Chapter 10: Demo Day
Dr. Yoon did not, as a rule, show emotions during student presentations.
She was famous for it—twenty minutes of poker-faced silence while students sweated through their demos, followed by three sentences of feedback that could range from “adequate” (high praise) to “revisit your assumptions” (academic code for catastrophic failure). Rumors circulated that she had once listened to a thirty-minute presentation on distributed hash tables, said “your hash function is broken,” and dismissed the entire group without further comment.
So when Dr. Yoon leaned forward during Dojun’s A-star demonstration and said “Show me that again,” three rows of students turned to stare.
“This is the base route from the engineering building to the library,” Dojun said, clicking the demo. On the projector screen, a campus map appeared with a clean path drawn between two buildings—the standard shortest path. “Dijkstra’s algorithm. Optimal for distance.”
He clicked again. The path changed—routing around the main walkway, through a covered passage, avoiding an open quad.
“Same route, same start and end. But now the context layer is active. It’s Tuesday at noon, which means the main walkway is at peak crowd density. It’s also raining.” He pointed to the sidebar Hana had designed—a clean panel showing toggle switches for weather, time, and crowds. “The A-star heuristic adjusts edge weights in real time. The new route is two minutes longer but avoids the crowd bottleneck and stays under cover.”
Silence in the classroom. Thirty students stared at the screen.
“And this,” Hana said, stepping forward with the confidence of someone who had rehearsed this moment in front of her bathroom mirror, “is the user interface. Every context factor is a toggle. Users don’t need to understand the algorithm—they just tell the system their preferences, and the path adapts.”
She clicked through her interface mockups—now digitized from her hand-drawn wireframes into clean, functional screens. The design was warm and intuitive, with a color scheme that felt friendly rather than clinical. Icons were hand-drawn in her distinctive style. The campus buildings were represented as cute, recognizable illustrations rather than generic markers.
“We also added an accessibility mode,” she continued. “For students with mobility constraints, the system avoids stairs and routes through buildings with elevators. The edge weights for accessibility mode are separate from the standard weights, stored in a parallel graph.”
Minjae, who had been standing by the laptop managing the data feeds, added: “I personally walked every route on campus to collect the timing data. Forty-seven routes, twenty-two buildings, three time periods. Including the path behind the chemistry building that goes through a bush. That bush is in the graph. It has a weight of 1.8x.”
Scattered laughter from the class. Dr. Yoon’s poker face held, but there was something around her eyes that might have been the ghost of amusement.
“Questions,” she said. Not “any questions”—just the word, flat and expectant.
A student in the front row raised his hand. “How do you handle real-time data? The crowd density estimates—are those based on actual measurements or predictions?”
“Currently, predictions based on class schedules and historical patterns,” Dojun said. “But the architecture supports real-time data injection. If the university deployed foot-traffic sensors, the system could consume that data through a standardized API.”
“What’s the computational overhead of the context layer compared to standard A-star?” Another student.
“Approximately 15% increase in pathfinding time. Negligible for a campus-scale graph. The bottleneck is the heuristic evaluation, but since our context weights are precomputed per time window, the per-query cost is minimal.”
Dr. Yoon raised her hand—not to ask a question, but to stop the questions. The room went silent.
“The assignment was to implement Dijkstra’s and A-star, compare their performance, and build a simple UI,” she said. “You built a contextual pathfinding system with dynamic heuristics, a parallel accessibility graph, real-time data architecture, and an interface that I would describe as—” She paused, searching for the word. “—polished.”
“Is that good?” Minjae whispered to Dojun.
“I don’t give compliments,” Dr. Yoon said, apparently having the hearing of a bat. “I state observations. My observation is that this project exceeds the scope of the assignment by a significant margin. The algorithm implementation is clean, the data collection is thorough, and the design work is—” Another pause. “—not what I expected from an algorithms course.”
“I’m a design student,” Hana said. “I believe algorithms deserve good interfaces.”
“They do. Most algorithm students don’t realize that.” Dr. Yoon looked at the three of them in turn. “Full marks. And Park—see me after class. I want to discuss your A-star variant.”
The class burst into whispered chatter as the next group set up their presentation. Minjae collapsed into his chair with the boneless relief of a man who had just survived a firing squad.
“Full marks,” he breathed. “Full marks from Yoon. That’s like getting a hug from a cactus. It shouldn’t be possible, but it happened.”
“We earned it,” Hana said, but she was grinning so hard it looked like it hurt. She turned to Dojun. “That A-star modification you added last night—the precomputed context windows—that’s what pushed it over the edge. When did you even have time to code that?”
“Last night. Between midnight and three.”
“You coded for three hours the night before the demo?”
“I couldn’t sleep. The original implementation bothered me. The per-query cost was too high.” He shrugged. “So I fixed it.”
“You ‘couldn’t sleep’ and your solution was to optimize an algorithm.” Hana shook her head. “Normal people watch TV when they can’t sleep. Or read a book. Or stare at the ceiling.”
“This was more productive than staring at the ceiling.”
