# What I learned publishing a paper > Source: > Published: 2026-07-12 07:02:29+00:00 # What I learned publishing a paper — [musings](/tags/musings/) — 5 min read In August 2024 I enrolled in a Master's degree in Artificial Intelligence (AI) at the University of Edinburgh with no prior research experience, and in April 2026 I published my first research paper at a reputed conference hosted in the US, studying how AI models learn languages. 1 In this post I want to share the story of how I did it. This is not a post on how to write a good research paper; instead I want to share some ways of thinking that helped me achieve something difficult that mattered to me. Below I've illustrated a timeline of events that occured between enrolling in college and publishing. At each point plotted, I describe what I was doing based on notes that I took at the time and on the y-axis, I've represented 'visibility', which captures how clearly I felt I could see my actions directly leading to achieving the end goal of publishing. **Aug-Dec 2024.** I enrolled in the program and took a natural language processing (NLP) class in my first semester. I also wrote my first literature review and learned research basics like how to structure research notes and use software to organise papers (shout-out Zotero). I was enthusiastic, and doing well in my classes seemed like a reasonable milestone. **Jan–Feb 2025.** For my summer dissertation requirement, I applied to work on a topic with my NLP professor since I'd enjoyed her class and found her work interesting. Securing a good dissertation project was the next milestone. **March-April 2025.** I got allocated her project, and spent March and April turning a rough dissertation topic into a full proposal alongside her and a PhD student advisor. I focused on trying to bring something meaningful to every weekly update meeting, and learned how to read and deconstruct a paper. My proposal was well received, and I was pumped for the work ahead. **Summer 2025.** From May to August I did the groundwork for my research: forming hypotheses, running experiments, and reading papers, figuring much of it out as I went, under the guidance of my professor and PhD student advisor. The next milestone was to write and submit a great dissertation that I would be proud of. **June 2025.** A visiting professor, who was a senior author of the paper my study was based on, was in Edinburgh to give a talk. I attended it and pitched a one-on-one meeting with them to share my work. They received it well and liked my ideas, which boosted my conviction that this work was atleast somewhat novel, though I still didn't know if it was paper-worthy. **Late Aug 2025.** I submitted my dissertation, happy with it after a marathon final day of polishing arguments. When I asked my professor if it could be a paper, she gave the fair answer that it had potential but needed far more work and guidance than the summer had allowed. Because of everyone's busy timelines and my own uncertainty regarding what I was doing after graduation, it was unclear to me when and whether this would actually happen. **Sept 2025.** With the dissertation submitted and the MSc program completed, I didn't know what my next step was. I decided to keep reading the literature and improve my understanding of what my results meant in the context of the field. At this time, I also moved back to Chennai, my hometown in India. **Oct 2025.** I remembered the visiting professor's interest in my work, and shot them an email of my completed dissertation; they liked it, and I asked if they'd be interested in extending this work together. I drafted a paper with a plan to improve it and sent them an email, CC'ing my professor and PhD student advisor-turned-collaborator. After some discussion, we decided that it was best to continue this work without the involvement of the visiting professor, due to this project being an extension of my dissertation. **Nov 2025.** Back in Edinburgh for my graduation ceremony, I met with my professor and PhD student collaborator to discuss our next steps. We aligned on the framing of the paper and follow-up experiments to conduct once I was back from an upcoming holiday. It finally felt like we were moving forward again, and I was getting closer to the goal. **Jan 2026.** In January, I ran more experiments to round out our existing results and refined drafts of the paper with the professor and PhD student collaborator. Later in the month, we settled on a target conference with a February deadline and for the first time in the whole journey, I could see the finish line. **Feb 2026.** February was a intense push — sharpening arguments, plotting figures, and assembling everything in our LaTeX template. My collaborators made invaluable contributions and I learned an enormous amount in a short time. Watching it come together and submitting the paper felt surreal. **Apr 2026.