Priyanshu

I poured my 10 years of teaching experience into a skill. It's called /teach, and it can teach you anything. Here's how it taught me to solve a Rubik's cube:

Ex-Meta data scientist Tina Huang just shared her entire Claude Cowork setup.🤯 Most people are guessing their way through Cowork. She showed exactly how she runs it daily in 15 minutes. 15 minutes here saves you weeks of guessing. Watch it once. Set it up right. 👇

@SharmaShradha @Java__sports @TrinaMu84309226 @Java__sports Do you have your own website or any online presence I am a Full-Stack developer and Can take you website from building to production or can maintain one also If need any help Please connect. How more can I help you ?

Reverse migration in Bihar. ❤️💪 One manufacturing startup in Patna has shown the way. @Java__sports , co founded by Trina and Nishant - they said we can build the world class sports products right here from Bihar. I am supporting them with an angel cheque. Join me in investing, collaborating and supporting them to be the biggest “made in Bihar” sports brand. Do visit them. Watch this 👇

Andrej Karpathy spent 2h showing how he actually uses AI day to day he's a co-founder of OpenAI and led AI at Tesla, so when he shows how he works, it’s worth watching and the whole session is just him telling the machine what he wants in simple terms, like he's briefing a coworker watch what's actually happening the entire time: > he describes the task in normal words > it goes off and does the work > he glances at the result and nudges it with one more sentence that's the whole skill, and you've had it since you learned to talk the only gap between that and a worker that runs on its own is handing that sentence a schedule and the tools to act check his work, then build the version that keeps working when you stop

Everyone can use @elonmusk's "Magic Wand Number" and "Idiot Index" They're universal ideas, helpful in any industry.

jia bin dropping the best lectures

@arpit_bhayani @arpit_bhayani and If right now I am unable to pay such heavy amount then is there any chance I can learn because its 50000 my two months salary. So how I can be one the learner Please do guide

My first Applied AI cohort kicks off in 2 hours; excited and nervous. To be honest, I had this feeling 5 years ago when I started my System Design cohort :) By the way, it is a 3-week cohort (6 sessions), super hands-on with 24 prototypes and demos, 7 real systems, and filled with production stories. brainstorming, and probing. Most of the stuff that I am going to cover comes from my actual experience building AI systems and, of course, spending a ton of time understanding them through and through (not the internals, but the applied side of things). Hence, it is not an ML or DL course. There is no model training or fine-tuning. The focus is entirely on applied AI: how do you take LLMs and build reliable, production-grade systems with them? Memory architecture, tool use, multi-agent coordination, evaluation, prompt failure modes, cost attribution, and more. Almost every single concept and pattern is backed by working code. By the way, if you are interested, the 2nd cohort starts on July 4th, and enrollments are now open. Sessions happen on Saturdays and Sundays from 8:00 pm to 11:00 pm IST, making it easy for folks from the west to join as well. Also, the early bird discount is on, and the curriculum and other details are linked below.

This article is literally wow. i read it 2 years ago, and coming back to it today, it still feels new. few tutorials teach computers in a way that permanently changes how you think. this is one of them. If you've never built a VM before, you're missing one of the biggest "aha" moments in computer science.

As a Backend Engineer, make sure you have a strong understanding of: • SOLID Principles • Multithreading & Concurrency • Immutability Concepts • Streaming & Message Queues • Caching Strategies • Security Fundamentals (SSL, JWT, OAuth) • Design Patterns (Factory, Decorator, Singleton, Observer) • Test-Driven Development (TDD) These aren't just interview topics—they’re fundamentals that help you build scalable, maintainable, and production-ready systems. 🚀

idk about others but the last two interviews I gave ... the questions were simple related to ai + infra > my opinions on long horizon eval like how will i evaluate a agent that is running 20+ steps to complete a task > agent architecture on prod level ... like a prod ready agent architecture and how do I stop a infinite loops

curious: what are some AI system design questions that companies are asking?