“Debatable.” But she was still smiling. “Come on. Dr. Yoon wants to see you, and I have wireframes to show you after.”
Dr. Yoon’s office was the opposite of Kim Taesik’s—neat, minimalist, with exactly three items on her desk: a laptop, a notebook, and a pen. No coffee machine. No stacks of papers. No personal effects of any kind except a single framed photograph of what appeared to be a mountain trail.
“Your A-star heuristic,” she said without preamble. “The precomputed context windows. Where did you learn that technique?”
“I derived it for this project,” Dojun said. The lie was easier now—practiced, smooth. “The standard A-star heuristic doesn’t account for temporal variation. So I split the day into thirty-minute windows and precomputed the edge weight adjustments for each window. At query time, the algorithm selects the appropriate window and applies the precomputed values.”
“That’s a time-partitioned heuristic cache. It’s not a standard technique.”
“I wasn’t aware of that.”
“It’s used in logistics optimization—truck routing, delivery scheduling. But I haven’t seen it applied to pedestrian pathfinding.” She made a note in her notebook. “Have you considered publishing this?”
“Publishing? I’m a sophomore.”
“Publication is not restricted by academic year. If the work is novel and the methodology is sound, it can be published. I’d like you to write it up—formal paper, ACM format. I’ll review it and, if it meets standards, submit it to the Korean Conference on Computing.”
A publication. At twenty. In his previous life, his first publication had been at twenty-six, as a junior engineer at his first company. This was six years early.
“I’d be honored,” Dojun said.
“Don’t be honored. Be rigorous. I don’t put my name on sloppy work.” She looked up from her notebook. “Professor Kim tells me you’re in the accelerated track now. Graduate courses next semester.”
“Yes.”
“Good. You don’t belong in undergraduate algorithms. You’ll take my graduate seminar on Advanced Graph Theory. It’s challenging. I expect you to contribute, not just attend.” She closed her notebook. “That’s all.”
“Yes, Dr. Yoon.”
He stood to leave. At the door, she said: “Park. The design student—Lee Hana. Her interface work was exceptional. If she’s interested in the intersection of design and computer science, send her to me. We don’t have enough people who think about both.”
“I’ll tell her.”
“Do that.”
Hana was waiting in the hallway, leaning against the wall, sketchbook open.
“How was it?” she asked.
“She wants me to publish the A-star variant. As a formal paper.”
“Publish? Like, in a journal?”
“A conference. Korean Conference on Computing.”
“Park Dojun, published academic author at age twenty.” She closed her sketchbook. “Your resume is getting absurd. Contest winner, Kim Taesik’s research assistant, accelerated track, now a publication. What’s next, a Nobel Prize?”
“They don’t give Nobel Prizes for computer science.”
“They should. You’d be a strong candidate.” She fell into step beside him as they walked toward the campus exit. “Also—Dr. Yoon said something about your interface work?”
“She said it was exceptional. And she wants to talk to you about the intersection of design and CS.”
Hana stopped walking. “Dr. Yoon said my work was exceptional?”
“Her exact word was ‘exceptional,’ yes.”
“Dr. Yoon. The woman who once described a student’s thesis as ‘not entirely without merit.’ She said my interface was exceptional?”
“Would you like me to get it in writing?”
“I would like a moment to process.” Hana stood very still for five seconds, then let out a sound that was somewhere between a laugh and a scream. Two passing students startled. “Okay. Processed. That is the single best thing anyone in academia has ever said to me. I’m going to frame it.”
“You should talk to her. Seriously. She doesn’t offer meetings lightly.”
“I will. Tomorrow. After I’ve calmed down enough to form coherent sentences.” She grabbed his arm and pulled him toward the campus gate. “Come on. We’re celebrating. Jjigae place. My treat.”
“You treated last time.”
“And I’m treating this time. Consider it a pattern. Mathematically, you should be able to extrapolate.” She was practically bouncing as they walked. “Full marks, a publication offer, and Dr. Yoon calling my work exceptional. This is the best day of the semester.”
“It’s only April.”
“Then the bar is set high. I like high bars.” She turned to him, eyes bright. “Okay, celebration first. Then—I want to show you the portfolio project wireframes. I’ve been thinking about what we discussed. Invisible technology. Technology that works the way people already think.”
“What did you come up with?”
“Something ambitious. Maybe too ambitious. But you told me your mentor said the best technology is invisible, and I couldn’t stop thinking about it.” She pulled out her sketchbook and held it against her chest like a shield. “Promise you won’t say it’s impossible.”
“I promise.”
“Promise you’ll be honest if it’s bad.”
“I promise that too.”
“Good. Jjigae first. Revolution later.”
They sat in the same basement restaurant, at the same small table, with the same bubbling stone pots of kimchi jjigae. The ajumma recognized them now—she grunted with slightly less hostility than the first time, which Hana interpreted as affection.
“Okay,” Hana said, opening her sketchbook to a spread she had clearly spent hours on. “The portfolio project. I’m calling it ‘Bridge.'”
