Intro
Welcome to the first issue of Evgenii Pendragon newsletter! 👋
Thank you for subscribing to this newsletter! I'm excited to share with you things that I think and write about, educating content, as well as interesting reads that I find online. My hope is that this newsletter will provide you with loads of value and help you learn something new.
In this issue you will see how I hacked an online word guessing game, dive into LLMs drawing pelicans riding bicycles, watch how a musician trains a model to obfuscate his music to make it impossible for LLMs to train on it, and many other interesting finds.
If you have any comments, questions, or suggestions, please reply to this newsletter or reach out to me on LinkedIn. I would love to connect and hear back from you! 🤙
Hope you enjoy! 😉
Blog
"Hello, World" or "There is a new blog on the block" | Evgenii Pendragon
I decided to start a blog to write on topics that are interesting to me or something that I would like to explore. The topics in this blog should cover tech, software development, AI, cybersecurity, business, recent news, and occasional book reviews. Hopefully the learnings that will be discovered in this blog will be useful to many more people.
Value Judgement vs. Productivity in the Age of AI | Evgenii Pendragon
Like many other people in the industry, I became wary of the news that AI was going to replace us started circulating in software professionals’ circles. To understand the proposition of the AI overlords and compare it to the reality, we need to examine what AI is truly good at and what it is at best subpar at.
Reverse Engineering Midword.com | Evgenii Pendragon
MidWord.com is a game where you guess a word from a dictionary developed by Mina Naguib. Whenever you make a guess, the program tells you if your guess is before or after the secret word. After many attempts you narrow it down to find the word that is hidden. After playing this game for a bit, I reverse engineered its secret word generation algorithm and came up with a one-liner JS exploit to get the word instantly.
Articles
The last six months in LLMs, illustrated by pelicans on bicycles
In his article, Simon Willison explores a benchmark for LLMs that he has created - requesting a model to create an SVG of a pelican riding a bicycle. This ingenious request demonstrates how different models fare against each other. The results that some of these models produced were pretty hilarious.
Bruteforcing the phone number of any Google user
"From rate limits to no limits: How IPv6's massive address space and a crafty botguard bypass left every Google user's phone number vulnerable". In this article, the author explores a vulnerability that allowed a brute force approach to work on one of Google's services. It is fascinating to see that some old tricks still occasionally work.
Turning Feedback Into Features: Building My npx Business Card
This type of project is not something that comes to mind when you want to do something small and simple. Yet creating a small business card via npx is small enough to be doable even for beginners while also creating value by sharing it with others on your socials.
Videos
'Forbidden' AI Technique - Computerphile - YouTube
It is amazing to see how new research is following the latest developments in AI. One such research noticed that in chain-of-reasoning when motivated to get to the answer "no matter the cost", the model started to omit the reasoning that it would estimate might get it in "trouble". A very interesting dive.
The Art Of Poison-Pilling Music Files - YouTube
One of the biggest issues being discussed these days is the issue of AI models scrapping every last bit of the Internet that it can. Benn Jordan found a way to "poison" the music files in a way that is unrecognizable to human ear, which makes AI produce a very faulty output. This is one of the great applications that I have seen in an attempt to protect intellectual property from being plundered without the creator's consent.
The Only Video You Need to Get Started with Neovim - YouTube
I have tried using Vim and NeoVim in the past a couple of times and just got frustrated with the amount of configurations and base knowledge for entry into that space. That changed once I gave kickstart.nvim a chance. In this video, TJ DeVries walks through the kickstart configuration and explains how it all works. This gave me the most confidence and helped me dive back into NeoVim.
Resources
Scrappy: make little apps for you and your friends
I found this amazing resource to create small one-purpose apps by accident. It instantly caught my attention. The idea that a super simple app development should be affordable, easy to learn, and understandable for most people is straightforward and relevant today. It is encouraging to see small tools/frameworks pop-up to make app development fun and accessible.
ossu/computer-science: 🎓 Path to a free self-taught education in Computer Science!
I discovered this amazing resource on GitHub recently - Open Source Society University. It provides everyone access to a set of courses and materials like they would be offered in a regular college for those interested pursuing an education in Computer Science. All of it for free. All of it curated by the community. Just an amazing testament to the power of the Open Source. And the fact that it can be applied to things outside of software.
Daily Jailbreak
There is this impressive daily contest platform where users can submit the most effective jailbreak prompts for AI models to make them behave in a way they are not supposed to. It is a great way to learn about prompt injections by example of other people. This area where AI intersects with security will only continue to grow as time goes by.