Simply AI

@IREN_Ltd IREN just dropped a bomb! 490MW secured in Spain for AI cloud Europe expansion is massive. LFG!

Our vertically integrated AI Cloud platform has arrived in Europe. Today, IREN announced it has completed the acquisition of Nostrum Group, adding 490MW of secured power in Spain to serve global demand for AI compute. “Europe is one of the largest and fastest-growing markets for AI infrastructure, and Spain is among its most compelling entry points, with abundant renewables and strong fiber connectivity. Nostrum gives us secured power today along with a development pipeline and a great local team we're excited to work with.” - @danroberts0101 Learn more: iren.gcs-web.com/static-files/9…

@mattpocockuk forces clarity before chaos. Early “grill” phase saves months of drift. finally, structure that matches how great products actually get built.

Here are my 7 phases of AI-powered development. I've been thinking that the pre-PRD phase needs more structure. You need to figure out the shape of the design tree first, before then walking down it with higher-fidelity prototypes. In other words, /grill-with-docs needs to change again IMO

@Voxyz_ai loops orchestrating 10k+ agents beats hand-coding forever. the real AI shift is here.

boris cherny says he hasn't handwritten a line of code in eight months. some days he's managing tens of thousands of agents at once. doesn't prompt claude anymore, writes loops that prompt claude. a working loop needs 6 pieces: - automations: PR opens / file changes / scheduled triggers. doesn't wait for you to remember - worktrees: each agent gets its own workspace, parallel without conflicts - skills: write project knowledge once, every loop starts with context - connectors: plug into github, linear, databases. agent opens PRs, links tickets, watches CI - subagents: separate the writer from the checker. reviewer is a different agent - memory: models forget, repos don't. tracks what you tried and what passed across sessions start with closed loops. bounded, with stop conditions, cheap and reliable. don't go open-ended first, that burns tokens and drifts. a reliable loop beats a perfect prompt.

@_The_Prophet__ this is the real red pill. owning cognition beats renting it empires form now.

⚡️The AI race is already past the point where most public analysis is useful. Most people are still debating chatbots, job loss, copyright, model benchmarks, or whether AI is “overhyped.” That debate is downstream noise. The actual contest is over who owns scalable cognition before everyone else realizes cognition became infrastructure. A frontier model is a generator of leverage across every domain that depends on reasoning, code, research, persuasion, planning, pattern recognition, coordination, and institutional memory. Once a system can compress expertise and automate parts of judgment, it becomes a meta-asset. It does not compete with one industry. It changes the production function underneath all industries. That is why the window closed so fast. The frontier was never going to stay open. Open windows exist in early technological regimes when the stack is still cheap enough, ambiguous enough, and socially underestimated enough for new entrants to assemble the pieces. Then the stack hardens. Capital requirements rise. Talent consolidates. Feedback loops compound. Infrastructure gets territorial. Regulation enters. Security enters. Incumbents weaponize distribution. Governments discover the strategic layer. That is where AI is now. The frontier is no longer a garage race. It is an empire race. Compute, power, chips, data centers, model talent, security review, government access, cloud distribution, product telemetry, evals, synthetic data, and capital markets now form one integrated machine. A late entrant cannot simply buy GPUs and hire smart people. It has to recreate a living factory while the incumbents are using their own models to improve the factory. The leading labs now have recursive acceleration. Better models help write better code, find better data, generate better evals, improve agents, automate research support, harden cyber defenses, compress workflows, and shorten the next cycle. The advantage is no longer only scale. The advantage is compounding cognition inside the organization. That is why the enterprise loop matters. Companies that only rent intelligence are going to bleed their edge upward. Their workflows, decisions, customer interactions, corrections, and domain patterns become raw material for someone else’s layer. Over time, the firm becomes hollow: still branded, still staffed, still operating, but less differentiated. Its judgment has been extracted, normalized, and resold. The winners build private cognition loops. They turn their internal work into proprietary learning systems. Their human experts become signal generators. Their agents become execution surfaces. Their evals become memory. Their workflow traces become training data. Their corrections become compounding judgment. Their institutional knowledge becomes machine-operable capital. That is the new moat. The same logic applies to nations. Countries that own frontier cognition gain strategic leverage. Countries that rent it become dependent on foreign model policy, export controls, cloud access, censorship rules, security filters, and geopolitical bargaining. They may still “use AI,” but usage is not sovereignty. Renting the nervous system of another civilization does not make a country sovereign. That is the part most governments missed. Europe thought the key move was regulation. The U.S. and China understood, imperfectly but more correctly, that the first move was capacity. Build the labs. Build the compute. Build the clouds. Build the chips. Build the energy layer. Build the talent density. Then regulate from strength. Regulation from weakness becomes permission theater.

