Chapter 5: The Cracks

Three months in. May.

David’s checking account: $38,950.

The number sits on his phone screen. He refreshes the banking app. Still $38,950.

He’s been careful. Brutally careful. Rice and eggs. No coffee shops unless he’s meeting someone. The heat stays off even when the apartment is cold. He walks instead of taking the subway when he can afford the time.

The last severance payment came two weeks ago. That safety net is gone now.

Still. Three months. Nearly nine thousand dollars gone from savings, even with the severance cushion.

At this rate: eight months left. Maybe ten if he stops eating out entirely.

The calculator in his head runs constantly now. Every purchase translated into days of runway. That sandwich: half a day. New headphones to replace the broken ones: three days. The domain name renewal: two days.

His mother texted yesterday: Your cousin just bought a condo. Two bedrooms.

David didn’t respond.

The prototype is almost done. That’s what they tell themselves.

“Almost done” has meant different things over the past month. First it meant “core features working.” Then “ready for internal testing.” Now it means “showable to real users without it crashing immediately.”

The bar keeps lowering.

David sits at his desk—the IKEA desk that cost three days of runway—staring at the adaptive learning algorithm. It’s supposed to detect how students learn. Visual, auditory, kinesthetic. Fast-paced or deliberate.

The problem: it’s guessing. Badly.

He runs the simulation again. A test student clicks on a video. The algorithm decides: visual learner. Serves more videos. The student skims the text quickly. The algorithm adjusts: prefers text, fast-paced.

But what if the student clicked the video because the text was poorly written? What if they skimmed because they were bored, not because they prefer fast-paced content?

The algorithm can’t tell the difference. It just guesses and hopes.

David closes the file. Opens it again. Stares at the code.

Behind him, Alex is on a video call with a potential beta tester—a teacher friend from college. David can hear half the conversation.

“Yeah, we’re targeting high school students initially… Personalized learning paths… Uses AI to adapt to how each person learns best…”

The words sound good. Polished. Like they know what they’re doing.

David returns to the algorithm. Changes a parameter. Runs the test again. Slightly better. Still guessing.

He changes it back.

At 7 PM, they order pizza. Too expensive, but neither of them has eaten since breakfast and the code isn’t working and they need something.

Alex pays. David doesn’t argue. He’s started keeping a mental tally of who pays for what. Alex has paid for the last four meals. David paid for the domain and hosting. Alex covered the API costs. David bought the whiteboard.

They don’t talk about it. But the scorecard exists.

They eat in silence for a few minutes. The pizza is mediocre. David eats three slices anyway.

“I think we should push back the user testing,” Alex says.

David looks up. “We’re already behind schedule.”

“I know. But the prototype isn’t ready. If we show it now and it crashes—”

“It won’t crash. The core functionality works.”

“The adaptive algorithm—”

“—will improve with real data. We can’t get real data without real users.”

Alex sets down his pizza. “David. It’s not good enough yet.”

The words hang in the air.

Not good enough.

David knows he’s right. The prototype is held together with duct tape and optimism. The UI is clunky. The algorithm guesses more than it learns. It crashes if you click certain buttons in the wrong order.

But if they wait longer, they burn more savings. More days off the runway. And for what? To make it slightly less broken?

“How much longer do you want to wait?” David asks.

“Two weeks. Maybe three.”

“That’s a thousand dollars.”

Alex pauses. “I know.”

“Do you? Because I’m looking at my account every morning and watching it go down and we haven’t made a single dollar yet. We haven’t validated anything. We’re just building in the dark and hoping.”

“So what do you want to do? Show them something broken?”

“I want to find out if we’re building the right thing. Before we run out of money building the wrong thing.”

The silence is different now. Heavier.

Alex picks up another slice. Doesn’t eat it. “Okay. When?”

“Next week. I’ll recruit the testers.”

“How many?”

“Ten. Maybe twelve.”

“High school students?”

“Yeah.”

Alex nods. Sets down the pizza. “Then we better fix the button crash bug.”

David spends the next three days recruiting. He posts in local education forums. Reaches out to teachers he vaguely knows from college. Offers $25 gift cards to participants.

Twelve students sign up. Ages 14 to 17. Mix of school backgrounds.

The sessions are scheduled for Saturday. Two weeks earlier than planned. Three weeks later than David wanted.

The compromise that satisfies no one.

Thursday night, David stays up until 4 AM fixing bugs. The button crash. The infinite loading screen. The feature that accidentally deletes user progress.

He makes a list of things that are still broken:

1. Algorithm guesses too much
2. UI confusing on mobile
3. Content library too small
4. Onboarding flow skips steps randomly
5. No way to track actual learning outcomes
6. Sometimes serves the same question three times in a row

He stares at the list. Closes his laptop. Opens it again.

Adds:

7. Might not solve the actual problem

Friday evening. Alex leaves around 6 PM—his mother’s birthday dinner. David says he’ll keep working on the prototype.

He does, for about an hour. Then he stops.

Opens his banking app. $37,200.

Opens LinkedIn. Scrolls through updates. Someone from his old company just got promoted. Senior Staff Engineer. David wonders what the salary is. $300k? $350k?

He closes LinkedIn. Opens his email.

Three unread messages from his mother. He’s been avoiding them. He knows what they say without reading. Questions about the job search. Suggestions about companies hiring. Worry disguised as helpfulness.

He opens the most recent one anyway.

Your father is asking about health insurance. When you get a new job, make sure they have good coverage. Remember when you needed that specialist? Don’t pick a plan with high deductibles.

David closes the email.

