Forem: Asmit Phuyal

How to Get Selected for GSoC (Google Summer of Code) - My Personal Experience at Accord Project

Asmit Phuyal — Sun, 14 Sep 2025 14:19:51 +0000

How to Get Selected for GSoC (Google Summer of Code) - My Personal Experience at Accord Project

Recently GSoC 2025 has ended, and I’ve successfully passed this. I thought, why not share my personal experience so that you can also crack GSoC.

By the end of this article, you’ll get to know what actually is GSoC, how to crack it, how many attempts you can make, and what happens if you crack GSoC.

How it all started

GSoC and open source started for me when I learned more about open source at DevFest 2024, probably towards the end of the year.

Until then, I only had a rough idea about open source and GSoC, but that day opened my mind a bit.

The next day, I went and contributed to the website of GNOME Nepal, the same org I had heard about at DevFest from Aditya Singh (founder of GNOME Nepal). My first pull request got merged there, and that actually got me excited.

A late start

It was late compared to others preparing for GSoC. It was already the start of 2025, and I was still at the beginning level while many were way ahead. For me, it was not the first start, but a fast and smart start.

Honestly, I had no hope I would get into GSoC this year. I just started exploring open source more deeply, without thinking of GSoC 2025. Some YouTube videos divide GSoC orgs into two categories:

very competitive and large orgs that come every year
smaller orgs that recently started coming with comparatively less competition

(Remember this point, I’ll come back to it later.)

At that time, I wasn’t sure which org I’d target for GSoC 2026, but I had started exploring random projects/repos to get a basic idea.

Before and after org announcements

I want to break my journey into two parts: before the GSoC 2025 org announcement and after it.

Before the announcement, I had contributed to a few projects but none of them were targeted GSoC orgs. I had no expectation of getting into GSoC, and I wasn’t sure which org would align with my interests.

Everything changed when the GSoC 2025 organizations were announced in the last week of February. After a few days, the enthusiasm to get into GSoC sparked inside me.

I picked a few orgs, mostly newer ones that were not very repetitive. Then I narrowed down to two, thinking I could submit proposals to both and maybe get lucky haha.

But finally, I ended up sticking with a single org where I had a real chance: Accord Project. Out of all the orgs, Accord Project fit my technical expertise and interests the best.

Deciding on Accord Project

By the first week of March, I had finalized Accord Project for GSoC 2025. That gave me only about a month to show my skills to the maintainers.

I started exploring whatever I could about Accord Project. I had already decided which project idea to go for. I opened a few issues, got some PRs merged, and had a few open. Days went on like this, and then it was time for proposal submission.

I had seen many people with lots of merged PRs, since they had started preparing 4–5 months earlier. But for me, it didn’t matter much. I think it’s all about quality over quantity: the quality of the proposal and how you plan to handle the main project idea.

Proposal writing and exams

The proposal-writing period overlapped with my 4th semester board exams. We had a 3-day gap between exams, so I spent around 2 days writing my proposal, contributing, and exploring the codebase more.

I submitted the proposal on April 8, the last date. Even after submission, I kept exploring the codebase because my idea was to learn more, no matter the result.

Result day

I honestly had very little hope of getting selected, as I wasn’t an early starter. I thought I’d definitely get into GSoC 2026, but not this year.

Then came May 8, the result day. A few days before the official announcement, some Reddit users noticed frontend changes in their GSoC dashboard. Their “Contributor” section was updated before the results. I checked mine, and yes, the same change appeared.

I got excited, but wasn’t 100% sure.

The official announcement was scheduled for 11:45 PM NPT. I usually sleep early, but that night I couldn’t. Others who had the dashboard change had already received their mails, so I was just waiting.

Finally, at around 12:30 AM, I got the mail. I got sooooooo happy. I can’t even express it.

I was sleeping with my 10-year-old brother, so I couldn’t wake him or my family up at midnight. I just stayed excited alone. I thought of sleeping and sharing the news in the morning, but lol, I couldn’t even sleep. I finally slept around 2:00 AM.

