TánMoN

Becoming a parent changes your lens. You stop chasing noise, and start seeing time in decades — not days. Parenthood teaches you to zoom out from propaganda and focus on what truly shapes the future.

@adxtyahq Using BMAD?

i’ve stopped caring about API bills - claude sonnet 4.6 builds it - claude opus 4.6 (high thinking) reviews it - gpt 5.4 codex designs the architecture and i just sit back sipping coffee

Visuals from #NarmadaGhat, Narmadapuram. Please do the needful to create a system & fix this forever.. 🙏 @mppcb_official @CPCB_OFFICIAL @CMMadhyaPradesh @SwachhBharatGov @moefcc #SaveNarmada #CleanNarmada #SwachhBharat

Uninstalled @ixigo — not just for pathetic service by @abhibusIN, but for disabling reviews to protect them. If you hide feedback, you support bad service. No more bookings via @ixigo or @abhibus.

Omkareshwar Temple on Weekend: 4‑hr jammed queue, unsafe for kids/elderly, no security/water/food, narrow passage, poor crowd control. Need better management. Very high probability of stampede. @OmkarJyotirling @DM_Khandwa @MPTourism @KhandwaSp @nvdamp @CMMadhyaPradesh

pagarparchi.in Demonstration Video youtu.be/YmKCaTsGzqI?si… via @YouTube

@iyoushetwt VB

What was the first programming language you learned ? Mine : C++

The Only AI That Reads Huge, Complex PDFs — Fully Offline and Private. #RAG #OfflineAI #PrivateAI #LocalLLM #AIEngineering #Java #SpringBoot #FullStack #CUDA #VectorSearch #DocumentAI #PDFIntelligence #AIDev #SearchEngineering #Microservices #LLM #ConfidentialComputing #AIInfra

Becoming a parent teaches you the real meaning of “long-term.” You stop living only for yourself… and start living for someone who carries your entire world in their tiny hands.

मातृत्व या पितृत्व सिर्फ बच्चों को बड़ा करने के बारे में नहीं है, यह खुद को उस इंसान में बदलने की यात्रा है — जो शोर से परे, अहंकार से ऊपर, और आज से आगे सोचता है।

Parenthood isn’t just about raising children. It’s about raising yourself — into someone who thinks beyond noise, beyond ego, beyond today.

get into FPGAs. they’re not as hyped as AI or robotics, but they’re the hidden backbone of both. you want speed? determinism? ultra-low latency? hardware-level parallelism? that’s FPGA territory. every serious system; autonomous vehicles, defense tech, high-frequency trading, industrial control; runs faster because some engineer somewhere mastered reconfigurable logic. software gives you flexibility. hardware gives you performance. FPGAs give you both. learn VHDL or Verilog. build a simple CPU. implement a PID controller. accelerate an ML inference. you’ll suddenly see how computing really works; not just in code, but in electrons. FPGAs are the bridge between hardware and software thinking. and if you can master that bridge, you become dangerous.

You only need 20 videos

A drone that flies, drives, and… switches modes in 0.1 seconds! ⬇️ Build it yourself: [📍 Bookmark for CAD + parts] No extra actuators, no deformation, just clever mechanics and full control. DUAWLFIN is a ground-aerial robot with unified actuation: flying like a quadcopter, rolling like a car, and transitioning seamlessly between modes. ✅ Climbs 30° slopes ✅ Hits 2 m/s on wheels with just 15W ✅ Only 3% added energy in flight mode ✅ Mode switch in 0.1s ✅ Fully open-source and 3D-printable Perfect for urban logistics, indoor nav, or just rethinking what drones can be. Thank you, Ruiqi Zhang, for sharing! Paper: arxiv.org/pdf/2505.13836 Website: sites.google.com/view/duawlfin Build it yourself: CAD + parts list in the paper 📍 BOOKMARK FOR LATER This is how you merge air and ground without compromise.

