My ASPLOS experience has just come to an end, and it was unforgettable. Before this trip, I had never attended an academic conference. The only conferences I had ever been to were industrial ones in China, where I helped my father hand out advertising materials when I was a little kid.
I want to divide this reflection into two parts: the journey and the learning. Since this was my first academic conference, I feel that learning about cutting-edge research was just one part of the experience. The non-technical side mattered just as much.
Journey
For the journey part, I want to organize the story around the feelings I experienced. Even so, the events still roughly follow the order in which they happened.
Tired
I do not like sitting in a car for too long because it makes me carsick, so I decided to take the GO Train to Toronto. Since it was a Saturday, I had to catch the earliest train, because there are only two weekend departures. The cost is that I had to wait at the airport until 3 p.m. for my flight.
I sat in the waiting area for hours with very little to do, only to find that my flight had been delayed. Fortunately, the plane just needed a new tire. After another half hour of waiting, I finally began my journey to ASPLOS.
At that moment, all my tiredness disappeared. It was the first time I had traveled abroad by myself, and even the first time I had taken a flight alone.
Energetic
The flight was so short that the attendants seemed to be racing against time: walking up and down the aisle, handing out drinks, collecting garbage, and preparing for landing almost immediately.
After arriving in Pittsburgh, I took an Uber to the hotel and rested for a while—not because I was tired, but because I needed to decide where to go for dinner.
The options were not great: an expensive yet mediocre restaurant in the hotel (or, more accurately, in the casino, since the conference was held there), some fast food nearby, and a few bars. In the end, I chose a burger place, and it turned out to be an excellent decision. The beef was juicy, tender, and full of flavor. It was the best burger I had ever had, and it gave me all my energy back.
Embarrassed
I embarrassed myself twice, which was not surprising given my lack of experience.
The first time was at the hotel. When I arrived, I told the receptionist that I had an appointment. I noticed a brief moment of hesitation on her face before she understood what I meant. I regretted my wording the second it left my mouth and felt embarrassed for a while. Fortunately, the staff was so kind and warm that I recovered quickly.
The second time was more consequential. I had not registered for the workshop because I did not realize that workshop registration was separate from registration for the main conference. As a result, I had nothing to do for the first two days.
In hindsight, that mistake was not entirely bad. I walked around Pittsburgh and explored the city a little. It felt energetic yet quiet, and I liked it a lot.
See? Interesting stories can come out of embarrassing moments. Minor accidents during a trip are not always a bad thing. Sometimes they make the experience more memorable.
Excited
Socializing is usually not my cup of tea, but being pushed into a social situation at a conference turned out to be exciting. People genuinely care about what you are working on, and if they happen to be in a similar area, they really want to understand your ideas in depth.
In the past, I often resisted social situations because I did not want to talk about things that did not interest me, or that felt empty and meaningless. A conference was different. I realized that there are many people who genuinely share my interests. That was the most exciting part. In an environment like that, there is no reason to avoid social opportunities.
That said, I still need to overcome my fear of starting conversations with strangers in English. I cannot say that I was fully satisfied with my performance, but I definitely did much better than merely surviving. I connected with Kirill, who is working on build system performance, and we discussed my recent work. I also asked Yann, an expert in formal verification, a question about refinement relations.
It is always amazing to talk with people who have spent years thinking deeply about the same kinds of problems. There is so much to learn from conversations like that.
Nervous
Although I had already given the same talk three times at school before ASPLOS, I still felt a little nervous. I went through my slides again on the evening before the presentation, which helped, but only a little.
I was the last speaker in the session, and my heart was racing throughout the entire session before my turn. Still, I knew I would be fine.
I know this pattern in myself well: as long as I can get the first sentence out, I gradually calm down. That was exactly what happened.
Q&A was probably the most challenging part, but it also went very well. No one challenged our work harshly. Instead, I received many questions about future directions, which I found both encouraging and inspiring. The audience really grasped the key point of our work.
I was not entirely satisfied with my answers, because I do not think they were very informative. I may even have sounded a little evasive at times. Still, the overall experience was good, and I genuinely enjoyed it.
