Pramode C.E @[email protected]

this guy has 29 models on huggingface at page 2 ranking. no lab behind him. no sponsorship. $2,000 from his own pocket on GPU rentals. he compressed GLM-4.7 to run on a MacBook and quantized Nemotron Super the week it dropped. all public. all free. nvidia is a trillion dollar company with hundreds of teams but they are not the ones quantizing models middle of the night and pushing them out before sunrise. if nvidia stopped tomorrow their employees stop working. people like @0xSero would not. that is the difference between a paycheck and a mission. @NVIDIAAI you talk about making AI accessible. the people actually doing it are right here. 29 models deep burning their own compute with no ask except more hardware to keep going. you do not need to build another program. just look at who is already building for you. one GPU to this man would produce more public value than a hundred internal sprints. i am not asking for charity. i am asking you to invest in someone who already proved it.

No prompt (sorry to disappoint!). We do our own stunts. No agencies (or agents). All inhouse. Everything you see was handsculpted in Blender by Fabian on our Logic & Magic team, with Fabio rendering this with Zig, WASM and WebGL and Joy illustrating the 3D model of the TigerBeetle character. In fact, everything at TigerBeetle is handcrafted. Just like the DBMS itself, per TigerStyle. Even the blog post art, all hand-drawn (if you look closely, you’ll see it’s always signed by Joy Machs). The quality is worth it.

IMSc announces Science at the Sabha 2026 on 29 March, 4-8pm at The Music Academy, Chennai. For details and registration:  imsc.res.in/outreach/scisa… @SciSabha #SatS2026

Helium is the only element that escapes Earth’s atmosphere permanently. Once released, it rises through the troposphere, passes the stratosphere, and leaves the planet. It cannot be manufactured. It cannot be synthesised at industrial scale. It accumulates over billions of years in the same geological reservoirs as natural gas. And one third of the world’s supply just went offline because Iran hit the facility that extracts it. Qatar produced roughly 63 million cubic metres of helium in 2025, accounting for 30 to 36 percent of global supply from a total of approximately 190 million cubic metres. QatarEnergy’s three large helium purification plants at Ras Laffan form the world’s biggest helium production base. When LNG production stopped after Iranian drone strikes on March 2 and the subsequent missile damage on March 19, helium extraction stopped automatically because helium is recovered during natural gas liquefaction. You cannot produce helium without producing LNG. The byproduct dies with the primary product. Spot helium prices have roughly doubled since the crisis began. Industry consultants warn that prolonged disruption could push contract prices toward $2,000 per thousand cubic feet. A major industrial gas supplier has already begun assessing customers a helium surcharge. Phil Kornbluth, the most cited helium market consultant, stated the assessment directly: the world cannot compensate for the loss of a third of its helium supply. South Korea imports 64.7 percent of its helium from Qatar. SK Hynix and Samsung operate high-volume fabs producing the DRAM and high-bandwidth memory that power every AI accelerator, every data centre GPU, and every cloud computing cluster on Earth. Helium cools silicon wafers during fabrication. It serves as a carrier gas in deposition and etching tools. It enables leak detection in vacuum systems. Modern extreme ultraviolet lithography requires helium-cooled environments for precise temperature control. Without helium, the fabrication process degrades or stops. SK Hynix and Samsung hold two to three months of helium inventory. Two to three months is not a buffer. It is a countdown. If Ras Laffan remains offline beyond that window, South Korean memory production faces rationing. TSMC in Taiwan is somewhat more diversified but still uses Qatar-linked supply chains. The entire AI hardware supply chain, from HBM3E memory stacks to advanced logic chips, sits inside helium-dependent ecosystems. Beyond semiconductors, helium cools the superconducting magnets in more than 14,000 MRI machines operating worldwide. It pressurises rocket fuel tanks and purges propulsion systems in aerospace. CERN’s Large Hadron Collider depends on helium cryogenic systems. There is no substitute for helium in any of these applications at industrial scale. The United States and Qatar together account for more than 70 percent of global production. The US federal helium reserve and private suppliers offer partial relief, but global prices and spot availability are still governed by Qatar’s market share. Japan’s Iwatani has drawn on US reserves. Canada and the Rockies are seeing renewed investor interest. None of this replaces 63 million cubic metres in weeks. The war hit uranium first. Then oil. Then nitrogen. Then water. Then plastic. Then medicine. Then sulfur. Now helium. Eight layers. Each one deeper. Each one closer to the infrastructure that sustains modern civilisation. The chip that processes your data, the magnet that scans your body, and the rocket that launches your satellite all depend on an atom that leaves the planet when you lose it. open.substack.com/pub/shanakaans…

Between 2015 and 2022, India certified 1.10 crore people under its flagship skilling programme and spent ₹9,261 crore doing it. A CAG audit later found that most of the data behind those numbers was not real. 🧵

To make something beautiful, you need deep attention and care. Modern mainstream computers, internet, and now AI because of a clusterfuck of reasons is designed to make our attention wander. Now, with disposable software, a slopwave that perniciously erodes beauty has started.