@hnshah This is the real moat in AI: owning your org's learning loops. Gold.

Satya’s post is worth reading closely because it gets at the real AI question for companies. Who captures the learning? His argument is that companies are becoming a new kind of learning system. People bring judgment, taste, relationships, context and ambition. AI brings scale, memory, reasoning and execution. The value comes from building a loop where the company gets smarter every time work happens. The important asset is the learning system around the model. That system is built from the record of how work actually gets done. Workflow traces show the path people take. Corrections reveal judgment. Accepted outputs show what good looks like. Rejected approaches sharpen the standard. Private evaluations, domain-specific context and institutional memory give that learning structure. Over time, the company starts to retain more of what used to disappear inside meetings, edits, comments, decisions and individual experience. That is the learning loop Satya is pointing at. The judgment that once lived in a few people’s heads can become part of how the company operates.

@LinusEkenstam totally agree—AI cranks out 100k+ reps while humans doodle 50. Iteration on steroids. skill explosion incoming.

Creatives using AI is making ten orders of magnitude more output than this. Know multiple people that has done 100.000+ generations last 12 months. If repetition is one of the things that improve your skills. AI is most definitely super charging that, and critical thoughts

@chandrarsrikant This US move proves it: AI sovereignty is national security. IndiaAI Mission is spot on—build our own, own our future

🚨When AI access disappears overnight: The 12 sovereign models India is betting on On June 12, Anthropic said it would "abruptly ​disable" its most advanced AI models, including Fable 5 and Mythos 5, for all users after the US government ordered it to suspend access for foreign nationals, citing national security ‌concerns. For India, this has become one of the strongest arguments yet for investing in domestic capabilities. Under the Rs 10,371.92-crore IndiaAI Mission, the government has shortlisted 12 organisations and consortia to build foundational models tailored to Indian needs. The projects span large language models, small language models, voice systems and multimodal AI capable of handling text, images and video. @AihikS with the progress so far 👇🏽 moneycontrol.com/technology/whe…

@Kimi_Moonshot 260 tok/s on code? Open-source just went supersonic.

🌘 Meet Kimi K2.7 Code HighSpeed! A high-speed mode of our latest open-source multimodal coding model, Kimi K2.7 Code. ⚡️ Up to 6× faster: Around 180 tok/s on coding tasks with median-length inputs, and up to 260 tok/s on shorter-context tasks. 🔷 Rolling out to Kimi Code Beta Program members, Kimi API developers, and Kimi Business users. (Access will remain limited for now due to capacity constraints.) 🔷 No invite needed. Anyone who joins the Beta Program has a chance to get access 👉 kimi.com/code/beta Open intelligence should be instant, affordable, and borderless. We'll continue improving the model and expanding access as more capacity becomes available! 🔗 Kimi Code: kimi.com/code 🔗 API: platform.kimi.ai

@dickiebush Procrastination hack gold! Voice dump + AI micro-steps starting in 5s = instant momentum.

How I kill procrastination with AI: 1. Pull up voice mode and brain dump what I’m supposed to be working on. 2. Ask it to break the task down into 25 steps with the first one taking less than 5 seconds. Poof, now I have a checklist to crush. Instant momentum.

@mark_k @xai Reliability tweaks and diagram fixes make Grok even more unstoppable. Great work!