John Carmack explains how he applies Nassim Taleb's "anti-fragile" concept to his work, enjoying the thrill of new ideas while accepting that many won't succeed. Source: Deep Thoughts Engineering Speaker Series: John Carmack

I love when Karpathy drops something new! This time it’s NANOCHAT. This is how to build a full LLM for ~$100 with a simple 5-step process in one evening. 👇 What is the Value? Right Now: Too many tools. Too many buttons. Ideally: You just want a bot that talks back — without 100 dependencies. Karpathy’s 5-Step Blueprint Solve this: 1️⃣ Set up a cloud computer. You just rent a strong computer for a few hours like booking a supercomputer on demand. → Technically, that’s an 8×H100 GPU box (for example, on Lambda Cloud). No setup pain. 2️⃣ Download nanochat. You grab Karpathy’s new open-source project small, simple, and fast. → Then you just run one file: speedrun .sh. Behind the scenes, it sets up your environment, downloads the data, starts training, and even builds the chat interface. 3️⃣ You teach it to read. First, you help your model learn the basics how to break sentences into pieces it can understand. → Then, under the hood, nanochat trains its own tokenizer (in Rust) and learns from FineWeb, a clean web-text dataset. 4️⃣ You teach it to talk. Now you train it to have real conversations. It practices answering questions, solving small math problems, and chatting naturally. → This part is called midtraining, using datasets like SmolTalk, MMLU, and GSM8K. 5️⃣ You open the chat window. After about 4 hours, you can talk to your model in a simple web UIjust like ChatGPT. It can write, explain, and even solve simple logic problems. → If you let it train longer (~12 hours), it can beat GPT-2 on standard tests like MMLU and ARC-Easy. AI Community Loves it! 25K+⭐ GitHub stars in 6 days. It opens a thousand doors. Because once you get nanochat running, you can turn it into: - a new experiment, - a new skill, or even - a new AI Agent. 7 Experiments You Can Try: ① Train deeper. Change one number (--depth=26) to see reasoning improve. → Watch your chatbot get smarter. ② Swap data. Use your own text — finance, medical, or research. → Build a model that speaks your language. ③ Tweak the tokenizer. Add code tokens or symbols. → See how language compression really works. ④ Test tool use. Hook it to a Python sandbox or API. → Turn chat into action. ⑤ Add memory. Wrap it with LangChain or CrewAI. → Get a goal-driven, multi-step agent. ⑥ Benchmark fast. Run ARC-E, GSM8K, HumanEval. → Learn what truly moves performance. ⑦ Log everything. Edit the report. md with notes or charts. → Track your runs like real experiments. nanochat: github.com/karpathy/nanoc… ≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣≣ ⫸ꆛ Want to build Real-World AI Agents? Join My 𝗛𝗮𝗻𝗱𝘀-𝗼𝗻 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁 𝟱-𝗶𝗻-𝟭 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 — trusted by 1,000+ builders worldwide! ✔ Only basic Python required. ➠ Build Agents for Healthcare, Finance, Smart Cities & More ➠ Master 5 Modules: 𝗠𝗖𝗣 · LangGraph · PydanticAI · CrewAI · OpenAI Swarm ➠ Includes 9 Real-World Projects 👉 𝗘𝗻𝗿𝗼𝗹𝗹 𝗡𝗢𝗪 (𝟱𝟲% 𝗢𝗙𝗙): maryammiradi.com/ai-agents-mast…

Now anyone can pre-train your own model in 4 hours. Incredible work by @karpathy to open source and democratize some of the most important educational resources in the AI Era. The “Eurekas Per Second” in nanochat is something you must experience. The NVIDIA Brev team has created a launchable (see below) for you to try! Simply click “Deploy Launchable” and a GPU will be provisioned in the cloud, and nanochat will begin training! The first 10 developers to deploy will be able to train their own GPT 2 style model completely for free. Happy hacking!

Stay tuned with @NVIDIARobotics folks, we’re expanding @LeRobotHF’s sim capabilities! I can train & teleop my SO-101 from real → sim, drop custom assets, and collect data from home (or HF office). finally making progress on the robotics dataset problem. Project launches soon 👀

karpathy speaks like someone who’s running a mental compiler in real time with minimal interpretive latency & almost zero runtime garbage. he’s not verbose. he just threads complexity into compressed lossless statements. most smart people can be dense, but they lose clarity. karpathy keeps clarity while cranking the bitrate like an llm tuned with perfect temperature control. it’s so beautiful to listen to that i had to hear this three times already. i might add a fourth.

For anyone wanting to start from the basics. Get a copy. (I don’t do paid promotions. I got my copy and found the text to be something I’ll save for the next generation) Let me know if you have other recommendations.