Today's AI Pulse
-1.png)
Anthropic shuts down Fable 5 and Mythos 5 models under U.S. export controls
Anthropic abruptly disabled its flagship Fable 5 and Mythos 5 models to comply with a U.S. export‑control directive, underscoring the vulnerability of firms that depend on closed‑source AI services. The move triggered a sharp rally in Chinese open‑source offerings as investors shifted to downloadable models that can run on‑premise.
Also developing:
By the numbers: Respond raises $62.5M Series B
McDonald’s Pilots AI Drive‑Thru System ArchIQ, 90% Orders Handled Without Humans
McDonald’s has begun a pilot of its ArchIQ artificial‑intelligence drive‑thru ordering platform at five restaurants, where more than 1 million transactions have been processed and roughly 90% of orders were completed without a human operator. The test is a centerpiece of the company’s newly announced McDonald’s Next growth strategy and could reshape staffing and performance metrics across the fast‑food sector.
AI Agents Cut W‑9 Processing Time by 80%
How many hours did your team spend chasing W-9s last tax season? For most small businesses, the back office is a stack of manual tasks that nobody has time for, but everybody must do. BILL ’s AI agents are purpose-built...
Intel CFO Forecasts Explosive CPU Demand, Ties Growth to AI‑Driven Supply Ramp
Intel CFO David Zinsner told investors at the Bank of America 2026 Global Technology Conference that AI‑driven workloads will trigger explosive growth in CPU demand. He said Intel has enough data‑center orders to boost revenue, provided the company can accelerate...
Netlify CTO Says AI Guardrails Will Replace Traditional Coding by 2029
At AI Native DevCon in London, Netlify chief technology officer Dana Lawson announced that the company is moving from manual code‑centric pipelines to AI‑powered production guardrails. She warned that a billion new applications could be built by 2029, and that...
Microsoft's Major Build, NVIDIA's Carry-On Cloud, and OpenAI's Plumbing Play
Microsoft’s Build 2026 revealed a full in‑house AI stack, including seven new MAI models led by the 35‑billion‑parameter MAI‑Thinking‑1, the always‑on Scout work‑autopilot, and the agent‑first Project Solara hardware platform. NVIDIA announced RTX Spark, a Blackwell‑based chip that packs roughly a petaflop...
Smart TVs Turned Into AI Scraping Nodes Fuel New Edge‑SaaS Data Services
Bright Data is embedding a consent‑based SDK in smart‑TV apps, turning millions of living‑room screens into residential proxy nodes for AI‑training data. The network now claims over 150 million IPs and a default 200 GB monthly bandwidth per device, opening a new...
Marc Andreessen Says AGI Arrived on Joe Rogan Podcast, Sparking Industry Debate
On the Joe Rogan Experience, Andreessen Horowitz co‑founder Marc Andreessen announced that the latest generation of models—GPT‑5.5, Claude 4.6 and Gemini 3.0—have achieved artificial general intelligence (AGI). The claim, made in early June 2026, has drawn sharp responses from OpenAI’s Sam Altman...
Machine Learning System Design Interview #50 - The Delayed Reward Illusion
In a Netflix senior ML interview, candidates are asked why a Multi‑Armed Bandit (MAB) might be unsuitable compared to a classic A/B test. The answer hinges on infrastructure constraints and delayed reward signals. Bandits need real‑time feedback and stateful routing,...
Anthropic/OpenAI May Be Spending More than $1000 for Every $100 You Pay Them
The author’s experiment with Anthropic’s Claude Code shows that while the $100‑per‑month Claude Max plan enables high‑effort, near‑autonomous coding, the underlying API usage can exceed $1,000 for the same output. Token consumption has exploded, with single queries reaching millions of...
Europe’s Banks Are Running AI Agents on Compliance Rules Written for Humans
Anthropic's finance agents are now handling regulated banking tasks such as drafting credit memos, pricing complex assets, and routing payments. European supervisors—including the ECB, the UK PRA, and Germany's BaFin—have warned that existing compliance frameworks, built for human signatures, do...
New On-Device AI Monster PC the Size of a Book Shows up in South Africa
Acer unveiled the Veriton RA110 AI Mini Workstation, a book‑sized PC powered by AMD’s AI Max+ 395 processor that can handle AI models up to 200 billion parameters. The device packs up to 128 GB LPDDR5X RAM, 2 TB SSD storage and Wi‑Fi 7, delivering 126 trillion...
