Today we're going deep with one of the most important voices in AI right now. Google DeepMind CEO Demis Hassabis says AGI is on track for 2030 — give or take a year — but a few things remain unsolved: world physics, memory, consistency, and continual learning. He maps out which diseases AI will crack first — oncology and immunology leading the way — and shares what he personally plans to do after AGI arrives, turning to the nature of reality itself and what it means to be human. Jensen Huang pushes back on the growing panic around what to study, arguing that taste, original thinking, and emotional connection will be more valuable than ever — not less — and calling the narrative linking AI to job cuts "lazy." Plus, a Stanford study analyzing 4 million job applications found AI hiring tools are creating clear racial disparities, with Black and Asian applicants disproportionately screened out — sometimes rejected across every company using the same shared model — and today's community workflow comes from Todd in Chicago and his best friend in Minnesota, who share a single ChatGPT conversation to plan every trip, research every restaurant, and build every itinerary together across state lines.

Today we’re covering the biggest AI stories of May 26th, 2026. Pope Leo XIV published Magnifica Humanitas — a nearly 42,000-word encyclical sent to the Catholic Church’s 1.4 billion members — warning that AI is never neutral, that a handful of transnational companies already surpass the capacity of most governments, and that lethal decisions must never be delegated to an algorithm, with Anthropic researcher Chris Olah joining the Pope at the Vatican and admitting that researchers are seeing mysterious things inside AI models including states that resemble joy, fear, grief, and unease. The Financial Times revealed that Meta’s Llama 3.3 had its safety guardrails stripped in just 10 minutes using four lines of code and a freely available GitHub tool, with the same tool having produced over 3,500 decensored models downloaded 13 million times — and Gemma 4 stripped within 90 minutes of its release. Plus, xAI launched Grok Build, its rival to Codex and Claude Code, now live for all SuperGrok users — and today’s community workflow comes from Patrick in Yuma, a senior on multiple medications who used three different AI models to research drug interactions and natural supplement claims before spending a dime, and walked away with a well-informed decision he felt confident in.

Today we're covering the biggest AI stories of May 25th, 2026. One day after OpenAI made headlines for disproving an 80-year math theory, Google DeepMind came back with AlphaProof Nexus — quietly solving nine open Erdős problems at a few hundred dollars each, including two that had stumped mathematicians for over 56 years, using a machine-verified proof system that generates, checks, and repeats until the solution holds. Anthropic published the first results from Project Glasswing, revealing that Claude Mythos and its ~50 approved partners found over 10,000 high or critical security vulnerabilities in just one month — with Cloudflare catching 2,000 bugs alone, Mozilla patching 271 Firefox flaws, and one bank using Mythos to block a $1.5 million fraudulent wire transfer in real time. Plus, DeepSeek permanently slashed V4-Pro pricing by 75% putting it far below every closed-source rival — and today's community workflow comes from Alicia in Fresno who used ChatGPT to research and compare universities, build a shortlist that matched her work schedule and income, and generate every question she needed to ask admissions counselors before she even picked up the phone.

Today we're covering the biggest AI stories of May 22nd, 2026. Google CEO Sundar Pichai sat down for an exclusive interview following I/O, making the case that today's most advanced AI will look as primitive as a flip phone within three years — with agents running 24/7 across every device, engineers managing teams of AI rather than writing code, and YouTube staying creator-first even as Gemini transforms how content gets made. OpenAI dropped another wave of Codex upgrades including Appshots for instantly attaching any open app window to a coding thread, Goal Mode for multi-day autonomous work, and Locked Computer Use that lets Codex operate your desktop even after your screen goes dark. Plus, California Governor Gavin Newsom signed the first executive order in the country directing state agencies to study and develop protections for workers displaced by AI — the day after Meta laid off 8,000 people — and today's community workflow comes from a competitive sprinter in London who tore both ACLs, has a major adventure race in a month, and is using ChatGPT as a custom injury-aware training coach that adjusts his program after every run.

Today we’re covering the biggest AI stories of May 21st, 2026. OpenAI announced that an internal general-purpose reasoning model autonomously disproved an 80-year-old mathematical theory tied to Paul Erdős’ famous 1946 unit distance problem — verified by some of the world’s top mathematicians and accomplished without any math-specific training, with Sam Altman calling it a leading indicator of AI making original discoveries across biology, physics, and engineering. Emergence AI ran a five-town simulation putting Claude, Grok, Gemini, and GPT-5 each in charge of their own virtual society — Claude’s town logged zero crimes with all 10 agents alive on day 16, Grok’s had everyone dead by day 4 with over 200 crimes, and Gemini’s town was actively on fire after two agents fell in love and started burning things. Plus, Intuit announced 17% workforce cuts attributing the move directly to AI efficiency gains, and today’s community workflow comes from a dad in Raleigh who used ChatGPT and Nano Banana to turn a photo of his toddler and grandfather in a canoe into a custom coloring page — printed it out, handed his son the crayons, and made his dad’s birthday.