Daksha is learning. Powered by Viveka Decision Core and an indigenous VLA architecture, Daksha observes, understands, and executes real-world tasks while improving with every interaction. Building the future of embodied AI and general-purpose humanoid robots @ihubrobotics

Training an LLM from scratch is easier to study when the whole path is in one repo. Train LLM From Scratch is a PyTorch repository for learning how a transformer language model is built, trained, saved, and used for text generation. It helps you move from “I understand attention on paper” to a runnable training pipeline by pairing model code with data download, preprocessing, config, training, and generation scripts. Key features: • Transformer components from scratch – separate PyTorch modules for MLP, attention, transformer blocks, and the final model • Pile-based data path – scripts download The Pile files and preprocess JSONL.ZST text into tokenized HDF5 datasets • Configurable training setup – model size, context length, heads, blocks, batch size, learning rate, and file paths live in config.py • Hardware guidance – README compares common GPUs for 13M and 2B-class training runs • Generation workflow included – generate_text.py loads trained checkpoints and produces sample text outputs It’s open-source (MIT license). Link in the reply 👇

Fun fact: Redis does not just rely on the standard operating system memory allocator. Even though it's written in C, it overrides the native `malloc` implementation with highly optimized third-party memory allocators like `jemalloc` (by Facebook) or `tcmalloc` (by Google) to achieve efficiency and low-fragmentation performance. I firmly believe that understanding theory is fine, but things become interesting when you sit down to implement it. This is where you find gaps in your understanding and understand the nuances. and hence... Today, we dive into the source code of Redis to look at how and why Redis overrides the default `malloc` implementation. This is the 18th video in the Redis Internals series. Like always, we keep our focus on execution and not just theory, looking closely at how an in-memory database handles massive amounts of small object allocations and deallocations without crumbling under memory fragmentation. In the video, I talk about memory pages, how native allocators interact with the OS, and the classic problem of fragmented memory. We also look directly at the Redis source code and see how they wrapped malloc with zmalloc to abstract out the implementation. By the way, 18 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 16. Implementing the Approx LRU Algorithm 17. How Redis Caps Its Memory Usage 18. How and Why Redis Overrides Malloc If you have ever wondered how scale-critical systems avoid performance degradation over time and handle advanced manual memory management with custom concurrency and defragmentation controls, this deep dive is for you. Hope this helps you better understand database internals and spark that engineering curiosity. Give it a watch.

Fun fact: Redis does not just block at the total number of keys level when it runs out of space. It caps its capacity at the level of actual memory consumed, which means the database needs a highly efficient, real-time way to know exactly how many bytes it has allocated. I firmly believe that understanding theory is fine, but things become interesting when you sit down to implement it. This is where you find gaps in your understanding and understand the nuances. and hence... Today, we dive into the source code of Redis to look at a file called `zmalloc` and explore exactly how Redis tracks and caps its memory usage. This is the 17th video in the Redis Internals series. Like always, we keep our focus on execution and not just theory, looking closely at how Redis wraps standard allocation functions to maintain an atomic counter of its exact memory footprint. In the video, I talk about the separation of concerns and eviction logic by actually taking you through the Redis source code. We also discuss the practical nuances of implementing this, and why we handle it differently when writing a replica in a garbage-collected language like Go. By the way, 17 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 16. Implementing the Approx LRU Algorithm 17. How Redis Caps Its Memory Usage If you have ever wondered how low-level systems handle manual memory management without constantly firing slow system calls to check process memory, this deep dive is for you. Hope this helps you better understand database internals and spark that engineering curiosity. Give it a watch.

How should software engineers learn AI?

What if you could take three completely different model families… and distill them into one tiny model? 🤯 📜 Paper: arxiv.org/pdf/2605.21699 MOPD (Multi-Teacher On-Policy Distillation) has become a standard procedure in post-training. We already distill multiple specialized variants of the same model into a single set of weights. But what if we could go further - and distill models from entirely different families? Turns out, it is possible. Today we’re releasing a paper on cross-tokenizer distillation - our first steps in this exciting direction. 📄 We distilled Qwen3-4B, Phi-4-Mini, and Llama-3B into Llama-3.2-1B. MMLU jumped from 32.05 → 46.32 when using multiple teachers. 📈 The team is now working on Nemo-RL integration so the community can try this method in their own settings. Plus, we are scaling experiments up. 🚀

Senior backend interview question: You deleted the data. You deleted the backups. You deleted the logs. A court order says produce the records. They still find them. Where?

Just be good with advance search of github and you will get to know there is soo much more to CS rather than just frontend, backend and DSA