Sourav Das

Data scientist|
Sanatani💪

[Siraj Hashmi @SirajAHashmi]

God grant me the strength of a lone male juror facing 11 retards

[Aarav @itsme_aaravvv]

he is the brother of a 17 year old NEET girl who committed suicide because of reservation.
she wrote in her suicide note "papa, this system has failed me, i scored well.. but they selected those who scored less than me".

"feminist" leftists are mocking brother of that dead girl.

[Nick Collins @nickcollins1953]

Migration out of India seeded the ancient world.

Two hills in southwest Ireland are named after a Rig Veda goddess. They are called the Paps of Danu. Every spring, farmers lit fires around them and drove cattle between the flames for protection. The Vedas describe the same ritual.

Danu is the Rig Veda mother goddess of the flowing waters. Her name is written into the geography of half of Eurasia. The Danube. The Dniester. The Dnieper. The Don. The Donau. Greek Demeter. Northern Greeks once called themselves the Dannuni. In Ireland, the pre-Christian pantheon called itself the Tuatha de Danaan. Children of Danu.

The mainstream story that pale steppe nomads invaded India around 1500 BC and imposed Sanskrit on a dying Indus Valley civilisation has been steadily dismantled over the last thirty years. The recent genetic and linguistic papers still cited in its defence describe contact at the edges, not replacement at the centre.

The direction of travel runs the other way. Out of India, in pulses, over thousands of years.

Iran took the name of the people, Aryans. Ahura Mazda is the Vedic Asura. The magi are the Vedic magoi.

The Germanic and Slavic pulse moved overland from the ninth millennium BC. They carried the sky-father Dyaius, who became Tyr in Scandinavia, Zeus in Greece, Deus in Rome.
The Italo-Celtic pulse moved by sea from the seventh millennium. When the Celtic branch turned up the Atlantic coast around 5,400 BC, it left place names all the way from Iberia to Ireland. Portugal. Galicia. Gaul. Pays de Galles. Cornwall. Galloway. Galway.

Then the Sarasvati river dried around 2000 BC. A military wave moved west into the Near East. The Kassites ruled Babylonia for 500 years. The Mitanni signed a treaty with the Hittite king Suppiluliuma in 1380 BC swearing by Mitra, Varuna, Indra and the Nasatyas. Four Vedic gods, on a Hittite tablet, thirteen centuries before the earliest date the invasion theory allows for the Vedas.

The Greeks came in the same window. Zeus is Dyaius. When Alexander crossed the Hindu Kush he met people at Nysa claiming descent from Dionysus. Some of their heroes, they said, had come this way.

Link to full Substack essay, with sources, in the comments below.

[a16z @a16z]

Cursor, 4 years ago

[उदा_sin @thevitraag]

@aravind arguably no civilization more ungrateful and malevolent than the Western world,they appropriated Indian knowledge, plundered our wealth, and utilized our people in the World Wars,only to leave us impoverished and partitioned,They remain just as racist today as they ever were.

[Astrobase Space @astrobase_space]

EVEREST — India’s first full-flow staged combustion (FFSC) rocket engine.

80-tonne class. Fully integrated. Designed and manufactured in Bengaluru, India.

[Mary @mxryboo]

job literally means “the persecuted” in hebrew

[Jitni Aabadi Utna Fuck @Parikshit_K_]

They're being clever, they know killing 30 ppl in one go will lead to Indian armed forces being involved. So they'd rather kill 2-3 ppl in 6-7 incidents round the year.

[Sourav Das @100rav009]

@HOVERCRAFT50167 Root

[CA Vivek Khatri @CaVivekkhatri]

🚨Last night, India switched on a reactor.

Here are 9 numbers nobody is talking about:

→ 72 years: Time since Homi Bhabha conceived this plan
→ 22 years: Time to actually build it
→ ₹7,700 crore: Final cost (started at ₹3,492 crore)
→ 500 MW: Power it will generate
→ 2nd: India's global rank only Russia had this before
→ 25%: India's share of world's thorium reserves
→ 400 years: How long those reserves can power India
→ 200+: Indian companies that built it. Zero foreign designs.
→ 3: Countries that tried and quit - USA, Germany, UK

🧵 A thread that will blow your mind:

[Sourav Das @100rav009]

@AshtonForbes This will age like milk.

[Sourav Das @100rav009]

@mufaddal_vohra

[Out Of Context Cricket @GemsOfCricket]

97, 89, 89 >>>>> 100, 100, 100

[Mathonymics @Mathonymics]

The Madhava series (also known as the Madhava–Leibniz series) is the infinite series for π/4:

π/4 = 1 − 1/3 + 1/5 − 1/7 + 1/9 − 1/11 + ⋯

It is the first known series representation of π and is named after the Indian mathematician Madhava of Sangamagrama (c. 1350–1425), who discovered it in the 14th century.

[Champ✨ @Ib_ra_himm]

In a court of justice, both parties know the truth... it is the judge who is on trial.

[Young⁠ @Brayohki]

if you pay attention to the patterns of your life, you'll realize everything always works out. everything always takes you to a greater destination. you always grow and the things you think you can't survive, you somehow divinely make it through. that's life. always remember that

[atharv @notathrv]

> invented JavaScript
> co-founded Mozilla
> built Brave to block ads by default
> made the internet faster
> said privacy isn’t optional
> didn’t sell out to surveillance capitalism

absolute legend 🐐

[Aakash Gupta @aakashgupta]

Everyone’s missing the real story here.

This isn’t a “prompting technique.” The paper is called Recursive Language Models, not “Recursive Meta-Cognition.” And the authors aren’t random researchers grinding for citation counts.

Omar Khattab created DSPy, which has 31,000+ GitHub stars and changed how people build compound AI systems. Tim Kraska pioneered learned index structures with Jeff Dean at Google and runs MIT’s Data Systems and AI Lab. Alex Zhang is a PhD student working at the intersection of both.

The actual paper addresses a specific problem: LLMs degrade on long context. The graph shows GPT-5 performance collapsing as input length scales from 2^14 to 2^21 tokens, while RLMs maintain stable performance. They’re handling inputs 100x beyond context windows.

The key insight from Khattab’s own Twitter: “Most people misunderstand RLMs to be about LLMs invoking themselves. The deeper insight is LLMs interacting with their own prompts as objects.”

This extends Khattab’s entire research arc. DSPy turned prompts into programmatic modules. ColBERT made retrieval smarter. RLMs turn context itself into something the model can manipulate like data in memory.

Prime Intellect, one of the leading decentralized AI labs, is already building on this. They wrote that RLMs will let them “teach models to manage their own context end-to-end through reinforcement learning” for agents running across weeks or months.

The real trade? This solves a constraint that every AI lab faces: context windows are a hard ceiling. Expanding them through architecture and training is expensive. RLMs offer an inference-time approach that works with existing models.

But watch for the catch: this requires models that can write and execute code reliably. The benchmarks use GPT-5 in a Python REPL. Models with weaker code generation will struggle to implement the recursive decomposition cleanly. The technique scales with code capability, not just reasoning capability.

[Historic Vids @historyinmemes]

Anthony Kiedis & John Frusciante singing "Under The Bridge" on a boat (Amsterdam, 1991)

[Women Posting W's @womenpostingws]

IM OBSESSED I’ve watched this on a loop 5 times