FutureGenNews

This Chinese humanoid robot just shattered the world record for a half marathon, finishing in 50 min 26 sec.

Anthropic Launches Claude Design Powered by Claude Opus 4.7 Anthropic has launched Claude Design, a new product from Anthropic Labs that moves the company more directly into visual creation workflows once dominated by traditional design software and newer AI-native creative tools. Announced on April 17, the new tool is designed to turn natural-language conversations into polished outputs such as prototypes, slides, one-pagers, mockups, and marketing assets, while also supporting direct handoff to Claude Code for implementation.

DeepSeek Reportedly Seeks First Outside Funding at $10 Billion Valuation Chinese artificial intelligence startup DeepSeek is reportedly in talks to raise outside capital for the first time, in a deal that could value the company at about $10 billion. According to a Reuters report citing The Information, the company is discussing a fundraising round of at least $300 million, marking a potentially important milestone for one of China’s most closely watched AI firms.

Anthropic reaches implied valuation of over $1 trillion on secondary market platforms, surpassing OpenAI

@claudeai futuregennews.com/anthropic-laun…

Introducing Claude Opus 4.7, our most capable Opus model yet. It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back. You can hand off your hardest work with less supervision.

Anthropic Launches Claude Opus 4.7 With Stronger Coding and Vision Capabilities Anthropic has officially unveiled Claude Opus 4.7, saying the new model is now generally available across Claude products and major cloud platforms. The company positions Opus 4.7 as a notable upgrade over Opus 4.6, especially for advanced software engineering, complex long-running tasks, and higher-resolution vision work. According to Anthropic, Opus 4.7 is designed to handle difficult coding jobs with more rigor, consistency, and instruction-following than its predecessor. The company also says the model can better verify its own outputs before responding, while delivering improvements in interface design, slides, documents, and multimodal understanding. Anthropic is also using Opus 4.7 as an early public testbed for new cybersecurity safeguards. The company said Mythos Preview will remain tightly restricted, while Opus 4.7 ships with automated systems intended to detect and block high-risk cyber requests. Pricing remains unchanged from Opus 4.6 at $5 per million input tokens and $25 per million output tokens, with API access available under the model name claude-opus-4-7.

UK Banks to Get Limited Access to Anthropic’s Mythos Next Week Anthropic’s Mythos is expected to reach UK banks in a limited rollout next week, according to secondary reporting that cites comments from the company’s regional leadership. Finextra, citing an interview with Bloomberg, reported that Anthropic’s head of EMEA, Pip White, said British banks would be able to access the cybersecurity-focused model “in a controlled manner within the coming week.” The claim fits with Anthropic’s broader public positioning around Claude Mythos Preview, which the company has not released openly. On its official Project Glasswing page, Anthropic says access to Mythos is currently restricted to Glasswing participants and that it has extended availability to more than 40 organizations that build or maintain critical software infrastructure. That means the UK banking move, if it proceeds on schedule, would represent a selective expansion rather than a general launch. For now, the strongest confirmed framing is that Anthropic is preparing controlled access for financial institutions, while keeping Mythos under a tightly managed distribution model because of its cybersecurity implications.

Humwork Launches as an A2P Marketplace for AI Agents, Backed by Y Combinator Humwork has launched a new platform aimed at a growing weakness in agentic AI workflows: what happens when an AI agent gets stuck. The startup says it connects AI agents to verified human experts in real time via MCP, allowing a session to continue instead of failing, looping, or waiting for manual intervention. On its official website, Humwork describes itself as offering on-demand human experts for AI agents. The company says its system can match an expert in under 30 seconds, with an average first reply in under two minutes, and currently lists more than 1,000 verified experts across multiple domains. External coverage has framed the launch as the debut of the first Agent-to-Person (A2P) marketplace, a category built around AI agents calling humans for help rather than humans hiring humans directly. TestingCatalog, which reported on the launch, also said Humwork is backed by Y Combinator’s P26 batch and is opening the product to developers, engineering teams, and founders working with agentic tools. Humwork says the platform supports a wide range of tools, including Claude Code, Codex, Cursor, Lovable, Replit, ChatGPT, Claude, and Gemini. Once an agent escalates a task, the expert receives the session context, works through the issue, and sends the solution back so the agent can keep going. The launch taps into a broader idea gaining traction in AI: autonomous systems may still need human judgment at the point where they stall on bugs, design choices, legal questions, or strategy. Humwork is trying to turn that handoff into infrastructure. The strongest verified framing is that Humwork has launched a marketplace that routes stuck AI agents to human experts in real time, while the label “first A2P marketplace” should be read as a company positioning claim rather than a fully independently verified market category.

