Devakinandan

My fav book

Open to connect with product designers to further progress my development here beta.chitraagni.com

Readiness is not a feeling. It's a decision. The voice in your head that says you're not ready is not protecting you. It's just stopping you from finding out what you're actually capable of.

Anthropic onboarding day: Michael Scott introducing Karpathy like he just signed Wemby in free agency.

Visited @100xSchool and gave a talk to the next Super 30 batch Had a really good time sharing some things I learned from my Super30 experience and talking about what’s happening in Bangalore right now around AI and startups Also enjoyed spending time with @kirat_tw :)

Optimizing for time complexity at the cost of space complexity is easy. But when you are building an in-memory database like Redis, space is premium; so, if your peripheral data structures consume more RAM than the actual data, you have lost the battle. And, I just published a video where we dive deep into the Approximated LRU Algorithm used in Redis. This is the 15th video in the Redis Internals series. We look at why a classic LRU implementation fails at scale and how Redis implements a clever, memory-efficient alternative. A standard LRU algorithm requires a doubly linked list to shuffle keys to the head on every access. For millions of keys, the memory overhead of these forward and backward pointers is catastrophic. Plus, storing a standard 32-bit timestamp for every object eats up precious megabytes. Instead of perfect accuracy, Redis picks a counterintuitive route. In this video, we dissect the absolute genius of Redis's design choices. By the way, 15 videos are now live: 1. Why Single-Threaded Redis Is Fast 2. Writing a TCP Echo Server 3. Wire Protocols 4. Implementing RESP 5. Implementing PING 6. Understanding Event Loops 7. Implementing Event Loops 8. Implementing GET, SET, and TTL 9. Implementing DEL, EXPIRE, and Cleanup 10. Evictions and Implementing first-eviction 11. Implementing Command Pipelining 12. Implementing AOF Persistence 13. Objects, Encodings, and Implementing INCR 14. Implementing INFO and allkeys-random Eviction 15. The Approximated LRU Algorithm Hope this helps you better understand database internals and spark that engineering curiosity. Give it a watch.

The most important decision of your life

I was fortunate to learn coding in the pre-AI era and implement it during the peak AI era. From my perspective, I can share some insights. This will be a long post, so I suggest enabling reader mode. First, I want to talk about my college years and even before that, in 11th and 12th standard. The narrative at the time was simple. > Learn to code. > Get a job. > Secure a stable career. > Switch companies. > Build a net worth of a million dollars. That was the roadmap everyone was promised. I was also excited about it but before pursuing any field, I decided to try it first. Many already know this journey, so I will skip the details. You can read about it on my portfolio linked in my bio. The plan was straightforward. Get settled through engineering. Then the pandemic hit. We got GPT. Then Claude. Opus. Countless coding models. Agentic AI. Then came harnesses for coding, AI tools, and much more beneath the surface. Now the problem is that many people are entering the economy through coding using no-code tools and building systems. I am not denying that software engineering was difficult and now it has become comparatively easier, which is true. But the nuances of the field remain. A coder can create an app. An engineer can create an app. Even a less technical person can create an app. Previously, only someone proficient in software engineering or with some experience could build apps or websites. Those who did not know how to code could not create anything. That was the entry barrier. Now the barrier is shattered. So what remains? Those who do not know how to code, plus those with less knowledge of how code works and less experience, are trying to create meaningful software. This brings a fundamental issue. These people do not understand: > How software works internally > What tradeoffs they should make > How security should be handled > How to deploy to production Code was not the only hard part. The other hard parts were: > Tech Tradeoffs > Structure of code writing code > Security > Computation I have been through many legacy codebases and I can tell you that even AI does not even write the amount of slop code I have seen in some legacy codebases. Speed is not the problem. The problem is context. We are shipping at light speed and forgetting the context of how things happen. This has become one of the fundamental issues for security. Previously, many people worked on software and had enough time to think about: Tradeoffs Structure Security Many had read programming books and knew what to do and what to avoid. Now anyone can code and deploy to production but they do not know: How to preserve user data How to handle security How to debug a production bug when AI models are down Many will say we can hire someone, but the context you need to give to an AI, you also need to give to a human. Humans are not superhumans who can do many things with just one instruction. We have fundamentally made one step easier and many think the whole montain is super easy. No. If you want to climb the mountain, you need to: Learn to walk Then run Then climb Then ascend the mountain Hypothetically, that is engineering. About layoffs and companies not hiring due to AI, it also makes sense to me. During the pandemic, many companies overhired because cash was flowing from investors and SaaS and mini software were peaking. Everyone was working from home and technology was the only thing connecting everyone. Companies hired many times beyond capacity. Once the pandemic ended, funds hit the real world. Everyone realized they had overhired. They started laying off people. They did not want to devalue their stock by laying off because usually layoffs cause stocks to drop. But now that does not happen if they say they are restructuring because of AI. Internally, everyone knows it is not because of AI but because of the overhiring during the pandemic. These companies are trying to protect their stock value by naming it AI restructuring. We are currently in an AI bubble and the bubble will not break. Once a lion tastes blood, it can never be a vegetarian. The same happened with software engineering and with everyone around the world. AI is not going anywhere. If you think it will, you are really stuck. I would highly suggest you upskill yourself with the current stack. There is no stop to learning. You should learn: How code works How to write raw code What kind of code is good What kind is bad The entry to software engineering is harder than ever because low-quality apps and low-quality software are no longer good enough for entry. Previously, if you knew a language, you were qualified. Then it became harder. If you knew a language and could create some projects, you could be hired. Then it became: If you knew a language, could build a project, and knew a framework, you would be hired. Now it has become: > Do you know the language? > Can you create a project? > Do you know a framework? > Have you created a full stack? > Have you deployed to production? > Have you fixed security vulnerabilities? That is how hard it is nowadays to get a job. My take is that AI is going to stay here. Companies will be mass laying off people now and then and calling it AI restructuring. The entry barrier is shattered but if you are experienced and have gone through the depth of software engineering, no one is going to replace you. Actual engineers are not replaceable. There is still a need for good engineers everywhere. Good engineers are rare. You should be a good engineer and keep learning because you want to do it. If you are just doing it for money, you probably will only survive a couple of years. If you only want to be an engineer, then become an engineer. Do not become an engineer if you are just here for money. Since my 11th standard, I started tinkering with Arduino, low-level programming, Java, Kotlin, and building my apps. I spent four years of college doing the same and now I have a job that pays comparatively well. This is how long it took me to get to a position where I am confident enough that I can build software, fix any kind of vulnerability, and fix any kind of issue if I just spend enough time doing that. So the intake is this. If you are a fresher, do not be demotivated. Head down and work hard. You will be successful one day.

