of Israel’s largest law firms, Goldfarb Gross Seligman, was recruiting for the second cohort of its early-stage founder bootcamp, run with Deloitte and Discount Tech. He sent a one-pager and got in, alongside 30 other entrepreneurs, some of whom had nothing but an idea. For a guy who’d been building alone in the metaphorical basement, it was the first external stamp: you’re an entrepreneur now.

Then came the demo day. Real investors — five or six representatives of the pre-seed funds that write Israel’s first checks. Ziv — a marketing manager, a man who’s given the elevator-pitch lecture himself — showed up with a 27-slide deck. He didn’t rehearse. He got to slide four. The one that says every second person on Earth is a gamer. Time. He never reached the product.

“There are no sentiments,” he says. “They said thank you very much, do you have questions? And obviously none of them had questions, because I didn’t get to the point.”

Two investors scanned the QR code anyway and added him on LinkedIn. One told him he seemed unusually creative — she’d noticed a slide of photos from his trips to Hong Kong and Japan. Which produced maybe the most Ziv piece of founder advice I’ve heard: when you pitch, tell them a bit about yourself, not only the project. Sometimes it saves you.

Since then he’s approached exactly one venture fund. Initial interest, a long questionnaire, then a rejection with an unusual shape: they think it can be a great business — they just don’t think it’s a VC business. Ziv disagrees. But he’s also not bleeding. He consoled himself with a statistic ChatGPT served him — the average startup takes 40 to 80 approaches before a term sheet — rolled his eyes, and went back to shipping features.

His actual strategy is quieter: land two or three design partners, prove the demand, and let that do the talking to investors. One conversation with the philanthropic fund is in motion. He’s committed to funding the venture himself as long as possible — and, he says, Quest IL will fulfill its mission with or without external money. He’s even open, without embarrassment, to the idea of eventually being acquired by someone like CET, Israel’s veteran Center for Educational Technology, whose learning software every Israeli schoolkid since the 80s remembers.

And for the first time, after months of deliberate solitude — his own best friend passed on joining (“he’s not a gamer”), his own brother told him “I wouldn’t use your product and certainly wouldn’t pay for it” — Ziv recently had lunch with a prominent figure from the Israeli gaming industry about possibly joining forces. He won’t name names. But the soloist is considering a duet.

Why This One

I write about people who build. Most weeks that means someone with a term sheet, a team, a Delaware entity and a rehearsed narrative.

Ziv has none of that. What he has is a working product with two AI agents on staff, a grandmother’s tune for a studio name — Takatushtush, after the melody his 93-year-old grandmother played to entertain him as a toddler — an inbound design partner, a rejected pitch, a supportive-but-brutal brother, and a habit of publicly documenting the humiliations along with the wins. His LinkedIn reads like a founder’s journal written by someone who never learned that founders are supposed to pretend.

He’s also a live experiment in the question every solo builder is asking right now: how far can one person get when the marginal cost of software drops to the price of lunch? Ten months in, the answer is: further than the industry assumed, and not yet far enough to know how it ends.

When he used Quest IL for a final project in his master’s course and it worked, he wrote that he was “happy as a Super Mario picking mushrooms.”

The plumber never asked permission to enter the pipe either. He just jumped.

💜