** In mid-April I received email confirmation of the paper'sacceptance, with great reviews. Later we learned it had also been awarded best paper. When I began my degree, I was eager to learn and push myself towards mastery. Publishing a paper seemed both challenging and cool in ways that mattered to me, so I decided to try to achieve this hard thing. Hard things are goals that have a beautiful capacity to simultaneously excite and scare you, and what is a hard thing for me may not be one for you and vice versa. Another defining characteristic is that in order to achieve them, you have to become a different person in the process of achieving it. That's because they are zero-to-one type goals, where you are trying to do something that you've never done before. This process of changing your identity can be messy and uncomfortable because of the questions you have to ask yourself, especially if you are an emotional person like I am. It involves both new kinds of decisions, and making different choices to ones you've encountered many times before. This is the tricky, underappreciated aspect of going after these types of goals, and the part that caught me most by surprise. Looking back, I found two techniques that were critical to navigating this. **The first technique is to hold an unwavering belief that you can do it.** The trick here is to be fully commited to the idea that there is a version of you in the future that has achieved the thing that you are working towards. This may sound counter intuitive because you haven't actually done it yet, but belief is a very powerful motivator, and even just knowing that it's possible for **you** can push you over the line (I think of it like using your future self to induce the Bannister effect ). This belief is what gave me the energy to persist and 'find a way' at many points even when it wasn't clear what to do, as illustrated in the graph above. [2](#user-content-fn-2)**Belief by itself doesn't get you anywhere, so the second technique is translating that belief into the right actions.** This undoutebly involves your own hard work, perseverance, and help from others; they are the 'domain-specific' things that have to go right for your own hard thing. In my case, some examples are getting better at reading papers, designing experiments, and academic writing. The trick here is to focus on the right *lead* indicators, which are meaningful ways to measure your effort towards the goal. A simple one I used was the amount of time spent in a day working on the project. 3 Relatedly, a good way to know if your daily actions are directionally appropriate is to ask yourself from time to time, "Is it unreasonable to think that someone who does these things every day for a long period of time will eventually achieve the goal?". If the answer is yes too often, it is a good sign to change things up. Being less prescriptive in my day-to-day activities complemented my style of working, but your mileage may vary. Importantly, it forced me to shift my idea of progress from achieving the big thing to achieving little things. Each individual day started to be measured not by whether the goal was achieved, but by whether I did my work for the day as intended. This style of working also helps build momentum, which is a powerful asset. After a while, the goal starts to feel almost 'inevitable' even while the actions themselves are changing, and this is when you know you are on the 'right' path. I would be remiss to not mention some components that are out of your hands: time, luck and people. Time captures how long you are able and willing to pursue the goal, and the longer the time frame, the increased surface area for luck to strike. As you can't directly control luck, it's important to have an open conversation with yourself about how long you give yourself to pursue your goal. This isn't always necessary, but it can help adjust your expectations, especially since hard things take a longer time than you might initially expect. I also had amazing co-authors whom I had access to via the institution I was a part of, and five years of work experience prior to enrolling in the MSc degree. These were some of my unfair unadvantages, and it is worth pointing them out and the privilege they afforded me. Keeping that in mind, it's important to identify your own unfair advantages and make the best use of them in pursuit of your hard things. If you liked this reading this post, let me know [here] ## Footnotes - The paper is titled "When transformers learn "impossible" languages, what do they learn?" ( [link](https://openreview.net/forum?id=zllTuNWQti#discussion)). It won "Best Paper Award" at the Conference on Computational Natural Language Learning (CoNLL) 2026.[↩](#user-content-fnref-1) - Once the first man completed a sub 4-minute mile that had never been done before, 24 men did it within a year. This psychological effect has come to be known as the [Bannister Effect](https://learningleader.com/bannister/), seen also in[mountaineering](https://www.nimsdai.com/project-possible)and weightlifting. Update:[Here](https://www.scienceofrunning.com/2017/05/the-roger-bannister-effect-the-myth-of-the-psychological-breakthrough.html?v=47e5dceea252)is an article that argues instead that this effect is a myth. While it rightly clarifies the 'neat' narrative commonly used when citing this effect, I think it doesn't exhaustively negate the existence of the effect itself.[↩](#user-content-fnref-2) - I practiced a style of pomodoro working to maintain focus, using this app: [https://www.flow.app](https://www.flow.app)(I love them, and I'm not being sponsored to say it).[↩](#user-content-fnref-3)