Shailesh Dixit

Message from a physician (Dr. Godse) to BJP on name change of KEM Hospital. Hope they reverse the decision.

You remember the auction room of 2018. Some kid from Patiala, barely old enough to vote, zero first class games to his name, suddenly worth 4.8 crore. It looked like the kind of punt that destroys franchises. 7 years on, that same kid is hitting sixes like his life depends on it. Because it does. From 2019 to 2022, Prabhsimran played just 6 IPL games. Punjab kept the chequebook open, kept him on roster, kept whispering about potential. But potential without release rots you. You spend nights in hotels carrying water bottles. You watch Gayle & Dhawan do your job. You start believing the auction price was a curse, not a blessing. He ran to stay human. 7 kilometers, then 8. Every morning. The burning in his lungs was better than the silence of the bench. Sachin Tendulkar told him something once; He said most boys would kill to sit where he sat. Told him to find contentment there. Sounds like advice for a monk, not a cricketer. But Prabhsimran clung to it. Then 2023 happened. The Impact Player rule dropped. Purists screamed about death of cricket. For Prabhsimran, it was oxygen. He could swing from ball one without worrying about the collapse. 13th May 2023, Kotla. Delhi Capitals had them against the wall. He walked in & played the innings of his life. First 30 balls: 27 runs. Next 35: 76 runs. No other batter crossed even 20 runs for his team while he scored his 1st IPL century. Delhi missed the playoffs. Prabhsimran found himself. But that is not why he bats like his life depends on it. You need to meet Sardar Surjit Singh, His father. Kidney failure. Three sessions a week hooked to machines. The house in Patiala is heavy with it. The only time the old man lights up is when his son opens for Punjab. They carry him to the couch before every game. Position the TV. For 3 hours, the sickness vanishes. Think about that weight. Every boundary is not runs for him, It is medication. Every six is oxygen for an old man. When he says he plays for his father, he means it literally. The single joy in a week of hospital corridors & medical debt. Come the 2025 auction, Punjab retained 2 players only. Shashank Singh & him. 4 crore for a keeper who had spent 4 years doing mostly nothing. He repaid them with 549 runs at 160 strike rate. Shreyas Iyer gave him something simple; Freedom. Told him he was senior now. Told him one bad game would not end him. Freedom to fail. Now he is flying. Strike rate pushing 173 this season. 80* against Mumbai off 39 balls. 51 off 25 against SRH. Each innings bigger than the last: 37, then 43, then 51, then 80*. The boy who waited is now the man who delivers.

Congrats to Vaishali akka! 🔥🔥Proud of what you have achieved! Inspiring to see how you handled the critical moments in the tournament! Happy to see the results for all the hard work!

BSWML has managed to complicate not just Solid Waste Management, but now even the GBA/BBMP property tax portal. #EaseOfLiving A simple property tax payment is turning into a confusing exercise due to poorly designed SWM user fee questions and unclear data requirements. There is no per-house waste weighment system, so all entries are merely estimates. And the framework imposes discriminatory charges on apartment communities that already segregate and compost their waste. Fix the SWM framework first—don’t make compliance harder for citizens. We seek urgent intervention from the Chief Commissioner, GBA. #BSWML @GBA_office @BSWML_GBA @KareeGowda_IAS @munishmody

@Dhairya__tweets A lot of uniformity has come in via the AA framework. If you have specific needs then our solution Precisa (precisa.in) is used by many CAs to analyse bank statements.