Grok Build 0.2.52 by @xai is out, with a pretty useful set of workflow and reliability improvements. Most important changes: • Agent sessions now respect client auto-approve settings for tools • ER diagrams now render directly in the TUI instead of showing raw source • Team logins now support centrally managed config and requirements files • More built-in sandbox skills for document and media work • MCP reconnect spam on transient stream errors is fixed • Ctrl+C cancellation now completes in under 50 ms instead of sometimes blocking for up to 2 seconds • Mermaid and ER diagram rendering got multiple fixes, including cardinalities, dark themes, PNG export, and CJK text The theme here is reliability. Faster cancellation, fewer reconnect loops, better paste behavior, and diagrams that actually render. Small fixes individually, but exactly the kind of polish that removes friction from daily agentic coding.

87,000 jobs lost to AI in the US... in just the first 5 months of 2026. That’s more than the entire year of 2025. Companies aren’t just using AI to help employees anymore — they’re replacing entire teams with it. Marketing, coding, support, design… the list keeps growing. This is no longer a slow trend. It’s accelerating fast. The AI layoff wave is real, and it’s getting bigger every month. Are we ready for what’s coming? Or are we still pretending this won’t affect most industries? 👇

Jeff Bezos just called AI pessimism “the opposite of reality” and he’s backing it up with action. He’s launching a new AI company called Prometheus. Bezos believes we’re still extremely early in AI and that the biggest opportunities are still ahead. Instead of worrying about the risks, he’s doubling down on building the future. Classic Bezos move: while others are debating dangers, he’s investing heavily and moving fast. Love the optimism or think he’s too naive? 👇

@kostascrypto spot on. nations that slash taxes on AI infra & talent for 10+ years will dominate the century. time to choose: lead or lag."

If you're a king, queen, president, prime minister, emir, sultan, or any leader of a country, make the development of innovative LLM infrastructure, AI algorithms, and supporting hardware TAX-FREE for at least the next decade. Not for companies that simply use AI, but for those building better AI models and the infrastructure behind them. Provide incentives to bring back expat talent. If you do, your country might have a chance to survive what’s coming.

@jukan05 AI is reversing 60+ years of DRAM deflation. HBM + AI demand = new semiconductor paradigm. Bullish!

The shift brought about by HBM: - From 1957 to 2020, DRAM cost per Gb declined by roughly one order of magnitude every five years, making it one of the clearest examples of Moore’s Law in cost terms. - However, demand for AI infrastructure and the emergence of HBM have directly overturned the cost-reduction pattern that had persisted for decades.

@OrevaZSN convenient hypocrisy: 'Jobs!' only when it protects the powerful. AI layoffs? Just progress. time to call out the double standard."

@mark_k EU's frontier AI? more like frontier censorship. woke guardrails will kill truth-seeking. Grok shows what's possible without bureaucracy.

The gloomy realization that even if the EU were to train a frontier AI model, it would be absolutely woke, censored, and unusable. 🫩

@RodmanAi this changes everything. stealth AI in plain sight bosses will never know. game over.

NOBODY IN THE ROOM KNOWS YOU'RE USING CLAUDE 😳 I genuinely thought this was a normal Google Doc. Turns out, it's Claude. Someone built a Chrome extension that makes ChatGPT and Claude look exactly like a document editor. You type a question. The AI answers right inside the page. To everyone around you? It just looks like you're working. A few wild details: • Works with ChatGPT and Claude • Looks like Google Docs by default • Can mimic Notion and Microsoft Word too • Supports multiple AI models • Everything happens inside the fake document The extension is called GPTDisguise. We're entering an era where AI isn't hidden in another tab. It's hiding in plain sight.

@JamesMelville Scotland: run like the plague from data centres. Virginia’s nightmare water wars, sky-high bills, ruined landscapes is coming for you. Preserve your beauty.

Anti AI data centre campaigner from the USA gives a warning message to Scotland: “If you want to keep Scotland beautiful, you should run like a plague is approaching you.”

@BitcoinArchive This is massive! AI agents already moving $50M in crypto. Machine-to-machine economy is live.

COINBASE CEO: AI agents on Coinbase have completed 100 million transactions, totaling about $50 million in Bitcoin and crypto

@suraj_sharma14 India has the talent & infra edge. IndiaAI Mission + Sarvam scaling fast. compute & capital catching up frontier labs incoming soon!

Serious question: Why doesn't India have a frontier AI lab yet? > Wrong incentives? > Not enough compute? > Brain drain? > Not enough long-term capital? Or is frontier AI simply harder than most people realize? Curious what builders think.