Recents
These are some of my recent stuffs.
Rust Benchmarking, Jun 25 2025
Libraries
Criterion.rs
- Has really good defaults and seems to work out of the box.
- The ecosystem around criterion is really good. purely because criterion has been around for so long
Rust iai
Even though it is supported by a lot of ecosystem tools. It hasn’t been in development for the past 4 years. and bheisler is the creator of criterion, so even more reason to use criterion.
...click to readAI media Editor, Mar 29 2025
The first animation movie, toy story was made in 1995. At that time this was a break through. It involved a lot of proprietary software (Ray tracing, shaders, motion blur) and decades of research and development. In 2025, for the first time a movie made in open source software (blender) received the oscar.
...click to readThe Immortal Life of Henrietta Lacks, Mar 15 2025
HeLa Cells, as they’re called by the science community. Has been through it all, exposed to radiation, injected with virus, exposed in space, treated with chemical burn. All of modern medicine depends on HeLa cells to create the medicine. HeLa cells are what are used to test vaccines, chemicals, and physics to see hwo humans cells would work under certain conditions, HeLa cells are being manufactured around the world in troves.
...click to readServer Setup, Mar 7 2025
Oh my zsh
- oh-my-zsh has hands down been a godsend for me.
for that though we first need to install zsh
Mobile Dev, Feb 28 2025
Mobile development is probably the most highest abstracted operation possible. because of this there are too many pitfalls that trip up
...click to readSource Code, Feb 26 2025
Bill Gates starts this story from his roots up until the point he found Microsoft and it became a really serious thing. I like that the book closes right after the founding of Microsoft. Because much has been written about that part of the story
...click to readReact Native, Feb 22 2025
I’ve worked with React Native before but never professionally.
Recently I’ve been getting into building a complete react-native app
And, that is when you see the dark underside of things.
...click to readAlmond, Feb 15 2025
the book starts with a tragedy and ends with a mild happiness/realization.
the protagonist Yunjae is incapable of understanding or expressing emotions because of a problem in his brain.
...click to readShiptalk, Feb 15 2025
Building a product goes something like this.
You find a problem | | You’re annoyed enough by it | | You decide you’ll bring an end to the problem | | You toil over the problem so that it gets solved for everyone | | It becomes a great product.
...click to readVSCode and shenanigans, Feb 15 2025
Why switch from Intellij
Intellij was my most common IDE but I’m so tired of having to be a second citizen for every single plugin. Intellij needs to do a huge platform shift and become more open or be consumed by vscode. The amount of people who start with vscode is insane.
...click to readMotto, Feb 7 2025
Be intense about the task.
Don’t lose focus on the bigger goal.
Be Fine with being uncomfortable.
All good things take time.
...click to readwebsites that are interesting, Dec 17 2024
tools
Weirdnesss
https://cult-escape.com/cult-test/ A site to check if you are in a cult
Site https://www.languagetransfer.org/
Games in arcane places
interesting design websites
- Really good design stuff
- https://www.framer.com/
- https://spline.design/
- https://2019.makemepulse.com/
- JS Design generation
- Corporate Memphis Humaaans for every situation
Programming weirdness
ML
- LLM visualization, https://bbycroft.net/llm
- Model visualizer, https://netron.app/
- https://chat.lmsys.org/
Scholarly links
...click to readProduct Building and fundamentals, Dec 10 2024
Just like any other tech-pilled software engineer. I’ve spent my free hours and days working on side projects in the hope that someday to live Independent to the corporate day job and working on something I own full time.
...click to readWhy I Build., Nov 26 2024
I’ve spent all my free time building arcane projects that don’t matter.
Despite being tired, or stressed from my day job. I’ve always had projects that consume all my free time.
...click to readSuriya's Newsletter, Nov 24 2024
Do Not Go Gentle into That Good Night, Nov 15 2024
Do not go gentle into that good night,
Old age should burn and rave at close of day;
Rage, rage against the dying of the light.
Poetry, Nov 14 2024
Python package management is a nightmare. Poetry somewhat solves it.
new project
poetry new <project_name>
new shell
poetry shell
setting up
pipx first
...click to readThe Practice, Nov 6 2024
Shipping, because it doesn’t count if you don’t share it.
Creative, because you’re not a cog in a system
Work, because it is not a hobby.
...click to readWriting, Nov 5 2024
I write a lot. and I forget a lot.
Taking a lot of notes is part of how I stay sane and get even part of the things that I want to get done.
...click to readMadurai, Oct 25 2024
I lived in Madurai till my first grade. So It’s a really nostalgic town for me. Madurai is a city with a rich 700 years of heritage, and is built around the Meenakshi temple and it’s lore.