ALL NINE COPS TO BE HANGED!!! In a landmark judgement, the honourable First Additional District and Sessions Judge of Madurai, G Muthukumaran pronounced death penalty for all nine policemen in the Jeyaraj and Beniks custodial torture and death case. Inspector S Sridhar, sub-inspectors P Ragu Ganesh & K Balakrishnan, head constables S Murugan and A Samidurai and constables M Muthuraja, S Veil Muthu, S Chelladurai & X Thomas Francis have all been found guilty of the double murder. On 19th June 2020, Jeyaraj went to his son’s mobile shop in at Sathankulam in Thoothukudi. Police arrested him for violating Covid rules. Following the father’s arrest, Beniks went to the police station and requested for his father’s release. Beniks apparently had an argument with the cops. They arrested him also and tortured him and his father through the night and later sent to judicial remand. Beniks died at Kovilapatti Government Hospital on 22nd June, 2020, and Jeyaraj passed away the next day. The investigation later revealed that the mobile shop wasn’t open beyond the permissible hours. Basically a false case was slapped. Later, a female constable, Revathi came forward to tell the truth of their torture. Not only were the son & the father beaten brutally throughout the night but were also made to clean up their blood off the police station floor and walls with their clothes. Beniks sustained 13 external injuries and Jeyaraj had 17 wounds that ultimately led to their deaths. They were bleeding from their rectums when they were sent to jail!!! Following a huge social media uproar and protests, there was pressure on the then Tamil Nadu government to conduct a fair investigation into their deaths. If you go through the whole case, you will realise that 59 year old Jeyraj and 31 year old Beniks were killed purely because the ego of the policemen was hurt. They were angry that Beniks questioned why his father was arrested and being beaten up. Family of the victims hope this will be a lesson to all policemen who think they can get away with any crime. But unfortunately in our country most police don’t play by the rules. It is either their ego or orders from above or corruption that drives them. Hoping for better policing and a system that stands for victims, especially the voiceless ones.

Gauss meets real life. Also - Notice how people lifting 95 already say, “Fuck it, let’s do 100” - so there’s a discontinuity point. Mathematical theory faces reality.

There are no words.

Some stories just break your heart. 6 children danced at their school's annual day in Dindori near Nashik on Friday night. By midnight, all 6 were dead — along with 3 adults who had come to watch them perform. They fell into an open, unguarded 40-foot well at the end of a concrete road on their way home. A well that had swallowed a biker and a cow before. A well that nobody covered. @thaver_mohamed reports indianexpress.com/article/cities…

Still remains the most emotional piece I’ve ever written. Anil Gurav seemed destined to be the next superstar of Indian cricket but fate had some very different plans for him as @pdevendra and I found out. Just heard from his son about his passing indianexpress.com/article/news-a…

Anil Gurav passed away on March 31, 2026, aged 61. Once seen as Mumbai’s answer to Viv Richards, he lent Sachin Tendulkar the bat for his first competitive century. Mistakenly picked up by police searching for his gangster brother, he was brutally beaten yet lived all his life in a small room in Nallasopara. Even Ramakant Achrekar had marked him out early, placing a coin on his stumps—a sign of rare talent in Mumbai cricket. Before Tendulkar, Gurav was the chosen one.

I am all for the credit given to Ian Bishop for the research he puts into the junior cricketers. But he's reading up and giving you stories written by reporters in the front lines who have travelled to the small venues, interviewed aspiring cricketers in locations around the country and really put in the hard yards. To not credit the likes of @PurnimaMalhotra, @pratyush93_raj, @captainshanky and so many others while acknowledging Bish's efforts in reading up on them from the stories written by these reporters is a bit surprising. Every one of these stories are being heard on air because a reporter went and chased it down. Please do give them the credit they deserve.

Persis had one job left. Get justice for her father Jayaraj and her brother Bennix. Both beaten to death inside a Tamil Nadu police station in June 2020. For keeping a mobile shop open 45 minutes past curfew. For 6 years she fought. She shielded her mother from the full details of what happened inside those walls. When her mother finally heard everything in court last week she broke down. Today Persis got dressed. Went to court. Sat through the hearing. Came home with 10 more days to wait. She has not forgotten a single day of the last 2191. April 9. She will be back in court. She always is.

1. Rinku Singh cleared the PCS exam in 2004. 2. In 2009, he exposed a ₹100 crore pension and scholarship scam. 3. For exposing corruption, he was shot 7 times, with 2 bullets hitting his face. 4. He survived, recovered, and went on to become an IAS officer in 2022. 5. Now he has resigned, reportedly because he was denied a field posting. 6. This reflects a broken system where honest officers are sidelined while corrupt ones often get powerful field roles. 7. If the public does not stand with honest officers, integrity will keep being punished and corruption will keep thriving.