Learning
I have to admit that most of the presentations were difficult for me to fully understand, which was only natural given how many different research areas were represented. However, the talks by senior professors and practitioners were often both accessible and insightful. I wrote down a few ideas that genuinely stayed with me.
Co-design
Co-design was the central theme of the first keynote, given by Partha Ranganathan, VP at Google working on system infrastructure. Although he mainly discussed hardware-software co-design, the broader lesson stayed with me.
It reminded me that we should not let ourselves be trapped by tunnel vision, even when we work primarily on software. It is respectable to focus deeply on one specific technical point, but that is not always the most effective way to solve real problems. Real systems are complicated. That complexity makes problems harder, but it also provides more possible levers for solving them.
Without co-design, more compromises have to be made, often at the cost of overall system performance. The talk reminded me to keep an open mind and pay attention to every layer that might matter, not just the one I happen to be working on directly.
Bottlenecks Move
One simple but powerful idea I heard was this: once you solve a bottleneck, that thing is no longer the bottleneck, but something upstream or downstream immediately becomes the new one.
I like this way of thinking because it is both practical and generative. It suggests that solving one problem is not the end of the story; it is often the best way to reveal the next important problem. For research, that feels like a useful instinct to cultivate.
Re-implementing Usually Fails
Another lesson that stayed with me came from the second keynote speaker, Professor Ion Stoica. His talk highlighted something easy to ignore in research: practicality matters, especially when you are trying to improve something incrementally rather than inventing something in a vacuum.
The cost of migration is often enormous. A technically better solution does not automatically win if it requires people to abandon an ecosystem, rewrite their workflows, or rebuild the surrounding infrastructure from scratch.
What I found especially striking was the importance of building on top of what already exists. Community matters. Adoption matters. Research does not become truly impactful just because it is elegant; it becomes impactful when other people can actually use it.
JSON Parsing
Besides the keynote talks, some paper presentations also triggered a few wild ideas in my head. One of them was about GPU-based JSON parsing.
The talk went through many design details that were needed to handle the ambiguous roles of characters such as quotation marks and commas. These characters are used both as structural syntax and, in some contexts, as part of string content. That ambiguity creates a lot of complexity.
At one point, I found myself wondering: why not completely separate these two use cases? Why not design a new JSON-like format in which structural characters can never appear inside string literals, and use some alternative escape scheme to represent them instead?
Then I immediately saw the obvious problem: this would mean re-implementing something that is already universal. And re-implementing widely adopted standards usually fails, no matter how elegant the new design may look on paper.
Still, I enjoy moments like that. Even when the idea is impractical, it is fun to follow the thought all the way through.
Neural-Symbolic Programming
Neural-symbolic programming was a new concept to me. The basic idea, as I understood it, is to use neural methods to extract facts—often with probabilities—and then use symbolic methods to reason over those facts and solve the final problem. Some recent work, including Lobster, combines differentiable Datalog with neural networks in this way.
What interested me most was the learning signal. In some sense, neural-symbolic methods feel a bit like reinforcement learning, except more structured and more targeted. In many cases, we are not searching blindly; we often have a fairly clear idea of which facts or predicates are relevant to the final decision.
That made me wonder whether the difficulty of symbolic search itself could be used as a learning signal. If some facts make the search much easier or harder, perhaps that information could be propagated back to help guide the neural part. I do not know whether this idea is new or practical, and I am definitely not an expert in this area. Still, I enjoy these moments of free-form thinking. They are part of what makes attending talks so rewarding.
Closing Thoughts
Looking back, ASPLOS was much more than a conference where I presented a paper and listened to talks. It was my first real experience of what an academic community feels like: the travel, the awkward moments, the nervousness before speaking, the excitement of finding people who care about similar problems, and the quiet mental aftertaste left by good ideas.
I did not understand everything, and I certainly did not do everything perfectly. But maybe that is exactly what a first conference should feel like. I arrived with curiosity and a bit of anxiety, and I left with more confidence, more questions, and a much clearer sense of why people find academic life so engaging.