Vijay krishna

@Holasoygeorgee @songjunkr Opus 4.7 is shit and expensive AF

@songjunkr When they released those prices, Grok was the best AI model up to that point, but it isn't anymore. Even Elon said they're two months away from reaching the level of Claude Opus 4.6; I suppose they're four months away from 4.7. It's disappointing, really.

욕하는건 아니고, 진짜 순수하게 궁금해서: 왜 Grok이 가장 비싼거야?

Apple just made a quietly stunning admission: its own Siri engineers need to go back to school. According to a report from The Information, the company is sending close to 200 members of the Siri organization to a multi-week bootcamp, where they will learn how to code using AI tools like Claude Code and Codex. Roughly 60 engineers stay behind to keep core development running, and another 60 handle evaluations and safety checks. This retraining wave lands just two months before WWDC in June, where Apple plans to unveil the long-delayed, Gemini-powered Siri overhaul.

🚨BREAKING: 🇮🇳Indian IT giant Wipro will not hire freshers from colleges for now as AI changes its workforce needs.

@jontwigge @kimmonismus I think you don't know how important this test is

@kimmonismus Clearly an indication we are close to AGI - why waste time looking for needles in haystacks, there is code to be written.

Hold on, something doesnt add up here. Opus 4.7 got much worse in needle in the haystack? need to dig into this

Give a man an LLM and he'll know everything. Give a man 2 LLMs and he won't be sure of anything.

BREAKING: 🇮🇳 India plans to build a massive cyber intelligence system to defend government networks from “state-sponsored” hackers. The project will collect and analyze huge amounts of digital data (petabyte-scale logs) across government systems to detect attacks early and respond faster. Source: Moneycontrol

Perplexity users just got nerfed. Deep research was unlimited. Now it's 20 a month. Normal search is gated to 200 a week. No email. No announcement. Nothing. What I subscribed to is not what I got. Asked for refunds. Denied. Complained on Discord. Banned. The moderator called it 'responsible evolution'. You paid for unlimited. You got robbed. And they won't even let you complain about it.

META DETHRONING GOOGLE AS WORLD’S LARGEST AD BUSINESS >google ad revenue: $239B >growth 11.9% and flatlining >meta ad revenue: $243B >growth 24.1% and accelerating META’s AI bets are already printing money: >ai boosted reels watch time up 30% >ai slop tools for advertisers printing $10B/year The Zucc is winning.

🚨 CCTV companies to sell only made in India products in India from this month.

🚨 Google researcher and a Turing Award winner published a paper that exposes the real crisis in AI. Tech industry is locked in a trillion-dollar race to build faster GPUs with insane amounts of raw compute. But this paper shatters this strategy. LLM inference is fundamentally different from training. During the "decode" phase, when an AI actually generates its answer, the model doesn't need to do more math. It needs to move data. The paper makes it blunt: The primary bottleneck for AI is memory and interconnect. Not compute. Imagine having a super-engine in a Ferrari, but trying to suck all the fuel through a tiny plastic straw. The engine isn't the problem. The straw is. Researchers propose a total paradigm shift. Instead of just stacking more raw compute, they outline four new architectural directions: - High Bandwidth Flash for 10X memory capacity. - Processing-Near-Memory (PNM). - 3D memory-logic stacking. - Low-latency interconnects. The era of relying entirely on massive, centralized cloud hardware is shifting. To get real efficiency, the kind that eventually lets you run complex, open-source models locally instead of constantly burning expensive API credits, we don't need a bigger calculator in a server farm. We need to fundamentally redesign how the hardware handles memory. For the last three years, we worshipped the GPU. But the next massive leap in artificial intelligence won't be won by whoever can do the most math.

The average 4-year-old laughs 300 times a day. The average 40-year-old laughs just four times a day.

@beknabdik @tembakhumalo @OfficialLoganK @SierraPlatform it can be optimised if they focus on it

@tembakhumalo @OfficialLoganK @SierraPlatform this is a bigger deal than people think. voice generates way more tokens than text because speech is verbose and redundant. context fills up fast. persistent memory across voice sessions is a real engineering challenge not just a missing feature toggle

Our latest Live model is # 1 on Tau Voice Bench! Excited to see this new frontier of voice models cross the chasm of usability in production.