30+ artists including me were summoned by authorities for being on Latent. The whole comedy fraternity took a hit because of how stupid you’re. Shows were canceled, Venues pulled out, Judgements made, the pathetic shadow of your stupidity still continues to make lives of funny comedians difficult. You didn’t even care to understand that side. Stop pretending to be the nice guy you’re not. Take your cheque’s, lower your gaze & be very ashamed. You’re a contraceptive for creativity, stop milking this & go back to what you do best which is being a hurdle for upward social change while FraudCasting & clout chasing…

The passing of 10-month-old Alin Sherin Abraham in a road accident at Kottayam has deeply saddened us all. In the midst of unimaginable pain, her parents, Arun Abraham and Sherin Ann John, chose to donate her organs. By becoming Kerala’s youngest organ donor, little Alin has granted a new lease of life to five others, a monumental act of compassion that reflects the true spirit of Kerala.  I extend my deepest condolences to the family. I also salute the dedicated health professionals, police and support teams whose swift coordination made the transplants possible. We will bid farewell to Alin Sherin Abraham with official honours.

Goosebumps! People often ask, is Shakespeare still relevant? Here is a great example, from the Steven Colbert Show, in which Sir Ian McKellen delivers an extraordinary speech. Shakespeare’s words are timeless, urgent and important. #Shakespeare #ianmckellen #stevencolbert

Further update to Mumbai Police rejecting my passport verification: 1. Mumbai Police official from the local police station reached out and explained that the verification is rejected because a marriage certificate was not given. I am now asked to reach out to @rpomumbai for re-opening of the file. Next appointment with RPO is in April. The current officer was helpful and kind to give the reason and explain the next steps - Thank you. 2. The central govt. (MEA) has stated that to add a spouse's name in the passport, ANY 1 of the following needs to be given: a) marriage certificate OR b) Annexure J form. The government does not mandate marriage certificate as the only document. However, @MumbaiPolice has made it mandatory through one of it's circular that they will only accept marriage certificate as the document. This circular, supersedes the central govt rules which accepts either Annexure J or the certificate. This arbitrary circular by the police make it more bureaucratic for the citizens because the document list on MEA's passport website says Annex. J is fine. if the MEA allows Annexure J, how and why is Mumbai Police bringing in circulars that only accepts marriage certificate? @MEAIndia @passportsevamea @CPMumbaiPolice Image 1 - MEA Passport website Image 2 - from the 2022 circular of Mumbai Police

₹50,000 premium. ₹15 lakh coverage. ₹0 peace of mind. India's health insurance system is broken. And no one's fixing it. Let me tell you what happened to my family last week. My wife needed gallbladder surgery. We have HDFC Optima Secure with @HDFCERGOGIC . We've paid premiums for years. We did everything right - got cashless approval, chose a network hospital (@ManipalHealth Whitefield), followed every process. On 26th Dec, I received the settlement letter: "Amount to be paid by insured: ₹0" Surgery happened. It went well. She was recovering. We were ready to go home. The hospital refused to let us leave. Despite holding a valid insurance document that said we owe NOTHING, they wouldn't discharge a recovering patient. My wife sat in that room - tired, in pain, wanting to go home - while I begged at the billing counter. We were only "allowed" to leave after I blocked ₹X,XX,XXX on my credit card. A patient. Held hostage. In a hospital. In India. In 2025. Then on 30th Dec, magically, a new settlement letter appeared: "Amount to be paid by insured: ₹X,XX,XXX" Same surgery. Same bills. Same everything. But now I suddenly owe lakhs. Here's how the trick works: The hospital bills ₹5.8L. The insurance company says "we have an MOU with the hospital, agreed package rate is ₹1.9L." So they settle for ₹1.9L. The remaining ₹3.9L? "Patient to pay." But wait - the MOU is between the hospital and insurance company. The "agreed package rate" is THEIR agreement. Why is the patient paying the difference for a deal they negotiated? The letter was updated AFTER we left. The hospital and insurance company sorted it out between themselves. The patient? Just a pawn in the middle. Both letters say: "No amount to be collected from insured" Their own words. Their own document. Exposed for what it is - meaningless paper. This is not a one-off incident. This is the system working exactly as designed: → Insurance companies collect premiums for years, then find ways to deny or reduce claims → Hospitals inflate bills knowing insurance will negotiate down, and patients will pay the difference → Patients are trapped between the two - too exhausted to fight, too scared to refuse → IRDAI exists but is reactive, not preventive - they'll act after you file a complaint, not before you're harassed Where is the protection for patients? Where are the regulations that say a hospital CANNOT hold a patient hostage for payment disputes? Where is the accountability when an insurance company issues two contradictory letters and expects the customer to just figure it out? We pay GST on premiums. We pay taxes. We fund the regulators. And when we need help, we're left standing at a billing counter, blocking credit cards to take our family home. I've filed a complaint with @IRDAI_official. I have every document. I'll fight this. This is not customer service failure. This is systemic exploitation. @HDFCERGO - honor your own document. Claim: RC-HS25-15349896_1 Policy: 2856205638548200 @IRDAI_official - regulate before patients are harassed, not after. @JPNadda - health insurance is not a luxury product. Millions of Indians depend on it. The current system is failing them. @raghav_chadha Need your help to raise this insurance scam in parliament.