The spread showed a product concept unlike anything in the current technology landscape. Bridge was an interface layer—not an app, not a website, but a bridge (hence the name) between existing digital tools that didn’t talk to each other. In Hana’s vision, it would unify email, calendars, notes, and files into a single, intuitive workspace that organized itself based on the user’s behavior patterns.
“The core insight,” she said, tracing her finger along a flow diagram, “is that people don’t think in applications. They think in tasks. ‘I need to prepare for Monday’s meeting’ is a task that spans email (checking the agenda), calendar (blocking prep time), notes (writing talking points), and files (reviewing the presentation). But current technology forces you to context-switch between four different applications. Bridge would understand the task and surface everything you need in one place.”
Dojun stared at the wireframes. His pulse was elevated. His palms were damp.
Because what Hana had just described, in a sketchbook, in a basement restaurant, in 2006, was the core product concept of Prometheus Labs.
Not exactly. The details were different—Prometheus Labs had used AI to power the task understanding, while Hana’s version relied on rule-based pattern matching. The scale was different—Prometheus served millions of enterprise customers, while Hana imagined individual users. But the fundamental insight—that technology should organize around human tasks, not around application silos—was identical.
She had arrived at the same destination from a completely different starting point. Dojun had gotten there through decades of building and iterating and failing. Hana had gotten there through pure design intuition, sitting alone with a sketchbook and a question: How should technology actually work?
“Say something,” Hana said. She was watching him with a mixture of hope and anxiety. “You’re staring at my wireframes like they contain a secret message. Is it good? Is it terrible? Is it—”
“It’s the best product concept I’ve ever seen.”
She blinked. “Really?”
“Really. The task-centric approach is exactly right. People don’t think in apps—they think in goals. And the interface you’ve designed—the way the workspace reconfigures based on context—that’s not just good design. That’s a paradigm shift.”
“A paradigm shift.” She said it slowly, testing the weight of the words. “That’s a big claim for a portfolio project.”
“It’s a big idea. Bigger than a portfolio project.” He leaned forward. “Hana, this could be a real product. Not a class assignment—a real thing that people would pay for.”
“It’s a sophomore project for a design class.”
“Right now, yes. But the concept is sound. The interface is intuitive. All it needs is the technical backend to make it work.” He paused, choosing his words carefully. “I could build that backend. The task-understanding engine, the cross-application integration, the context detection. It’s ambitious, but it’s doable.”
“Doable for a sophomore CS student?”
“Doable for a team. You design it. I build it. Minjae does the data layer.” He was getting ahead of himself—he could feel it, the same dangerous momentum that had driven him to build Prometheus Labs the first time, the same intoxicating certainty that this idea could change everything. “We could have a working prototype by the end of the semester.”
Hana set down her chopsticks. She looked at him with an expression he recognized from a thousand previous encounters—the look she gave when she was deciding whether to trust someone with something important.
“You really think it’s that good?” she asked quietly. Not fishing for compliments—genuinely asking.
“I think it’s the kind of idea that people will look back on and say ‘that changed how we think about software.’ Maybe not this version. Maybe not this year. But the core insight—task-centric computing instead of application-centric computing—that’s the future.”
“You keep doing that,” she said.
“Doing what?”
“Talking about the future like you’ve already been there.” She held his gaze. “It’s either visionary or delusional. I haven’t decided which.”
“Can it be both?”
“That’s what you said about genius and insanity.” She picked up her chopsticks and stirred her jjigae thoughtfully. “Okay. Let’s do it. Bridge. You and me. And Minjae, if he’s willing to map more data.”
“He’ll be willing.”
“How do you know?”
“Because you’ll ask him, and nobody says no to you.”
She laughed—that bright, startled sound that he was becoming addicted to hearing. “Flattery will get you exactly one extra bowl of rice.” She waved at the ajumma. “Ahjumma! Rice one more, please!”
The ajumma brought the rice with her customary wordless efficiency. Hana distributed it between them, then raised her water glass.
“To Bridge,” she said.
“To Bridge.”
They clinked steel tumblers. The thunk echoed in the small restaurant.
Dojun ate his rice and felt the tectonic plates of his second life shifting beneath him. Prometheus Labs had started exactly like this—an idea, a partnership, a meal in a cheap restaurant, and the naive, beautiful belief that two people could build something that mattered.
The first time, it had ended in heartbreak. The company survived, but the partnership didn’t. The product succeeded, but the people failed.
Not this time, he thought, watching Hana sketch a new wireframe on a napkin, her pen moving with the quick confidence of someone who had just been told her ideas were worth building.
This time, the people come first. And the product will be better for it.
Outside the basement window, Seoul hummed its endless urban song. April was turning into May. The cherry blossoms were gone, replaced by the deep green of full spring. Somewhere, a new season was beginning.
And in a basement restaurant that didn’t need a sign, two twenty-year-olds had just decided to build the future together.
They didn’t know yet how far that decision would take them. But Dojun did. And for once, instead of fear, he felt something else entirely.
Hope.