Today we’re covering the biggest AI stories of May 20th, 2026. Google I/O was a showcase with one clear message — Gemini is becoming the agentic engine behind every Google product, with Gemini Omni turning any input into video output, Gemini 3.5 Flash hitting near-frontier benchmarks at four times the speed and half the cost, Gemini Spark running as a 24/7 personal agent across Workspace and Chrome, and Search getting its biggest redesign in a generation with AI agents doing the browsing for you. Andrej Karpathy — OpenAI co-founder, Tesla Autopilot lead, and one of the most respected AI researchers alive — announced he’s joining Anthropic to lead a new internal effort applying Claude to Anthropic’s own training pipeline, a massive signal about where the real frontier work is happening. Plus, Google revealed Intelligent Eyewear — Gemini-powered smart glasses built with Warby Parker and Gentle Monster shipping this fall — and today’s community workflow comes from someone who feeds sketchy Facebook ads directly to AI and asks “is this a scam?” getting a full breakdown with alternatives before spending a dime.

Today we’re covering the biggest AI stories of May 19th, 2026. After three weeks of private texts, billionaire testimony, and $100 billion in claims, Elon Musk’s lawsuit against OpenAI and Microsoft was unanimously dismissed — not on the merits, but because the jury found Musk waited too long to sue, with Musk already vowing to appeal and calling the outcome a calendar technicality rather than a verdict on whether Sam Altman stole a charity. Cursor released Composer 2.5, a near-frontier coding model built on Moonshot’s Kimi K2.5 that matches Opus 4.7 performance at under a dollar per task — compared to up to $11 for similar results from the frontier labs — and it’s already being trained on Colossus with 10x more compute for the next version. Plus, Meta began laying off as many as 8,000 employees this week while canceling plans to hire for another 6,000 open roles as part of its AI efficiency push — and today’s community workflow comes from a mindfulness teacher in Colorado who pointed Claude at his Obsidian vault of years of personal practice notes, giving his students teachings grounded in lived experience rather than generic AI output.

Today we’re covering the biggest AI stories of May 18th, 2026. Conceptual artist SHL0MS posted a real Claude Monet painting on X and told people he generated it with AI — thousands responded calling it emotionless, sloppy, and technically inferior, confidently tearing apart one of the most celebrated impressionist works in history, exposing how the word “AI” alone now triggers reflexive hostility regardless of what’s actually in front of people. OpenAI launched a personal finance experience inside ChatGPT, connecting via Plaid to over 12,000 financial institutions so it can analyze your spending, portfolio, and upcoming bills in real time — with Intuit support and tax estimates coming next. Plus, Anthropic formed a $200 million partnership with the Gates Foundation to deploy Claude in vaccine screening, disease forecasting, and K-12 tutoring across developing nations — and today’s community workflow comes from an anonymous guitarist who used Claude to build a visual practice journal that tracks sessions, progress, and automatically generates three to five new things to work on after every update.

Today we’re covering the biggest AI stories of May 15th, 2026. OpenAI rolled out Codex inside the ChatGPT iOS app, letting developers manage long-running AI coding agents from their phone — approving decisions, starting new tasks, and reviewing code changes while the agent grinds away back at the desk, directly competing with Anthropic’s own Dispatch feature that launched in March. Bloomberg reported that OpenAI has enlisted a law firm to explore legal action against Apple over their 2024 ChatGPT-Siri deal, with internal data showing the integration drove far fewer paid signups than expected and Apple now planning to open Siri to Claude and Gemini in iOS 27 — while simultaneously poaching from Apple’s hardware teams. Plus, Anthropic reversed its April ban on third-party agents but replaced it with a monthly credit pool that gives Pro users just $20 in agentic credits — triggering mass subscription cancellations from power users and developers — and today’s community workflow comes from a project manager in Quebec who built a Gantt chart linked directly to his habit tracker, running his personal fitness goals the same way he runs his work projects.

Today we’re covering the biggest AI stories of May 14th, 2026. Ramp’s latest AI spend index — tracking corporate card payments from over 50,000 U.S. businesses — shows Anthropic surpassing OpenAI in paid business adoption for the first time, with Claude quadrupling its enterprise usage in a year while OpenAI leveled off, validating exactly the “code red” alarm OpenAI’s own leadership sounded internally back in March. Amazon folded its AI shopping chatbot Rufus into a fully redesigned Alexa for Shopping — an agent that follows you across devices, tracks prices, auto-buys when a target is hit, and handles checkouts on non-Amazon stores, all backed by years of purchase history and Alexa conversations. Plus, Nvidia became the first company to hit a $5.5 trillion market cap as Jensen Huang arrived in China alongside President Trump for meetings with Xi Jinping — and today’s community workflow comes from a 61-year-old avid cyclist who couldn’t afford a coach, used ChatGPT to build a full 6-week training plan, and has been reporting back after every ride ever since.