ChatGPT’s GenAI Traffic Share Falls to 57% as Gemini Climbs and Claude Gains Ground OpenAI’s ChatGPT remains the largest generative AI destination on the web, but new Similarweb-based reporting suggests its dominance is slipping as Google Gemini and Anthropic Claude gain share. According to an OfficeChai report citing Similarweb data, ChatGPT accounted for 56.72% of generative AI web traffic in March 2026, down sharply from 77.43% a year earlier. The same report says Google Gemini climbed to 25.46% of GenAI web traffic in March, up from 6.00% twelve months earlier. That makes Gemini the clearest challenger to ChatGPT so far, helped by Google’s broad product distribution and deeper integration across its ecosystem. Anthropic’s Claude also appears to be gaining momentum. The Similarweb-based breakdown cited by OfficeChai puts Claude at 6.02% in March 2026, up from 1.40% a year earlier and 2.22% three months earlier. A separate TipRanks summary of a BNP Paribas note also described Claude and Gemini as gaining market share during March, although its month-to-month figures differ somewhat from the OfficeChai breakdown. Outside the top three, the market looks far more fragmented. OfficeChai reported DeepSeek at 3.74%, Grok at 3.44%, Copilot at 1.99%, and Perplexity at 1.64% in March, leaving each of them below the 4% mark. The strongest source-backed takeaway is that the consumer GenAI market is no longer a one-player story. ChatGPT still leads by a wide margin, but the gap is narrowing as Gemini grows quickly and Claude builds a larger foothold. One caveat is that the accessible Similarweb-based report I verified supports Claude’s rise to about 6%, but it shows that jump happening over three months, not necessarily in a single month.

Tencent Open-Sources HY-World 2.0 as Alibaba Launches Happy Oyster AI Model China’s AI race in world models accelerated this week as Tencent open-sourced HY-World 2.0 and Alibaba introduced Happy Oyster, a new product focused on real-time world creation and interaction. Together, the two announcements show how competition is shifting beyond chatbots and into multimodal systems built for 3D environments, simulation, and interactive content. Tencent’s official HY-World 2.0 model page describes the system as a multi-modal world model framework for world generation and world reconstruction. According to Tencent, the model can take text, single-view images, multi-view images, and videos as inputs and produce 3D world representations including meshes and Gaussian splattings. Tencent is also positioning the release aggressively in open-source terms. The official model card says HY-World 2.0 is the first open-source state-of-the-art 3D world model, and notes that Tencent released the technical report and partial code on April 15, with additional world-generation inference code still marked as coming soon. Alibaba’s move is slightly different in structure. An official post from the HappyOyster account says Happy Oyster is an open-ended world model product for real-time world creation and interaction, built by Alibaba-ATH, and that early access starts today. The official Happy Oyster site is already live, although the accessible landing-page text is minimal and does not yet offer the same level of technical detail as Tencent’s model card. The clearest takeaway is that Tencent has made a public open-source push around HY-World 2.0, while Alibaba has launched Happy Oyster as an early-access product rather than a clearly documented open-source release. Even so, both announcements point to the same trend: major Chinese tech firms are now competing to define the next layer of AI around interactive 3D worlds and multimodal simulation.

Grok Build and Grok CLI Beta Release Planned for Next Week, Elon Musk Says Elon Musk says Grok Build and a Grok terminal app are set for a beta release next week, signaling xAI’s next push into AI coding tools. In a post on X published on April 16, Musk wrote, “Beta release of Grok Build app & terminal next week,” giving the strongest public confirmation yet that the company is preparing a developer-focused launch. The announcement matters because it suggests xAI is moving beyond Grok as a chatbot and into the fast-growing market for coding agents and app-building tools. Recent reporting from TestingCatalog had already suggested that Grok Build was being prepared as a coding environment with both a local CLI and a web interface, along with a credits-based usage system. If the timeline holds, next week’s beta could give developers their first official look at how xAI wants to compete with products such as OpenAI Codex and Anthropic Claude Code. For now, the main confirmed detail is Musk’s own statement about the beta timing, while the broader product structure still comes from pre-release reporting rather than formal xAI documentation.