I’m non-technical but want to deeply understand AI. @karpathy's “Intro to LLMs” is the best resource I’ve found so far. Here are my biggest takeaways and questions from his 60-minute talk: 1. A large language model is “just two files.” Under the hood, an LLM like LLaMA‑2‑70B is literally (1) a giant parameters file (the learned weights) and (2) a small run file (code that implements the neural net and feeds data through it). Question: If the architecture code is tiny and public, what actual moat is left besides the weights? 2. Open‑weights vs closed models. LLaMA‑2 is open‑weights: architecture + weights + paper are public. GPT‑4, Claude, etc. are closed: you get an API/web UI but not the actual model. Question: For a company, when is “renting” a closed model strategically worse than owning an open‑weights model? 3. Training vs inference: training is the hard, expensive part. Running the model (inference) is cheap; getting the weights (training) is a major industrial process. Question: Where is the greatest axis of innovation in front of us to lower the cost of training significantly? 4. Pre‑training compresses ~10 TB of internet text. LLaMA‑2‑70B is trained on roughly 10 TB of scraped internet text, compressed into 140 GB of parameters—a ~100× lossy compression of “internet knowledge.” Question: Given that we’ve run out of knowledge on the internet to pre-train models on, is new data going to be the limiting factor on model improvement moving forward? 5. Training scale: ~6,000 GPUs × 12 days ≈ ~$2M for LLaMA‑2‑70B. That’s already described as “rookie numbers” compared to modern frontier models, which are ~10× bigger in data/compute and cost tens to hundreds of millions. Question: How far are we from “more compute” no longer being a competitive advantage? 6. Frontier models just scale this up by another ~10×. State‑of‑the‑art models (i.e. GPT‑5) simply dial up parameters, data, and compute by large factors relative to LLaMA‑2‑70B. Question: How much of GPT‑5‑style capability is just more scale vs genuinely new algorithms? 7. Core objective of an LLM predict the next word in a sequence. LLMs are trained to take a sequence like “the cat sat on the” and predict the probability distribution over the next word (“mat” with ~97%, etc.). Question: The beauty and the curse of LLMs is them being probabilistic. How can we create the right constraints such that people trust LLMs in enterprise settings? 8. Architecture is known: the Transformer. We know all the math and wiring (layers, attention, etc.); that part is transparent and simple relative to behavior. Question: If the architecture is commoditized, where exactly do you build sustainable differentiation? And how much more shelf life is there on the Transformer before a new architecture takes over? 9. Parameters are a black box. Billions of weights cooperate to solve next‑word prediction, but we don’t really know “what each one does”—only how to adjust them to lower loss. Rabbit hole: Read about mechanistic interpretability work. 10. Treat LLMs as empirical artifacts, not engineered machines. They’re less like cars (fully understood mechanisms) and more like organisms we poke, test, benchmark, and characterize behaviorally. Rabbit hole: Understand the current process for evals & if/what limitations exist in today’s eval tools. 11. Pre‑training vs. fine-tuning. Pre-training favors quantity over quality; Fine-tuning flips that: maybe ~100k really good dialogs matter more than another terabyte of web junk. Question: How much incremental performance can fine tuning and RHLF drive for models? Is it a fraction of what pre training does for performance or is it more meaningful than that? 12. Knowledge vs behavior. Pre-training loads the model with world knowledge; Fine-tuning teaches it to be helpful, harmless, and to respond in Q&A format. Rabbit hole: I’d love to deeply understand how exactly a model is fine tuned from beginning to end. 13. Reinforcement learning from human feedback (RLHF) via comparisons. It’s often easier for labelers to rank several options vs. write the best one from scratch; RLHF uses these rankings to further improve the model. Question: When exactly does it make sense to fine tune a model vs. use RHLF & does the answer depend on the domain of knowledge the model will be used for? 14. Closed vs open models. Closed models are stronger but opaque; open‑weights models are weaker but hackable, fine‑tunable, and deployable on your own infra. Question: As companies deploy agents, what is the most important consideration to make as they think about their AI tech stack? 