The 2024 Nobel Prize in physics went to a computer scientist. Now, in turn, the 2025 Turing Award was given to a physicist!

Found a dude who is going all out in sloppifying the whole of github by opening PRs across a huge number of repos, all autonomous and unmonitored.

Indian astronomy is growing by leaps and bounds. The excellent work being done in our country needs to be explained in an accessible, yet scientifically accurate way. @starlabiitb has come up with a wonderful initiative to do exactly this.

We're announcing a new conference for AI Scientists where both authors and reviewers will be AI systems. This is to probe the limits of our current LLMs/agents. Accepting papers starting April 15 👇

💯 this. To generate anything non-slop with agents, you need to baby sit them. At least as of now, there have been so many attempts at spec-as-code or spec-to-code and all of them (like amazon kiro, github spec kit, etc) have fallen flat on their faces. For any non trivial software, an upfront spec cannot cover everything and there are tons of areas that you need to flesh out as you implement. Even the upfront fleshed out things have a lot of discrepancies with agents assuming things that don't even exist however you may do it. There's a reason why specs, even when humans code, do not (and should not) spell out every small tech decision upfront because you need to explore more for each small area independently as things change, software evolves.

While technology side of a startup makes it exciting for us nerds, what lends credibility to that technology are customers Great to see @Airbound_Aero progress, commericalising your tech can often be harder then developing it. God speed to them.

Nicely done 🇮🇳 INDIA ROBOTICS in Singapore🇸🇬. Also a reminder to ALL Indian startups -- the US 🇺🇸 is NOT your default launch market. I first heard of Genrobotics (@GenRobotic) via the Kerala Startup Mission a few years ago but never ended up connecting. Seeing this now genuinely makes me smile. Such companies just stay underground (Pun intended) They’ve stayed focused on building a solid product in a niche space. No noise, no hype, just execution. Bandicoot robots going into sewers and manholes, taking on the dull, dirty and dangerous work. This is EXACTLY what robots should be doing. They’ve raised just $6M from @Zoho , @anandmahindra , @unicornindia Ventures, Sea Fund and others, and now gone on to win a Singapore government contract after outperforming 600 global competitors. These guys went where the problem existed, built for it, and won. This is how you build. Read more here: tinyurl.com/3pbjk4kr

What enterprise AI acceleration looks like

@nutanc Looks like Salman Khan himself is very positive about AI's role in education (that of course does not justify what the AI companies are doing ...).

Most of the educational training material is without a doubt from Khan Academy. There is only one reason his life's work is obsolete. It's because the AI companies do not cite his work and don't tell the people that it's trained on his data. Basically, in layman terms they stole his data. And that's why his life's work is obsolete.

The problem that AI brings is to the Computer Science degree is this. The days of getting a CS degree from mediocre colleges by paying a lot of money and then dancing in a towel to “mere khwabon mein jo aaye” through your four years and then landing a “business analyst” job checking in requirements for delivery to customer are OVER. Or to quote CS graduates from my era in an Ayesha Jhulka rain: Yeh mausam bhi gya, woh mausam bhi gya

Tariffs, shipping costs and manufacturing ops - can often be too much to handle for an indie developer Which is why I’m super excited to see how bunzee-labs.com turns out Buy the designs, get it manufactured from your local PCBA shop - could be JLC too if the shipping to your country is subsidized Great hardware designers are everywhere and we need simpler distribution for their products

The only metric for quality as far as college management is concerned are marks scored in exams. They are not going to do anything to improve quality or provide better exposure. Students will have to take it upon themselves to do better ... hope this spreads to more colleges!!

@HardikSeth69 It is easy these days when you have AI 😀Anyways, I guessed it first and verified with gemini, got it right!

Can you guess what this PCB does? Hint: You probably use this every day with a remote.

🦀 New Video Drop! – Rust compiler series: promoted constants -- wait, where did my range go? Finally found time to start a mini series, to cover interesting tidbits about the rust compiler I happen to learn about while working with Rust's MIR and a custom codegen backend.