...click to readNewsletters and blogs that are interesting, Oct 16 2024
Pg’s Essays
- Incredibly interesting if you are into the startup world and are generally curious about how the world operates.
- I highly recommend The Submarine and The Bus Ticket Theory of Genius
Joel Spolsky’s blog
- Joel’s Essays are super good.
- I suggest people start with Strategy Letter 1, 2 and 3
Hardcore Software - Steven Sinofsky
- Steven writes, incredibly well thought essays on software and related things
- I was very engrossed with his launching windows series
Ribbon Farm - Venkatesh Rao
- Venkatesh Rao Writes really long, deep essays.
- I started with The Gervais Principle, or The Office According to “The Office”. Set aside a couple of hours before you start reading this. because it is really long and incredibly engrossing.
Sriram Krishnan
- I stumbled upon Sriram Krishnan’s blog by reading his memo collection. Took me a day to go through all the memo in his blog. Wish I was taking notes this rigorously back then
- I liked the recent group chat one as well
Nabeel Qureshi
- I signed up right after reading Reflections on Palantir. Spent a good 4 hours just reading random stuff
Shyam Sankar
- Palantir’s CTO, writes very poignant essays which I really enjoy
- I really loved The Case Against Work Life Balance
Thinking about things
- Really high quality articles, every alternative day.
- This used to be the only sponsored Newsletter I signed up for. Highly Highly recommend
Internal Tech Emails
- Internal Tech Emails tingles my gossip brain a lot. ngl, it has provided me with some of the most intricate scoops when talking with people.
Work Chronicles
...click to read
- Fun thing that I read almost everyday.
- Really love the dedication from the artist
Organizational Wisdom, Oct 16 2024
What started with Nabeel Qureshi’s Reflections on Palantir led me into an incredibly deep and rich rabbit hole of companies, intensity, operating them and, a whole host of new (to me) industry operators within the palantir circles such as, Tyler Cowen, Shyam Sankar, etc.
...click to readPittsburgh, Oct 16 2024
Pittsburgh has the quaint old town vibe with its own special antiquated ecosystems. It definitely has the feeling of a once flourising industrial town, now in its midlife crisis trying to find it’s place in the world. I stayed in Pittsburgh from Aug 2023 - Dec 2024. Made really close friends here.
...click to readAwesome incubators/collectives for startups, Oct 15 2024
Socratics
website: https://socratica.info twitter: https://x.com/socraticainfo
Open collective for builders, present in Multiple cities in the US and Canada. Really cool demos from their hackathon
...click to readPhotogrammetry/VideoGrammetry or the dark art of capturing 2d and generating 3d models, Oct 11 2024
Capturing
Questions:
Is it possible to go from, photo/video to high fidelity 3d model, with strong support for texture, without losing a lot of details on noise?
...click to readLeaky Abstractions, or How we Understand the world, Oct 2 2024
Languages are a lossy communication framework. You meant one thing, but your friend understood something else. The same words don’t mean the same thing even a few hundred miles apart, etc. It is wildly unfair to try and represent billions of neurons firing together, in sound waves only a few amplitudes higher or lower.
...click to readinertia, Sep 12 2024
Inertia is the best way to keep oneself active and consistent. Any sort of plan, trumps having intense velocity at the task.
...click to readAtlanta, Sep 6 2024
Atlanta is what happens if you put too many white collars in the same space. Downtown was very immaculate and the roads were slick, not to mention wide.
...click to readBangalore aka Bengaluru, Sep 6 2024
The earliest memory of Bangalore I remember is the LalBagh. And tangy Grilled chicken (For some odd reason). I was in awe of the city. The weather was perfect, the people were nice and the food was amazing.
...click to readPostmortem from building an AI speech coach, Sep 1 2024
The problem
International Students have a hard time understanding and integrating with the culture. This is true even if there’s a good grasp of the language.
...click to readLLMs and unstructured data, Aug 31 2024
Close on the heels of running a batch processing job over 1M documents, I was thinking about how LLMs are changing the way we process unstructured data.
...click to readSolutions vs Feelings, Aug 31 2024
A friend of yours is going through a tough breakup.
What should you do?
I typically start offering solutions and ranting on and on about strategies to attack the problem.
...click to readjournal, Aug 20 2024
Journaling is going to become an important thing.
...click to readTell us about the time you most successfully hacked some (non-computer) system to your advantage, Aug 20 2024
YC is looking for evidence that you are clever. The ability to think out of the box and manipulate large systems to your advantage are critical to gaining an edge on the market when you have no capital advantage (which is usually the case with startups) https://medium.com/@zan2434/y-combinator-applicant-advice-289c58a2ca89
...click to readUnix, Aug 19 2024
If you are on a unix-like environment, ports < 1024 (like 80) will require superuser privileges.