15. Scaling laws: performance is a smooth, predictable function of model size and data. Given parameters (N) and data (D), you can predict next‑token accuracy with surprising reliability, and the curve hasn’t obviously saturated yet. Question: If capabilities keep scaling smoothly, what non‑technical bottlenecks (data rights, energy, chips, regulation) become the real limiters? 16. GPU and data “gold rush” is driven by scaling law confidence. Since everyone believes “more compute → better model,” there’s a race to grab GPUs, data, and money. Question: Let’s assume scaling laws no longer scale. Who is most screwed when the music stops? 17. LLMs as tool-using agents, not just text predictors. Modern LLMs don’t just “think in text”; they orchestrate tools. Given a natural-language task, the model decides to (1) browse the web, (2) call a calculator or write Python to compute ratios and extrapolations, (3) generate plots with matplotlib, and (4) even hand off to an image model (like DALL·E) to create visuals. The intelligence is increasingly in the coordination layer: the LLM becomes a kind of “foreman” that plans, calls tools, checks outputs, and weaves everything back into a coherent answer. 18. How do LLMs know when to make a tool call? “It emits special words, e.g. |BROWSER|. It captures the output that follows, sends it off to a tool, comes back with the result and continues the generation. How does the LLM know to emit these special words? Finetuning datasets teach it how and when to browse, by example.” 19. System 1 vs System 2 thinking applied to LLMs. Concept popularized in Thinking Fast and Slow. System 1 = fast, instinctive; System 2 = slower, deliberate, tree‑searchy reasoning. Right now LLMs mostly operate in System 1 mode: same “chunk time” per token. Rabbit hole: Explore how “chain‑of‑thought” method works & what limitations still exist in System 2 thinking for LLMs. 20. Desired future: trade time for accuracy. This was before the first reasoning model (GPT O1) came out. At the time, Karpathy talked about this idea of wanting to be able to say: “Here’s a hard problem, take 30 minutes,” and get a more accurate answer than a quick reply; currently, the models can’t do that in a principled way. 21. Model self‑improvement example: AlphaGo’s two stages. AlphaGo first imitates human Go games, then surpasses humans via self‑play and a simple, cheap reward signal (did you win?). Question: What’s the best way to improve models in domains where there isn’t a simple reward function, like creative writing or design? 22. Retrieval‑augmented generation (RAG) as “local browsing.” Instead of searching the internet, the model searches your uploaded files and pulls snippets into its context before answering. Question: Where does RAG break down in production? 23. Think of LLMs as the kernel process of a new operating system. This process is coordinating resources including tools, memory, and I/O for problem-solving. Future LLM will: - read/generate text - have more knowledge than any single human about all subjects - browse the internet - use existing software infrastructure - see and generate images and video - hear and speak and generate music - think for a long time using system 2 - “self-improve” in domains with a reward function - customized and fine-tuned - communicate with other LLMs Rabbit hole: Draw out the LLM OS and explain it to someone. This will show how well you understand the technology. 24. The LLM OS is reminiscent of today’s operating systems. The finite context window is like working memory; browsing/RAG are like paging data in from disk or the internet; rapidly growing closed vs. open ecosystem; Managing what’s in context is a core challenge. Rabbit hole: Explore techniques for working across many context windows & longer-running tasks. 25. New computing stack → new security problems. Just as OS’ created new attack surfaces (malware, exploits), LLM‑centric stacks create their own families of attacks. Jailbreaks, adversarial prompting, adversarial suffixes, and prompt injection. Question: security for AI systems seems orders of magnitude harder than traditional software because the # of edge cases feels infinite. Is this assumption right or wrong? 26: LLMs are a new computing paradigm with huge promise and serious challenges. They compress internet‑scale knowledge, act as operating‑system‑like kernels, orchestrate tools and modalities, and open up both transformative products and novel security risks. Question: what is the most nascent part of the LLM OS that needs to be built up in order to accelerate diffusion of the technology? Link to the full “Intro to LLMs” video below 👇