She doesn’t know he’s not looking for jobs. Doesn’t know he’s building something. Doesn’t know he’s choosing this.

He’s still telling her he’s “interviewing.” The lie gets easier each time.

His phone buzzes. Alex.

How’s it going?

David looks at his screen. The prototype is open. He hasn’t touched it in twenty minutes.

Good. Fixed two bugs. Should be ready for tomorrow.

Nice. Get some rest.

David puts the phone down. Looks around the apartment.

Two desks. Two chairs. Cables everywhere. The whiteboard covered in diagrams that made sense two weeks ago. Empty coffee cups. A blanket on the couch where Alex crashed last Tuesday.

This is his life now. This is what he chose instead of $300k and good health insurance.

He opens the prototype. Runs through the user flow one more time. Clicks through the screens. Everything works. Mostly.

Tomorrow they’ll find out if it matters.

Saturday. 10 AM.

Coffee shop private room. David early. Alex brings coffee.

“Ready?”

Not ready.

First student, 17, wants to study CS. Clicks around fifteen minutes. Feedback: “It’s cool. Kind of like Khan Academy?”

David’s chest tightens.

Second student, 15, struggles with math: “Pretty similar to stuff at school. They all kind of feel the same.”

Third: “Like a homework app.”

Fourth: “Reminds me of Duolingo.”

Fifth: “I probably wouldn’t use it unless I had to.”

By the sixth session, David stops taking notes. He knows what’s coming.

The seventh student is fourteen. Quiet. Barely makes eye contact. She uses the prototype for ten minutes without saying anything.

When they ask for feedback, she pauses.

“It’s okay.”

David waits. Something in her hesitation feels different.

“Was there anything you liked about it?” Alex asks.

She picks at her thumbnail. “I guess… there was that part at the beginning where it asked what I’m curious about.”

David leans forward slightly. “The onboarding questions?”

“Yeah. Like, it asked what I actually want to learn. Not what I’m supposed to learn.” She shrugs. “No one ever asks that.”

“Did it feel useful?”

“I mean, I don’t know. I just typed something random because I didn’t think it actually mattered. Does it?”

David and Alex exchange a glance.

“It’s supposed to,” David says. “To personalize what you see.”

“Oh. That’s cool.” She looks back at the screen. “The rest of it kind of felt like homework though.”

They thank her. She leaves.

Alex closes his laptop. The coffee shop is getting crowded—Saturday afternoon rush. They have six more sessions scheduled.

“She liked the curiosity question,” Alex says.

“She typed something random because she didn’t think it mattered.”

“But the idea resonated. That’s something.”

David doesn’t respond. His phone buzzes. Reminder: Next session in 10 minutes.

They do the remaining six sessions. The feedback blurs together.

It’s fine. It’s okay. It’s like the other apps. I’d maybe use it. Probably not. It’s cool, I guess.

At 6 PM, they’re done. Twelve sessions. Two hundred dollars in gift cards. Six hours of feedback.

They pack up in silence. Walk back to David’s apartment. The spring evening is warm. People are out—couples, families, groups of friends heading to dinner.

David and Alex walk without talking.

Back at the apartment, Alex sits on the couch. David stands by the window.

“So,” Alex says finally. “That was rough.”

“Yeah.”

“Not all bad though. The girl who liked the curiosity question—”

“One person. Out of twelve. And she still said the rest felt like homework.”

“It’s a first prototype. We can iterate.”

David turns from the window. “Into what? Another homework app? Another Khan Academy clone? They all kind of feel the same, apparently.”

“We just need to figure out what makes it different.”

“I thought we knew what made it different. We spent two months building it.”

Alex is quiet for a moment. “David. This is part of the process. User feedback. Iteration. It’s—”

“Don’t.” David’s voice is flat. “Don’t give me the startup playbook speech. I know the process. I know we’re supposed to embrace failure and iterate quickly and all that bullshit. But we just spent eight thousand dollars to build something twelve people said feels exactly like everything else.”

The words sit in the air.

Alex stands up. “I’m going to head out. Give you some space.”

“You don’t have to—”

“Yeah, I do. We’re both tired. Let’s regroup tomorrow.”

Alex grabs his jacket. Stops at the door. “For what it’s worth, I still think we’re onto something. We just haven’t found it yet.”

He leaves.

David stands alone in the apartment. The whiteboard mocks him. “MVP Complete: May 15” is still written at the top in blue marker. Below it, the timeline that seemed optimistic six weeks ago now seems delusional.

He opens his laptop. Pulls up the feedback notes.

Twelve users. One semi-positive comment about the curiosity question. Eleven variations of “it’s like everything else.”

His phone buzzes. His mother.

Have you heard back from any interviews? Your father knows someone at Microsoft if you need an introduction.

David sets the phone face-down.

Opens his banking app. $36,800.

Two hundred dollars for gift cards. Eighty dollars for the coffee shop. Twenty-two days of work. Eight thousand dollars in savings.

To build another homework app.

He closes the laptop. The apartment is dim—he never turned on the lights. The city glows through the window. Out there, people are having normal Saturday nights. Dinner. Movies. Drinks with friends.

David sits on the floor. Back against the couch. The ceiling has a crack he’s never noticed before.

His phone buzzes again. He ignores it.

The prototype is open on his laptop. The screen has dimmed to save battery. He can barely see it in the dark.

He doesn’t open it again.

Just sits there. In the dark. As the city continues outside and the crack in the ceiling stays unnoticed and the savings account ticks down and twelve people’s voices echo:

It’s like everything else.

They all kind of feel the same.

Another homework app.

The cursor blinks on the black screen.

David closes his eyes.

End of Chapter 5