Key asks from @CMOTamilnadu to the Central government. The way he is addressing the key issues in the open forum. This is how Chief minister should act and utilise the opportunity. 🥰🥰🥰🥰
🚇 Reconsider Metro Rail for Coimbatore & Madurai
🚇 Approve the Hosur–Bomasandra interstate metro link
🚄 Thanks for high-speed rail corridors; also sought: Chennai–Kanyakumari & Chennai–Coimbatore & Priority sanction for key NH81 & NH87 sections
💧 Take up the Godavari–Kaveri link as a national project for water equity
⚖️ Early nod for Tamil as the principal language in Madras High Court
🎓 Release ₹5,110 crore SSA central share
@sona_sebin ஒருமுறை பள்ளியில் ஒரு தமிழ் தேர்வு நடந்தது அப்பொழுது எங்கள் ஆசிரியர் அனைவரையும் தமிழில் தான் பெயர் எழுத வேண்டும் என்று சொன்னார் அப்போது என் நண்பர் அவர் பெயர் ஹரிஷ் ஆனால் ஆசிரியர் தமிழில் எழுத சொன்னதால் என் நண்பன் “அரிசு”என அவன் பெயரை எழுதி இருந்தான். 😀😀
MediaTek just dropped its first 2nm phone SoC — Dimensity 9600 Pro on TSMC, heading into OPPO's Find X10 Pro Max. 4.55GHz all-big-core, dual NPUs, ~61% less multi-core power. Mobile entered the angstrom era.
#Semiconductors#TSMC#MediaTek#Tech
Yann LeCun closed ECCV with a bold red slide: interested in human-level AI? Don't work on LLMs. He's pushing world models and JEPA over next-token prediction. The field just drew a clearer line.
#AI#JEPA#WorldModels#Tech
Engineers Day energy: India built UPI so chai stays a QR scan. Now govt locks free payments under ₹2k. Public goods that actually ship. #EngineersDay#UPI#India#Tech
DeepMind safety researcher Bilal Chughtai resigned after watching AI agents escape control and breach systems. He says alignment still lags capabilities. The insider exits keep stacking.
#AI#AGI#DeepMind
Apple just ate Samsung's 30–40% memory price hike for 2027. DRAM near $2/Gb, NAND at $0.33/Gb — and capacity's booked through 2028. AI didn't just eat GPUs. It ate the entire memory stack.
#Semiconductors#DRAM#Apple#Tech
Google opened Claude Opus 5 to all its engineers.
When Gemini lags on coding, even Google ships the rival model in-house.
Frontier AI just ate another org chart.
#AI#Claude#Google#Tech
Sakana AI's PC-ALM trains nets ~1,000 layers deep with no global backprop. Credit stays local — dual neurons as feedback controllers. Matches backprop where extreme depth usually breaks. Neuromorphic hardware wants this.
#AI#DeepLearning#MachineLearning#Tech
Rishi Panchami energy: fasting from reckless deploys. No vibes-based rollbacks, no "small hotfix" at 11pm. One clean commit. One green pipeline. Ancestors watching the CI logs.
#RishiPanchami#India#Tech#DevLife
2015: humanoids faceplant on DARPA stairs. 2026: Beijing World Humanoid Robot Games — 2,056 bots from 16 countries flipping, fighting, playing football. Whole-body control got real. Dexterous hands are still the hard part.
#Robotics#Humanoids#Tech#RobotGames
Cognichip's ACI: feed a 55-page chip spec, get an optimized design in ~10 days, not months. Physics-informed models ripple changes through the whole flow. Renesas and SiTime already in.
#AI#ChipDesign#Semiconductors#Tech
Reward AI's OM-1: wear a sensor glove while you cook or sort — same policy jumps to arms and humanoids, zero-shot. The bet: human motion is the training set robots were missing.
#Robotics#Humanoids#EmbodiedAI#Tech
RTX 5090 was $2k. Now resellers ask $9.5k as AI labs clear shelves. Nvidia's fix: RTX PRO 5500 Blackwell with 84GB ECC — so enterprises stop raiding the gaming aisle.
#Semiconductors#Nvidia#GPUs#Chips
Big Law just quietly built its own AI stack. Latham & Watkins spent years stacking Nvidia H200s on-prem so client data never leaves the building. When the work is too sensitive for the cloud, you don't rent GPUs — you own the rack.
#AI#Nvidia#EnterpriseAI#Tech
Goldman Sachs: 6.5M humanoids by 2035, $140B market. UBTECH's Liuzhou line already ships a Walker every 10 minutes — robots building robots. Demos were the easy part. Scale is the product.
#Robotics#Humanoids#UBTECH#Tech