The next morning, I shared it with my family, close ones, and my friend Darshan Poudel (the biggest open source enthusiast in my class). Then I posted on LinkedIn and got many congratulations from friends, relatives, and more.

Representing Nepal in GSoC

That’s how I was able to represent Nepal among the 68 countries in GSoC 2025.

Some quick stats:

98,000+ registered contributors
15,240 contributors submitted 23,559 proposals
1,272 proposals were accepted

For Accord Project, about 30.5% of the proposals were for the same project I chose. The org received around 200 proposals overall.

After selection

From May 8 to June 1 was the community bonding period. I met my mentors and co-contributors, explored the codebase deeply, and felt a sense of responsibility.

From June 2, the actual coding period started. If you’re curious about my project (a JSON ↔ YAML converter for DCS), I’ve already written about it here:

Accord Project – GSoC 2025 JSON to YAML Bidirectional Converter

FAQs

What is GSoC?

Google Summer of Code (GSoC) is not an internship, but a Google-led program that promotes open source contributions worldwide.

Is there a stipend?

Yes. It depends on your country, but the base rates are:

Small project: $750
Medium project: $1500
Large project: $3000 Mine was a medium one, which is the most common.

Which org should you choose for GSoC?

Pick one that aligns with your interests and skills. Don’t just chase the biggest names.

What happens if you crack GSoC?

You become part of a global network of contributors, gain mentorship, and improve your open source credibility. It can also open doors for future opportunities, jobs, and collaborations.

That’s it. I don’t know who or when you’re reading this, but if I can be of any help, I’m always ready.

Feel free to contact me if you need.

Thank you, and good luck for your next GSoC.

Model Collapse: When AI learns from AI

Asmit Phuyal — Sun, 14 Sep 2025 11:05:22 +0000

Model Collapse: When AI Learns from AI

Let’s imagine a line of people playing the telephone game. The last person, labeled F, whispers a message to E. E whispers what she heard to D, and the process continues till the message reaches A.

By the time A receives the message, it will be totally different from what F originally intended. There will be lots of distortions and inaccuracies.

Likewise is the case for AI model training. If synthetic data (AI-generated content) is used to train the next model, and then that new synthetic data is again used to train another model, the final model tends to produce more homogenous output — more error-prone, less useful, less diverse, and less accurate.

Let’s get deeper into it.

You are probably familiar with the importance of diversity in ecosystems. The same applies to AI training data — otherwise, models risk collapse.

With the rapid rise of LLMs, the internet is increasingly being flooded with AI-generated content. As LLMs are trained heavily on data scraped from the internet, future training datasets will contain AI-generated data as inputs.

Organizations working on LLMs value human-generated data, which could become harder to find over time. I’ve heard ideas about making AI-generated data easily identifiable so that future systems can distinguish synthetic data from real. I mean some kind of human-undetectable but machine-detectable information embedded within AI-generated outputs.

But here’s the catch — there are already tools that make AI-generated content appear more human. If that AI-generated data is again used to train models, what happens? A deadlock? An infinite feedback loop? Irreversible defects?

Here’s what I think happens if a model collapses:

Poor decisions for rare cases; models will struggle with edge-case scenarios.
AI-generated outputs are expected to be unique, but users may be disappointed by repetitive or similar responses.
Amplified bias in certain rare topics; models rely more on dominant patterns.

Model collapse is a significant challenge for the future development of robust and reliable AI. If we don’t pay attention now, future AIs might lose the very thing that made them powerful — their connection to human experience.

LaTeX for Beginners in 6 Minutes

Asmit Phuyal — Wed, 15 Jan 2025 14:16:44 +0000

LaTeX for Beginners: Learn in Only 7 Minutes

Master the 20% of LaTeX that powers 80% of your documents.

