Varun Mayya's Recent LinkedIn Posts

In this kata, I reveal why 99% of Indian companies are burning money on marketing that hurts their brand.  While the best brands spend ₹0 and still win. Watch to learn how to build a brand in India.

NVIDIA has announced groundbreaking DLSS 5 that makes games look like real life.

Sat down with Alex Wang the guy who dropped out of college, built Scale AI as a teenager, became the youngest self-made billionaire, and is now building superintelligence at Meta. We talked about Meta's approach to superintelligence, why distribution matters the most now, upcoming Meta glasses & why the next 5 years of AI will be monumental.

Jensen Huang has a keynote in three weeks. Every time he's done this, the people who paid attention knew what to build next in AI. NVIDIA GTC is about to drop the biggest tech announcements of 2026, and before I get into everything, we're doing a giveaway — an NVIDIA JETSON AGX THOR developer kit and Brev Code credits worth $100. I'll tell you exactly how to win at the end of this video, so stay till the end. So GTC is basically where NVIDIA shows the world what's coming next. This year the focus is on agentic AI, large language models, sovereign AI infrastructure, Omniverse, digital twins, robotics, and generative AI. Basically every major frontier in tech right now, all in one place. We're talking humanoid robots going from concept to production, AI agents that can actually write and review your code, digital twins for entire factories, and Jensen Huang's keynote which — if you've seen his past ones — is basically a masterclass in where technology is headed. But here's what I think is really cool. There are sessions on everything from autonomous driving safety to quantum computing integration to visual AI agents that understand video in real time. One session is literally called "Turning Models Into Engines: The AI Factory Era." Like we're not just building AI anymore, we're building the infrastructure for AI to build things. GTC is happening March 16th through 19th and you can attend virtually for free. Link to register is in the description. Now the giveaway — to enter, comment below with which GTC session you're most excited about, make sure you're subscribed, and I'll pick a winner after the conference. Register For Nvidia GTC: https://nvda.ws/3M9DJ3Z

Turns out Pokémon Go players accidentally helped train real delivery robots. Over eight years, the gamers helped generate more than 30 billion real-world scans, and now Niantic is using that data to guide food delivery bots through city streets with centimetre-level accuracy. The same tech that placed Pikachu on your sidewalk is now helping robots find your front door.

Claude can now control your entire computer for you!

Militaries are now using real, AI-powered “spy” cockroaches! A startup called SWARM Biotactics has just built something called a "bio-robotic scout." It’s basically a living cockroach with a tiny “backpack” housing a camera, a mic, a transmitter for communications, and an AI chip for local data processing.

Career advice for engineers right now is mostly noise. Everyone has an opinion, everyone has a take, and most of it is either obvious or just wrong. That's why I decided to sit down with Mohak Shroff, SVP of Engineering at LinkedIn, to figure out what was really going on. And to my delight, he was direct about things. For example, he said that the engineers who are going to struggle most are the ones whose entire value is in writing code. Reason? It’s not a skill shortage anymore. Any sufficiently smart LLM can do that. And because of it, the idea of a “job” has shifted more toward judgment. On the practical side, his advice for job seekers was pretty grounded. Verifying your identity, your skills, your background, it signals more than most people give it credit for. And the network thing is still true, most real opportunities come through people you already know, not through the applications people send. But the thing he kept coming back to, and honestly the part that stuck with me the most, was this: stop trying to figure out which role to chase. Get better at thinking clearly. Get better at expressing what you think. As execution gets cheaper and cheaper, the thing that holds value is the ability to make good calls and communicate them well. Part 1 of our podcast go live tomorrow. Link in the comments as soon as it's out.

Sam Altman just said training a human child takes more energy than training an AI model. yeah he actually said that. Altman's argument is simple. People complain about how much energy AI uses, but nobody talks about the 20 years of food and energy it takes for a human to become useful. He says if you compare one AI query to one human answer, the AI is already more efficient. And look, the energy numbers are real on both sides. Humans use massive amounts of energy, water and resources over a lifetime. But AI data centres also use billions of gallons of water just for cooling. They're pulling huge amounts of electricity from power grids. Energy prices are going up in towns near these facilities. And Altman's response is basically "humans use energy too, so what's the difference?" But here's what that framing actually does. It turns a child growing up into a training run. It turns food into fuel cost. It turns 20 years of life into compute time. And once you accept that a human is just an expensive biological computer, then spending any amount of energy, water and resources on a faster, cheaper synthetic version starts to sound logical according to Altman. even if it means strained grids, higher prices, and stressed ecosystems for actual people. The resources we're talking about aren't abstract. water. electricity. land. These are things communities need. So here's the question: Is it right to compare training AI to raising a human when both use massive resources, but only one of them is actually a human being?