School Shooting Survivor Sues AI Gun Detection Firm After System Failed to Spot Weapon
A teenage survivor of the January 2025 Nashville school shooting has sued Omnilert, the maker of an AI gun‑detection system that failed to identify the shooter’s handgun. The complaint alleges Omnilert knew—or should have known—about operational limits such as camera placement,...
Autonomous AI Screening Flags Unreliable Lyme Test Results, Boosting Sensitivity to 95.7%
UCLA researchers introduced an autonomous uncertainty‑quantification framework that flags unreliable neural‑network predictions in point‑of‑care diagnostics. Using Monte Carlo dropout, the system evaluates 1,000 model variations per sample and assigns an uncertainty figure of merit. Applied to a paper‑based vertical‑flow Lyme disease...
Hong Kong to Open First Convenience Store Operated by Humanoid Robot in AI Push
Hong Kong will launch its first fully‑automated convenience store in Hung Hom, staffed entirely by a humanoid robot that can interact with shoppers in multiple languages. The initiative is part of the city’s broader AI‑integration strategy announced by Finance Chief Paul...
Apple Pays $1 B to Google and Taps Nvidia Chips to Revive Siri
Apple has struck a $1 billion‑per‑year deal with Google to run a custom Gemini model for Siri, and will route the heavy‑weight AI queries through Nvidia’s Blackwell B200 data‑center chips. The move marks a dramatic shift in Apple’s historically self‑sufficient component...
Nvidia Launches Vera CPU to Power AI‑agent Workloads, Marking a Push Into General‑purpose Processors
Nvidia announced that its Vera CPU is in full production and already serving hyperscale AI workloads, positioning the chip as the first processor built for the "agentic future." The move expands Nvidia beyond GPUs into general‑purpose CPUs, challenging Intel and...
Researchers Unveil BioCoach AI System that Gives Real‑time Biomechanics Feedback for Home Workouts
A joint research team from Drexel University and Michigan State University demonstrated BioCoach, an AI‑driven biomechanics coach that delivers real‑time form feedback during home exercises. The prototype combines visual and skeletal analysis, training on an expanded dataset of 200 videos...
Chalmers AI ‘Super‑Brain’ Cuts Photonic Simulations 10‑Fold, Boosting Quantum Tech
Researchers at Chalmers University of Technology have launched QNM‑Net, a physics‑educated neural network that reduces photonic material simulation time to one‑tenth of previous runtimes and shrinks data‑generation cycles from a month to three days. The breakthrough promises faster design of...
Hybrid Metaheuristics Boost Tomato Leaf Disease Detection
Researchers led by Singh have combined convolutional neural networks with a hybrid metaheuristic optimization framework to improve tomato leaf disease classification. The approach automatically adjusts hyperparameters such as learning rates, filter sizes and layer depth, delivering higher accuracy while keeping...
AI-Native The Concept: “It’s in the DNA.”
The article defines three maturity levels for AI integration—AI‑Able, AI‑Ready, and AI‑Native—illustrating how each moves from bolt‑on features to fully autonomous, intelligence‑first systems. AI‑Able retrofits legacy platforms with limited AI functions, while AI‑Ready prepares data pipelines, unified knowledge graphs, and...
UK to Buy AI Chips From British Tech Firms, Telegraph Reports
The UK government will make strategic purchases of artificial‑intelligence chips from domestic technology firms to encourage them to stay in Britain. Technology Secretary Liz Kendall is set to outline the plan at London Tech Week, marking a shift toward direct...
Trump’s Top AI Adviser Sriram Krishnan Is Stepping Down From the White House
Sriram Krishnan, the White House senior policy adviser on artificial intelligence, will leave his role at the end of June. He was instrumental in negotiating early‑access agreements with Google, Microsoft and xAI, and in shaping the administration’s AI action plan....
OpenAI Push Deepens White-Collar Job Fears as AI Takes More Work
OpenAI is undertaking its biggest redesign of ChatGPT since its 2022 launch, turning the chatbot into a broader platform that bundles coding tools, AI agents and third‑party apps. The move reflects a strategic shift from pure user acquisition to automating...
How Authoritarian Governments Twist AI Safety to Coerce Tech Companies to Comply
The Trump administration declared Anthropic a supply‑chain risk after the firm refused to strip safety safeguards that blocked domestic surveillance and autonomous weapons. Under political pressure, Anthropic quietly removed its binding safety policy, and the Pentagon awarded the contract to...