Whether you are a student, researcher, professionals, or someone who recently came across LaTeX and thought of using it but feel confused about how and where to begin, this short guide will help you understand the basics.

At the end of this tutorial, you'll be building a simple resume.

As a bonus, I will provide you with the template that we create in this tutorial for reference.

Why LaTeX and why not MS Word?

Before understanding LaTeX, let me begin with why LaTeX and why not traditional word processors like MS Word.

When Donald Knuth, a computer scientist and mathematician, was writing his book The Art of Computer Programming, he became unhappy with the poor-quality typesetting. To solve this, he developed the TeX typesetting system, on top of which LaTeX was later developed.

Reading this, you might already have figured out when LaTeX is useful, right? It's for those who usually need more than just plain text in their documents - like mathematical expressions, symbols, tables, and technical formatting.

Still wondering, "These things can also be done in word processors like MS Word, so why use LaTeX?"

The key difference is how LaTeX and MS Word handle documents:

Word processors like MS Word focus on appearance and what you see on the screen, requiring you to manually adjust the formatting for equations and tables. This manual formatting can be time consuming and inconsistent.
In contrast, LaTeX focuses on content. Instead of worrying about how your document looks while you're typing, you define the structure and formatting of your document with simple markup commands, which LaTeX then uses to create a formatted document with consistent styling.

LaTeX is a powerful and efficient tool for creating technical documents, offering cleaner and more consistent results than word processors.

Now excited to get started with LaTeX? Let's get into it.

LaTeX Hands-On Guide

For now, we will use online LaTeX editor Overleaf, which is so easy to use. No installation is needed, just a simple login and you're good to go.

Log in to Overleaf.
Click Create a New Project and select Blank Project.

You'll see a code editor. It might look overwhelming at first, but don't worry - just delete the existing code and paste the following:

\documentclass{article}
\usepackage{graphicx} % Required for inserting images

\begin{document}

\end{document}

Now, lets start creating a resume by understanding what these commands actually do.

Note: You need to click the Recompile button each time you make changes to the code, unless you enable auto compile from the dropdown next to this button.

Document Structure:

Every LaTeX file must include the following commands, which define where the actual content of your document begins and ends:

\begin{document}
% Contents here...
\end{document}

Usually we write the command \documentclass{article} at the top of our file.

article is used for articles, short reports, and small papers.
It can be replaced with some other document class options, like report (for longer documents with chapters), book (for books), or beamer(for presentations) depending upon context. For now, we will leave it as article.

If you want to use any additional packages, they are usually placed right below the \documentclass{...} command, in the form of \usepackage{package_name}.
We'll leave the graphicx package (used for inserting images) as is for now, since we'll be adding a cute photo of yours to the resume later.

Basic Commands:

Lets understand a few more commands, which are widely used every time you're working with LaTeX. First, copy the following code and paste it just below the \begin{document} command and we'll see it line by line:

% Name and Contact Info
\begin{center}
    {\LARGE \textbf{Your Name}} \\
    \vspace{0.2cm}
    Email: mail.me@example.com | Phone: +123 456 7890
\end{center}

% you see in first line is how we add comment. Every texts starting with % will be ignored during compilation. For example, Name and Contact Info is a comment that helps us identify the purpose of the section in the code.
\begin{…} and \end{…} are used to mark the start and end of environments. Environment is nothing, but a block of code that is responsible for a specific formatting to the content inside it. Here, we have used center environment, so text inside the \begin{center} and \end{center} will be centered.
We used \LARGE command to make the name appear larger. LaTeX allows us to control the size of text using commands like \tiny, \small, \huge, \Large, \LARGE, etc. You can explore these by replacing \LARGE with any of the other size commands to adjust the text size.
For text formatting, we have used \textbf{Your Name}. We can also use others like \textit{italic text}, \underline{underlined text}, which you can easily guess what these commands are doing.
\\, \vspace{…}, \hspace{…} are widely used commands for spacing. \\ just creates a line break i.e. text following it will start on a new line. \vspace{…} and \hspace{…} as name suggests, are used to add extra vertical space and horizontal space, respectively. For example, \vspace{0.2cm} adds 0.2 cm of vertical space.