You can now move your entire AI memory to Claude within seconds! And the best part? You don't even need a paid plan anymore since Claude's memory is now free for everyone. You simply go to claude.com/import-memory, copy this prompt, paste it into your AI app (Gemini, ChatGPT, or whatever you’ve been using), and then paste that output into Claude’s memory! That’s it. Everything from your name, background, preferences, writing style, and recurring projects just gets imported to Claude instantly! But why now? Well, the timing is what makes it so significant. Because after a very public “fight” with the US government last week, millions of users switched to Claude! It even shot to #1 on the App store! And it makes sense too. Claude is genuinely one of the best models out there right now. So if there’s one thing that can make new users stay, it’s to ensure Claude already knows them. Having said that, it wasn't impossible before. But you had to manually ask other apps to give you your memory, and then copy paste it into Claude. Plus, you didn’t even know the right prompts and had to figure it out on your own… it was a mess. But with this update, that headache is simply gone.

Jack Dorsey just fired half his company because of AI! The stock even jumped 25% even though this was the largest layoff by percentage in S&P 500’s entire history!

This Indian startup just built an AI that finds you jobs! #AD Try tal here: https://lnkd.in/gKBpQvrw

Anthropic just released a study on which jobs AI is actually replacing. And there is one chart that every employee, and fresher, needs to stop and look at right now. Blue is what AI could theoretically do in each job category. Red is what people are actually using AI for right now. The further out from the centre, the higher the coverage — zero to 100%. And the gap is massive. Blue is way bigger than red across almost every category.

Apple just launched Macbook Neo, it's cheapest macbook ever!

Gucci just posted an AI-generated ad, and people on the internet are already trashing them! See these comments!

Trump just blacklisted Anthropic, kicked them out of all government systems, and even labelled them a national security risk.

Xbox just announced project Helix! A new Xbox which can run PC games.

This person created a custom cancer vaccine for his dying dog, using ChatGPT!

Is this the era of Indians finally looksmaxxing?

Google just launched an AI tool that lets anyone design a complete app with just a prompt. If you're a UI/UX designer, you need to rethink everything. The AI tool is called stitch and you can literally type something like "build me a swiggy-type food delivery app but for ghar ka khaana," and it generates the screens, the layout, the interactions, all of it in one shot.

Scientists just copied a fruit-fly brain inside a laptop and made it think!

OpenAI Launched a New Model called GPT 5.4!

Nvidia just announced a massive partnership with India to speed up the IndiaAI Mission! Focused on building “sovereign AI,” it targets three main areas: AI infra, AI agents, and manufacturing. Starting with infra, India is building "AI factories” completely locally using tens of thousands of Nvidia GPUs! Through companies like L&T, Yotta, and E2E Networks, it will now allow India to build more of its own AI models like Sarvam AI. On the agent side, Indian IT giants like Wipro and Infosys will now be using Nvidia’s “Nemotron” to not only build AI agents but also deploy them for their clients. For example, Wipro recently rolled out an AI-agent for a major US healthcare company that now handles over 42% of incoming calls on its own! And finally: manufacturing. Companies like Reliance Industries and Tata Motors will now be using Nvidia Omniverse to build "Digital Twins" to simulate everything from assembly lines to factory operations completely virtually! Honestly, it’s great to see India finally building its own AI… as it will not only save India billions of dollars but also speed everything up! And on that note, follow us to stay updated on the latest tech news.

Anthropic just caught three Chinese AI labs stealing from their AI model. like, at an industrial scale. Here's what happened. Anthropic identified what they're calling "distillation attacks" on Claude by three labs — DeepSeek, Moonshot AI, and MiniMax. These labs created over 24,000 fraudulent accounts and generated more than 16 million exchanges with Claude, extracting its capabilities to train and improve their own models.