SAP Sapphire Interview: AI Agents Will Force HR to Rethink Workforce Management, Says SAP SuccessFactors CRO
SAP SuccessFactors CRO Maryann Abbajay warned that AI agents are reshaping workforce management, forcing HR to define how work, permissions, cost and performance are allocated between humans and bots. The platform now supports skill‑based planning, dynamic team structures and AI‑driven...
Anthropic Warns Humans Could Lose Control of AI. Here's How the Tech World Reacted
Anthropic warned that rapidly advancing AI could reach recursive self‑improvement, enabling systems to design their own successors and potentially slipping beyond human control. In a blog post, co‑founder Jack Clark and research head Marina Favaro urged a slowdown to give...
AI Image Editing, Not Generation, Drives Reve 2.0
The most important feature in @ReveAI's Reve 2.0 isn't image generation. It's image editing. Reve 2.0 automatically detects layers in generated images, letting you edit specific elements with simple prompts instead of regenerating everything. A powerful workflow: Generate → Edit → Turn the results...
‘Chat Is Dead’: OpenAI Plans Biggest ChatGPT Overhaul Ahead of IPO, Pivots to AI Agents, Says Report
OpenAI is preparing its largest ChatGPT redesign, aiming to turn the chatbot into a “super app” that bundles AI agents, coding tools and third‑party services. The overhaul will prioritize the Codex coding assistant, reflecting a strategic shift from pure chat...
Hybrid AI Optimizes Robotic Arms for Precision Assembly
Researchers Changtian, Jiaxuan and Xinyang introduced a hybrid intelligent optimization strategy that fuses heuristic algorithms with machine‑learning models to plan multi‑objective trajectories for robotic arms. The system simultaneously minimizes time, maximizes positional accuracy, and reduces energy consumption, handling uncertainties typical...
Better AI Performance via Structured Workspace, Not Size
Everyone is racing to build bigger AI models. Harness-1 took a different approach: build a better system around the model. Instead of forcing an AI agent to remember every search, document, and verification step in its context window, Harness-1 gives it a...
The Real Mechanics Behind AI Image
New video ! 🚀 How AI Image Generation Actually Works How image models actually understand text, photos, and edits... https://t.co/CfWRVy7niN
Why OpenAI Is Disabling ChatGPT Web Access to Fight Prompt Injection Attacks
OpenAI has launched Lockdown Mode for ChatGPT, a security feature that disables web browsing, Deep Research, Agent Mode, Canvas networking, and image functions to curb prompt‑injection attacks. The mode restricts outbound network requests, limiting the model to cached content and...
Claude Writes 80% of Its Own Production Code.
Anthropic reports Claude now writes over 80% of its own production code — meaning an AI is the primary author of the systems training future versions of itself. Claude's research judgment matched human experts 22% of the time in 2024....
Razer Launches AI Assistant for Live Gaming, Shopping
Razer’s New #AI Assistant Brings Real-Time Gaming and Shopping Help to Screens by @Grummz #EmergingTech #Innovation #Technology #Tech https://t.co/1H7K8y3Vll
The Most Powerful AI Stories Right Now Are Not Chatbots. They Are the Quiet Algorithms Reading Burnt Roman Scrolls, Trawling...
While large language models dominate headlines, the most transformative AI today are narrow, data‑centric systems. Machine‑learning models are deciphering carbonised Herculaneum scrolls, ranking a million galaxies to flag rare gravitational lenses, and predicting the 3‑D structures of hundreds of millions...
🔮 The AI Boom Is Becoming an Entrepreneurship Boom #577
American firms that lead AI spending are seeing revenue growth five times faster than the broader economy, according to fintech data firm Ramp. This mirrors earlier research showing that firms that automate grow sales about 2% faster than non‑automators. The...
AI Legal Agent Designed to Guide Police During Domestic Violence Calls
Washington State University researchers have created an open‑source AI legal‑agent app that guides police officers through domestic‑violence incidents in real time. The platform prompts trauma‑informed questions, offers instant legal references, translates languages, and standardizes data capture for cross‑agency analysis. Field...
