Shubham Kashyap

The weirdest part of the #Anthropic deal? Elon's infrastructure is now powering a direct competitor to #Grok. xAI competes with #Claude, yet SpaceX’s #Colossus1 compute is being leased to Anthropic. Just like #AWS ended up hosting #Amazon’s competitors. Everyone winning!

AI rewards curiosity which unfortunately isn’t scalable for everyone.

Trust is a #dangerousgame: if you trust, you risk betrayal; if you don’t, you lose cooperation. So #society built “trust infrastructure” like banks, laws, ratings to reshape #incentives. The next big winners won’t just build #AI, they’ll make it #trustworthy

What if the next AI #breakthrough isn't a bigger model but truly autonomous agents that deploy apps on their own? #Cloudflare + #Stripe just enabled AI agents to create accounts, buy domains & launch production apps with #zerohumansteps

They want #AI outputs that look like expert work, without going through the #messy phase #experts went through.

People waiting for the “right time” to learn #AI will meet people who started at the wrong time! #start #learning #today

Playstore: play.google.com/store/apps/det…

A parent once told: “Dr., I didn’t know something was wrong. He seemed fine.” That’s how developmental delays get missed. I built Grow Baby Grow, a free, offline app to help parents track their child's #growth & #vaccination. Share it with a new #parent! #OpenSource #IAP #NIS

Repo: github.com/Xdev200/Grow-B…

@karrisaarinen True, but flip it once. You know exactly what good looks like, so AI's near-misses become frustrating. The noob gets 80% value effortlessly. The expert fights for the last 20% & are disappointed. It's not about perfect AI but understanding its limits and evolving around it.

A common dynamic I observe with AI: it feels most impressive when you don’t know much about the subject, don’t care or don’t have a clear idea of what the you want. This applies across design, code, legal, and more. If I don’t know code very well, every piece of code it writes feels very impressive. Once you know what something should feel or look like, it becomes almost impossible to guide AI there. And you definitely can’t one-shot it.

Google dropped #Gemini Enterprise Agent Platform at #GoogleCloudNext '26 The #shift is from working to designing work! Employees should learn to guide agents & not just do tasks. Break problems into intent + constraints. Have you built any agents yet? #Adapt or #GetAutomated

Hot take: the #AI #memory problem isn't about #storage or #context length. It's about retrieval #noise. Curious if anyone #building #agents has actually measured how memory quality changes over a long run and what they found.

New #AI models drop every other day, beating benchmarks, but the real-world output rarely improves. Better results come from better questions and it's not new wisdom. There is a separate literature just for the right questions- Prashna Upanishad #SeekingTruth #Intelligence

AI is spitting out more potential "drugs" than ever. A startup just raised $4.8M to figure out which drugs actually matter. I can see parallels coming up in unrelated industries. Switch the "drugs" with anything that AI generates in bulk, & Voila! You got an idea. #AI #Idea

@Nithin0dha We need a funnel to return the accumulated wealth back to lowest strata by enabling sustainable social ventures. You are doing it right by investing in the ecosystem. Hope more people join you🙌🏽

I'll admit this might sound odd coming from me, maybe even clichéd. But it's something I've been sitting with for a while, so here goes. When I started out, like most people, I had a simple wealth goal. I'd actually written it down: hit ₹5 crore, retire in Goa, beach shack, done. That was the dream. After the Zerodha journey, I find myself on a very different side of that equation, and the dark inequalities of wealth and opportunity are harder to ignore than ever. We all know the numbers on inequality. The concentration of wealth among the top 1% is severe and getting worse, and it's even starker among the top 0.1%. The post-2008 era of rising asset prices has likely made this worse, because the people who hold financial assets are, by definition, people who already have money. This isn't unique to India. Barring a few exceptions, it's a global phenomenon. I'm cautious about attributing every socio-political problem we face today to inequality, but it's hard to deny the role it's played in the political upheavals we're seeing across the world. History rarely shows that sustained, extreme inequality ends well. To me, it increasingly feels like sitting in a car with the brakes cut, watching a cliff approach. Btw, all of this even before AI, which has a non-trivial probability of making things worse. I'll stop short of prescribing solutions. It's too easy to reach for simple answers to complicated problems, and that's a separate conversation entirely. But I think we need to collectively acknowledge this: wealth that just sits in financial assets whose value keeps compounding upward doesn't do much good for anyone beyond those who already have it. And if that wealth isn't in motion, if it isn't doing some social good, the fabric that holds us together will only continue to fray and lead to cynicism, resentment, and worse yet, nihilism. We're already seeing all of it. What I am saying is that even if a portion of that wealth were channelled into things that could materially improve lives, that seems worth doing. Hoarding wealth, in the grand scheme of things, doesn't really help anyone.

I wonder if forgetfulness is actually what we need in AI memory. Imagine human memory without any forgetting… we’d probably hallucinate just like today’s AI. #AIMemory #AI #Idea

Models will become commodities. Real power will sit with those who control infra. Big Tech will gatekeep & smaller players will get squeezed unless open-source moves faster than the moat. What's your take? #AIInfra #Moat #AI

Google now let patients store & sync medical records directly in Fitbit. Wristband just became your personal health command center. #integrationOpportunity

Her Photo 51, an X-ray image revealing DNA’s double helix, became the breakthrough others used to win the Nobel. She saw the truth first, the world just caught up later. #RosalindFranklin #WomenInSTEM #DNA #ScienceHistory

OpenAI naming its life sciences model Rosalind is a powerful tribute to Rosalind Franklin. From uncovering DNA’s structure to decoding hidden biological patterns, this AI could shrink decade-long drug trials into months and unlock personalized cures for rare diseases. #OpenAI