You’re roughly half human. Your body is more bacteria than human 🦠 You carry ~38 trillion bacterial cells
vs ~30 trillion human cells That means you are only ~43% human by cell count. You’re not just one organism. You’re an ecosystem.

OH: some dude in Fujian (scam capital of China) uses hacked credit cards to subscribe to Claude Code Max, resell them as API at 20% cost, and then sell user traces to Chinese labs for distillation. Bruh.

Dubai has a massive problem. Iran will not open the Strait of Hormuz to the UAE. And most likely they will not open it for years. Now the UAE is demanding compensation from Iran for the damages. They said they'll keep attacking Iran unless they get paid. Highly unlikely Iran would pay a cent. Yesterday, the UAE bombed an oil refinery on Lavan Island. Now Iran will most likely keep that strait closed to the UAE. UAE's leaders are pressuring the UK and a few countries to fight Iran and take control of the Hormuz. Keir Stramer is in the UAE as we speak. This approach most likely will backfire. The UAE'S entire economy runs on oil flowing through those waters. Airports will not operate without fuel. Hotels could shut down. Tourism has already collaosed. Empty beaches. Empty malls. Empty stadiums. Every single thing could crash.

Do you understand what this means? For the first time, an Open Weight models is #1 on CyberSecutity. Sure there’s Mythos but we don’t have it :p Don’t let them trick you

Claude Mythos. Ten trillion parameters: the first model in this weight class. Estimated training cost: ten billion dollars. On the hardest coding test in the industry (SWE bench) it scores 94%. It found a security flaw in a system that had been running for 27 years, one that every human engineer and every automated check had missed. It found another bug that had survived five million test runs over 16 years. (It did so overnight.) It is so capable in cybersecurity that Anthropic will not release it to the public, instead it is launching Project Glasswing along with 100m in compute credits to help secure software. Only twelve partners currently have access: Amazon, Cisco, Apple, Google, Microsoft, NVIDIA, JPMorgan Chase, Crowdstrike, Palo Alto, AWS, The Linux Foundation, Broadcom. (I'm sure the Pentagon is on the line?) This is not a product launch: it is a controlled deployment of a system too powerful to distribute freely. Tell me this isn't (very expensive) AGI?

This is hilarious Clown Pakistan screwed the US🤣 ▪️ US wanted a ceasefire but didn’t want to look like it was losing, so a plan was prepared. ▪️The plan was that Pakistan would beg the USA to stop the war. ▪️The US would then accept Pakistan’s request and agree to a ceasefire. ▪️A draft message was prepared in the White House and sent to Pakistan’s PM Shahbaz Sharif to post. ▪️But Shahbaz Sharif made a huge mistake. He forgot to remove the "draft message for Pakistan PM" header. ▪️He just copy-pasted it as it was and completely spoiled the entire US-Pakistan backdoor plan. ▪️Now Donald Trump is searching for Shahbaz Sharif with a bamboo stick in his hand 🤣🤣

JUST IN: 🇷🇺🇮🇷 Russia provided Iran with 55 key energy infrastructure targets inside Israel, Jerusalem post reports.

BREAKING : Iran FM humiliating Trump 🔥 🇺🇸 Trump –– You have 48 hours, fulfill our demands 🇮🇷 Iran FM –– "We won't do that" 🗿 🇺🇸 Trump –– Okay let's take 5 days and make a deal together 🇮🇷 Iran FM –– "We aren't stopping here" 🗿 🇺🇸 Trump –– Let's do one month ceasefire 🇮🇷 Iran FM –– "Not needed" 🗿 🇺🇸 Trump –– Okay, then we will send US troops to Iran 🇮🇷 Iran FM –– "Come closer" 🗿 🇺🇸 Trump –– You will have to open the Strait of Hormuz at any cost 🇮🇷 Iran FM –– "Okay. Pay us all the compensation for our losses in the war" 🗿 🇺🇸 Trump –– You are trying to dictate us 🇮🇷 Iran FM –– "Better obey or we will close the Strait of Bab el-Mandeb too" 🗿 Iran and Abbas Araghchi have made Trump a global laughing stock 🤣🤣