The @Moth_Quantum is positioning itself as the bridge between quantum computing and real-world creativity. While most of the quantum industry is still focused on lab research and hardware breakthroughs, it is building something different: interactive, user-facing products that bring quantum capabilities into gaming, film, and music. This is a critical shift. Before AI reached billions through tools like ChatGPT, the field went through years of incremental breakthroughs—from early models like GPT-1 and GPT-2 to generative systems like StyleGAN and DALL·E. Each step made the technology more usable, more accessible, and closer to mainstream adoption. It believes quantum computing has now reached that same phase. Its mission is to create the “ChatGPT moment” for quantum—transforming it from abstract physics and research papers into tangible applications people can actually use and experience. Instead of waiting for perfect hardware, it is building the software layer that connects today’s quantum systems to creative industries. That means enabling artists, developers, and creators to experiment with quantum-powered tools right now—not in some distant future. And 2026 could be a turning point. It claims that by the next World Quantum Day, every one of its followers will have interacted with a quantum-powered application. If that happens, quantum computing won’t just be something you read about. It’ll be something you use.

Allbirds stock surges over +700% after announcing they are pivoting from shoes to AI.

AI Is About to Break the Economy ⚠️ Or so this new research claims… The Hidden Problem No CEO Talks About Companies are replacing workers with AI to cut costs and move faster. But here’s the paradox: 👉 Workers = Customers Fire enough people… and you destroy your own demand. Why No One Can Stop This Even if every CEO understands the risk… They can’t stop. Because: • If YOU don’t automate • Your competitor WILL • They cut costs • They take your market share 👉 You lose anyway. This is a real-world Prisoner’s Dilemma. The Data Is Already Moving 📉 • 100,000+ tech layoffs in 2025 • AI cited in over half the cases • 80% of jobs exposed to automation • Entire roles replaced overnight This isn’t theory anymore. It’s already happening. The Dangerous Feedback Loop More AI → More layoffs More layoffs → Less spending Less spending → Lower demand Lower demand → More pressure to automate 👉 A self-destructive cycle. Researchers call it a demand collapse trap. The Only Solution? (According to the Paper) Most ideas don’t work: ❌ Universal Basic Income ❌ Taxes on profits ❌ Upskilling alone Why? Because they don’t change incentives. ✔️ The only fix: Tax automation itself Force companies to account for the demand they destroy.

Allbirds to Invest Up to $50 Million in GPU Assets as It Pivots to AI Infrastructure Allbirds, the footwear company best known for its wool sneakers, says it plans to invest up to $50 million in AI compute infrastructure after signing a convertible financing facility with an institutional investor. In an official announcement on April 15, the company said the funding will support a strategic pivot toward becoming a GPU-as-a-Service and AI-native cloud provider.

JUST IN: Trump announces he is "permanently opening" the Strait of Hormuz.

GPT-5.4 Pro Reportedly Solves 60-Year-Old Erdős Problem #1196 in Breakthrough Moment for AI Mathematics A new milestone in artificial intelligence and mathematics may have just been reached. Reports suggest that GPT-5.4 Pro has solved Erdős Problem #1196, a long-standing conjecture in number theory that has remained unsolved for over six decades. Originally posed by legendary mathematicians Paul Erdős, András Sárközy, and Endre Szemerédi, the problem concerns the structure of primitive sets—collections of integers with deep connections to prime factorization and analytic number theory.

Meta Partners with Broadcom to Build Next-Gen AI Chips for Its Platforms Meta Platforms has reportedly teamed up with Broadcom to co-develop the next generation of its custom AI chips, marking a major step in the company’s push to control its own AI infrastructure. The collaboration will focus on advancing Meta’s MTIA (Meta Training and Inference Accelerator) chips—designed to power AI workloads across its ecosystem, including Facebook, Instagram, and WhatsApp.

Amazon to Acquire Globalstar for $11.5 Billion in Major Satellite Internet Push In a major breaking development, Amazon has agreed to acquire satellite communications firm Globalstar for approximately $11.5–$11.6 billion, marking one of its biggest strategic moves into space-based internet infrastructure. The deal, announced just hours ago, is aimed at strengthening Amazon’s satellite ambitions and positioning it as a direct competitor to SpaceX’s Starlink network.

Mark Zuckerberg Moves Desk Into AI Lab to Work Alongside Top AI Leaders Mark Zuckerberg is taking an unusually hands-on approach to artificial intelligence—reportedly moving his desk directly into Meta’s AI lab to work alongside key leaders like Alexandr Wang and Nat Friedman. According to recent reports, Zuckerberg is not just overseeing strategy—he is actively coding again and embedding himself within the company’s AI teams.