Disk mounting and unmounting
sudo mount -o discard,defaults <diskid> <dir>
you can get the disk id by
...click to readZSH, Aug 19 2024
Setting up new zsh with oh my zsh
firs installing zsh involves
brew install zsh
While I don’t really trust and run randomly downloaded shell files. This one is quite painless.
...click to readTravelling, Aug 18 2024
Library of Congress information, Aug 17 2024
Library of Congress is really an incredible repository of content. Just scrolling through library of congress is a fascinating mental exercise.
...click to readAlgorithms, Aug 8 2024
Elegant Algorhythms 🎶 are mesemerizing
...click to readGossip glomers, Aug 8 2024
I reused the functions provided at https://github.com/jepsen-io/maelstrom/tree/main/demo/go for interfaces.
repo: https://github.com/subygan/gg
1. Echo
This was a gentle intro to maelstrom and it’s tools. TBF the message passing using RPC was pretty interesting. Implementing the
...click to readread()
andwrite()
functions and then handing it off to the maelstrom Node made it work.ML journal, Aug 8 2024
09.27.23
- Tried ML model for face classification with,
self.backbone = torch.nn.Sequential( torch.nn.Conv2d(3,64,kernel_size=7,stride=4), torch.nn.Conv2d(64,128, kernel_size=3, stride=2), torch.nn.Conv2d(128,256,kernel_size=3, stride=2), torch.nn.Conv2d(256, 512,kernel_size=3, stride=2), torch.nn.Conv2d(512, 1024,kernel_size=3, stride=2), ) self.cls_layer = torch.nn.Sequential( torch.nn.Flatten(), torch.nn.Linear( 4096,num_classes), )
This model overfit spectacularly, training accuracy after 20 epochs was 95% validation accuracy was 15%. 🤯
...click to readTerminal, Aug 8 2024
Raw mode in a terminal is receiving the raw bytes from the
...click to readSTDIN
file. terminal usually gets opened in canonical mode which means input is fed to the program one line at a time, instead of character by character. This can be done by unsetting theICANON
flag. There are a lot of these bitflagsRight to be Forgotten, Aug 6 2024
This is a (mostly) search engine specific requirement where individuals are protected from the harm of their past actions.
Disparity in countries
Similar to any privacy issue, these countries are divided in the way these laws are implemented. Especially between developed countries such as United States and the European Union.
...click to readWord Embedding Models, Aug 4 2024
My aim for this blog was to keep it as not too technical as possible.
Word embeddings is a way to represent words as a vector. Which lets us do all kinds of trickery like how far a word like “King” is away from “Queen” and how they are related to each other with. Very useful when building applications with semantic understanding.
...click to readRead the Docs, Jul 23 2024
Failing to read the documentation and understanding the code underneath is the bane of all software bugs and I’ve guilty of it on multiple counts
...click to readShell hacks, Jul 23 2024
number of lines in a file
wc -l file.txt
split file into multiple files
split -l 1000 file.txt
append two files into one
...click to readcat file1.txt file2.txt > file3.txt
Privacy problems and attacks in Machine Learning, Jul 19 2024
Privacy problems and attacks in Machine Learning
- Membership inference attacks: These aim to determine if a particular data point was used to train a model. Researchers study how to detect and mitigate such attacks to protect individuals’ privacy.
- Model inversion attacks: These attempt to reconstruct training data from model outputs. Understanding these vulnerabilities helps in developing more robust privacy-preserving techniques.
- Data reconstruction in federated learning: This examines how malicious participants could potentially extract private information from other participants during collaborative training.
- Model extraction attacks: These involve extracting a model’s architecture or parameters, which could be used to launch further attacks or to steal intellectual property.
- Data Poisoning in Edge ML: This is a problem where the data that is being used to train the model is poisoned by an attacker. This is a huge problem in edge ML where the data is not curated and is coming from a variety of sources.
- Differential privacy techniques: Researchers explore how to add controlled noise to data or model outputs to preserve privacy while maintaining utility.
- Secure multi-party computation: This involves developing protocols for multiple parties to jointly compute on private data without revealing individual inputs. Homomorphic encryption: This allows computations on encrypted data, potentially enabling privacy-preserving machine learning on sensitive information. Privacy-preserving record linkage: Techniques to securely combine datasets from multiple sources without compromising individual privacy.
Papers and other readings
Membership Inference Attacks Against Machine Learning Models
...click to readMy Stack, Jul 15 2024
I used to deploy everything in an ec2 instance with docker-compose and nginx. Keeps it clean to apply small updates and also take backups. Where I’ll run different postgres containers for different containers, with some volume mapping and a backup container that triggers a backup push.
...click to read