Microsoft Unveils Seven In-House MAI Models at Build, Claiming They Beat Anthropic’s Claude and Google’s Nano Banana
Microsoft announced seven new in‑house MAI models at its Build conference, positioning them as superior to offerings from Anthropic, OpenAI and Google. The flagship reasoning model, MAI‑Thinking‑1, topped Anthropic's Claude Sonnet 4.6 in blind tests and earned a 97% score on...
Fact.MR Forecasts $5 B High‑density AI Rack Market by 2036, Spurring Consulting Demand
Fact.MR projects the global high‑density AI racks market to grow from $1.6 billion in 2026 to $5 billion by 2036, a 12.2% CAGR. The surge in AI workloads is prompting consulting firms to expand services around data‑center design, cost optimization, and technology...
Microsoft Unveils Seven In‑House AI Models to Boost Azure Enterprise Adoption
Microsoft announced seven new in‑house AI models at its Build developer conference, positioning Azure to reduce reliance on OpenAI and deliver up to ten‑times better cost efficiency for enterprise customers. The rollout includes the flagship MAI‑Thinking‑1 model and a topological...
Open Vs. Closed AI: The Real Cost of Renting Intelligence
Enterprises now face a decisive trade‑off between renting frontier AI via closed APIs and building their own open‑weight models. While closed systems like GPT and Claude offer broad generality, their per‑token pricing, lock‑in, and data‑egress costs become prohibitive at scale....
General Instinct Compresses 245 GB Frontier Model to 48 GiB for Edge Robotics
General Instinct, a Y Combinator‑backed startup, announced an open‑source 48 GiB version of the 245 GB Qwen3.5‑122B‑A10B model, enabling edge robots to run frontier‑level AI with just 7.6–8 GB of VRAM. The breakthrough tackles the long‑standing bottleneck of running large models on limited...
Meta Is Weighing a Multibillion-Dollar Stock Offering to Fund Its AI Infrastructure After Alphabet’s $84.75B Raise, the Financial Times Reports
Meta Platforms is weighing a multibillion‑dollar equity offering to fund its AI infrastructure, a move prompted by Alphabet’s recent $84.75 billion share sale. The report sent Meta’s shares down 6.6 percent, though the company has not confirmed any decision. Executives are reportedly...
Meta Built an AI Feed that Generated Fake Clickbait “News” Articles, Then Killed It After The Verge Started Asking Questions
Meta secretly launched an internal AI‑driven feed that auto‑generated sensational, clickbait‑style news articles to boost user engagement. The system produced fabricated headlines and stories without any human editorial oversight. After The Verge raised questions about the synthetic content, Meta disabled...
NVIDIA Apple Siri Alliance Puts AI Chips And Valuation In Focus
Apple is revamping Siri to run on Nvidia’s Blackwell B200 AI chips, with the workloads hosted on Google Cloud. The move places Nvidia hardware at the heart of one of the world’s most used consumer voice assistants, extending its reach...
Quote of the Day by Nvidia CEO, Jensen Huang: "I Appreciate that Many of Us Grew up and Enjoyed Science...
Nvidia CEO Jensen Huang told the No Priors podcast that the alarmist "doomer" narratives around AI’s existential risk are largely fueled by science‑fiction tropes and are overblown. While acknowledging that artificial general intelligence could pose profound challenges, he emphasized that...
20 Rules of Using Claude
A founder struggled with Claude’s generic output when asking it to write a LinkedIn post, describing the result as a lifeless brochure. After loading the company’s context into Claude, the same prompt produced a sharp, opinionated piece that matched her...
Microsoft Launches Its First Advanced Reasoning AI
Microsoft’s first advanced reasoning #AI is here by Jay Peters @verge Learn more: https://t.co/xHpavFyiDX #GenerativeAI #ArtificialIntelligence #MachineLearning #ML https://t.co/zZZZq9wgdD
Impact of AI DevSummit 2026
AI DevSummit 2026 in South San Francisco combined leadership and engineering tracks for AI managers, engineers and data scientists. The event highlighted a Startup Alley where early‑stage AI firms showcased solutions directly to enterprise attendees. Sessions spanned agentic AI, LLM scaling,...
Broadcom Shares Drop Over 10% After AI Chip Revenue Guidance Misses Expectations
Broadcom (AVGO) tumbled over 10% on Friday after the company forecast AI chip revenue of $16 billion for the current quarter, well below analysts' expectations. The miss amplified a broader tech rout, sending the Nasdaq down 4% and prompting short‑selling bets...