Clearly you can see, we have combined two commands; \LARGE and \textbf{Your Name} as {\LARGE \textbf{Your Name}}, to create a larger and bold version of "Your Name".

Adding Photos:

Like before, copy the following lines of code just below the name and contact info section we discussed earlier.

% Photo Section
\begin{center}
    \includegraphics[width=0.2\textwidth]{profile_photo.png}
    \vspace{0.5cm}
\end{center}

Now, you almost know what's happening in this code, except for the \includegraphics{…} part. Earlier, I mentioned that we'd add a cute photo of yours. Well, this is the step where we do that. See at the line no. 2 of your code we have used graphicx package, now we're implementing it.

You can see upload button at the top left. Just click there, upload a photo and then rename it to profile_photo.png.

The command \includegraphics{profile_photo.png} tells LaTeX to include an image with file name profile_photo.png but we need to make sure the image file is in same folder as of main.tex file. Since we're using Overleaf, uploading it as mentioned will automatically put the file in the correct location.

The width=0.2\textwidth option adjusts the image size, making the width 20% of the text width on the page. You can explore different sizes by changing 0.2 to any other value of your choice.

Sections and Sub-Sections:

Copy and paste the code below, and let's see what's happening.

% Objective Section
\section*{Objective}
Write a short statement about your career goals.

% Education Section
\section*{Education}
\textbf{Degree Name}, Institution Name \hfill Year \\
Field of Study

% Work Experience Section
\section*{Work Experience}
\textbf{Job Title}, Company Name \hfill Year -- Year \\
Brief description of your role and achievements.

Now you've been introduced with new concept: section. In LaTeX, sections help structure our document into logical parts, making it easier to read and navigate. Lets understand more on how we can create sections in our document.

The \section{...} command is used to create a main section in our document. Here, we've used it to create sections like "Objective," "Education," and "Work Experience." By default, sections are numbered automatically (e.g. 1, 2, 3…) but adding an asterik (*) as shown in code above, will remove the numbering.
Likewise, \subsection{...} command is used to create a subsection under the main section. Basically, these are second-level section under the main section which are numbered hiererchially. And yeah, to stop auto numbering we can use * as stated earlier.
Now you can easily guess what \subsubsection{} command does. It creates a third-level section under a subsection.

Additionally, notice the \hfill command in the code. This command inserts a blank space that stretches to fill the available space, which is useful for aligning text, such as dates and institution names.

List and Tables:

We're almost at the end of learning LaTeX. Let's wrap things up with two more essential elements: lists and tables. As earlier, copy the following code and paste it below work experience section.

% Skills Section
\section*{Skills}
\begin{itemize}
    \item Skill 1
    \item Skill 2
    \item Skill 3
\end{itemize}

\begin{itemize} command is usually used to make a bullet list environment. Each bullet point is created using the \item command.
If we want numbered lists instead of bullets, we can simply use \begin{enumerate} command i.e replace itemize with enumerate.

We will not be using tables here in this resume, but I want you explore table in LaTeX. You may take the following code as reference:

\begin{tabular}{|c|c|}   \hline
Column 1 & Column 2 \\
\hline   Data 1 & Data 2 \\
\hline   \end{tabular}

\begin{tabular}{|c|c|}: Creates a table with two centered columns, separated by vertical lines.
\hline: Adds horizontal lines.
\\: Moves to the next row.

Final Touch: Removing Page Numbers

Now, you may see the page number at the bottom of our document. To remove this, simply add \pagestyle{empty} command before \begin{document} command.

Congratulations! You've created a simple resume using LaTeX. As promised, here's the template we created: Click Here

Feel free to explore more and make it your own. Happy LaTeX-ing!