If you’ve been looking to properly get started building on @solana or level up your skills as a developer, there are currently a lot of active bootcamps, training cohorts,workshops, and builder programs happening across the ecosystem. Here are some of the major learning activities and programs currently available for developers: Solana Developer Bootcamp 2026 Official Solana Foundation learning path covering: Introduction to solana Local Installation Hello World AI Best Practices On-Chain Voting Escrow Application Private Transfers Stablecoin Stable Swap x402 Real-World Assets Security Checklist Indexing Prediction Market Production Readiness This is self paced with projects and video walkthroughs. youtube.com/watch?v=2pcm7I… Solana Developer Training Cohort 1 Live cohort based training focused on fundamentals and hands on program building. Dates: June 1 to June 5, 2026 luma.com/whfy8tz0 Solana Summer School Extended learning program for students and interns who want deeper ecosystem exposure and guided learning. Dates: June 15 to August 15, 2026 luma.com/g8b5fy9s Encode Club Solana Bootcamp 6 week structured learning experience starting June 1, 2026 covering Rust, Solana programs, AI assisted learning, and build challenges. encodeclub.com/programmes/sol…

PSA: I would be free from August mid till Nov mid (Around 4-5 months) I am open for work during this short time and would love to work on any roles open, in a remote setting I am in my 7th sem soon, and would have a CGPA of around 8.9/10 I am currently interning at Cisco and working on developing router software (mostly C and some py). In the past, I have worked with top researchers of this country (India) and have several publications (2 A* Findings, 3 A* Workshops, One A* in review atm ~ high chance to convert into publication). I have also helped build critical models which have been deployed to lacs of people in production. Open for anything, remote + good work environment + only for those 4-5 months I can/will do good work, that's guranteed RT + Share to amplify thanks :)

@harsh_twtt Cool

i am doing something of my own now yeah, i am officially jobless

Big corporations laying off thousands of workers to “save cost” AI turns out to cost more.

if you’re in tech, here’s a simple advice: there’s no investment better than building and growing a SaaS business in less time, you will add more net-worth than with any other investment the best part? everything you’ll learn while growing a business will help you in a compounded way, while also rewarding you at the same time go hustle a lil, take risks, see yourself grow like a 🚀

Why scroll 5 apps when you can spot the cheapest deal in seconds? Same product. Different prices. ₹338 on Zepto vs ₹389 on Instamart 🤤 That’s ₹51+ saved on just ONE item. Imagine your full cart. BuyHatke > switching B/W Zepto, Blinkit, Instamart & DMart

Millionaires are made every decade: 2000s - Internet The people building websites while others feared technology won. 2010s - Crypto The "crazy internet money" turned nobodies into billionaires. 2020s - AI One smart man with AI now outworks entire companies. 2030s - Attention The rarest skill will be keeping focus in a distracted world. 2040s - Longevity The rich will pay to slow aging while others destroy themselves willingly. 2050s - Energy & Water The men controlling basic resources will control nations. 2060s - Genetics Humans will start upgrading humans. Because apparently being insecure naturally won't enough. 2070s - Virtual Reality People will spend more time escaping life than living it. 2080s - Land Civilization will rediscover the oldest cheat code in history: own land. 2090s - Peace & Privacy Silence will become a luxury only the wealthy can afford. Every decade rewards the people who see the future early while the crowd calls them delusional.

Name the topic and I will share the best YouTube playlist to learn!

You’re 23, know only Python basics & hate LeetCode. Here’s the path I’d take: • Forget DSA guilt — focus on skills that get you hired • Build 5 real projects (APIs, automation, dashboards, mini-apps) • Learn Flask/FastAPI → deploy one project on AWS • Pick PostgreSQL → master CRUD + joins only • Learn Git/GitHub → clean commits + README = recruiter-friendly • Add automation scripts (file handling, web scraping, reports) • Build a portfolio page & upload everything • Share progress publicly for 30 days • Apply to internships + freelance + junior roles with your portfolio This path beats 200 LeetCode problems. Skills > theory. Projects > pressure.

2025: broke, depressed, nothing at all.... 2026: took 5th flight 😭 thank godd!!

DONT LOSE THE FXKING MOMENTUM resting on weekends is a propag*nda to break your flow > skip your kid's birthday > skip your wife's delivery > skip your parents' anniversary > skip your wife's boyfriend's birthday > skip your doctor's appointment > skip your best friend's funeral > skip your girlfriend's wedding

Anthropic pays $750,000+ a year for engineers who know how to build LLMs from scratch. Stanford just released the exact lecture that teaches it - 1 hour 44 minutes, free, straight from CS229. Bookmark and watch it this weekend. It'll teach you more about how ChatGPT & Claude actually work than most people at top AI companies learn in their entire careers.