⚙️ Alibaba cuts GPU usage by 82% with Aegaeon

Alibaba engineers have introduced Aegaeon, a new system that allows multiple AI models to share the same GPU resources, reducing total GPU demand by 82% in Alibaba Cloud.

🖱 Traditionally, each model (e.g., Llama-70B → 8× A100) is tied to dedicated GPUs that stay loaded even when idle. Alibaba found that 17.7% of its GPUs were serving only 1.35% of total requests, creating massive inefficiency.
🖱 Aegaeon changes this by pooling GPUs and dynamically switching between models at the token level, alternating between prefill and decode phases during generation.
🖱 Instead of reloading models from scratch, Aegaeon caches critical data in VRAM and loads the rest as needed, adding only 3–5% latency overhead while freeing up hundreds of GPUs.

Alibaba reports cutting GPU needs from 1192 to 213, and the system is already running at scale inside Alibaba Cloud.

The move turns model serving into a dynamic orchestration problem, and may become China’s smartest workaround for the U.S. GPU export ban.

✔️Powered by Trade Watcher

Forbes 30 under 30 in the USA

Three months after receiving investment.

✔️Powered by Trade Watcher

🧬 Anthropic launches Claude for Life Sciences to power biotech research

Anthropic has introduced Claude for Life Sciences, a specialized version of its flagship Claude Sonnet 4.5 model tailored for biotech and pharmaceutical R&D. The model is designed to help researchers read scientific papers, generate hypotheses, and prepare regulatory documents.

🖱 The system integrates directly with tools like Benchling, 10x Genomics, Synapse.org, and PubMed to process real lab data.
🖱 It can summarize studies, compare dosing plans, and generate experiment reports but Anthropic notes it won’t accelerate clinical trials themselves.
🖱 The company says its goal is for “a meaningful percentage of all life-science work in the world to run on Claude.”

Anthropic’s push marks a shift from generic AI assistants to deep vertical integration in scientific workflows — turning Claude from a chatbot into a lab assistant for biotech.

✔️Powered by Trade Watcher

💰 John Rockefeller turned management into a science

John D. Rockefeller (1839–1937), founder of Standard Oil, became the world’s first billionaire and a symbol of America’s industrial age, proving that discipline, systems, and scale could outmatch luck and instinct.

🖱 Born into a modest family, Rockefeller started as an accountant before moving into grain trading and then oil. By 1870, he founded Standard Oil, which by the 1880s controlled 90% of U.S. refining capacity.
🖱 He revolutionized business through vertical integration; owning everything from extraction to distribution and negotiated exclusive deals with railroads while buying out competitors.
🖱 After the 1911 antitrust ruling broke Standard Oil into 34 firms (including Exxon, Chevron, Mobil, and Amoco), Rockefeller’s wealth only grew, making him the richest man in history with an inflation-adjusted fortune of $400–500 billion.

Rockefeller later devoted his life to philanthropy, founding the Rockefeller Foundation, University of Chicago, and Rockefeller University, funding breakthroughs in medicine, education, and science.

His story is a blueprint for industrial capitalism where control, structure, and long-term vision built not just a company, but an entire economic era.

✔️Powered by Trade Watcher

🏠 OpenAI trains models to replace junior bankers’ grunt work

Bloomberg reports that OpenAI is developing a secret project, “Mercury,” aimed at automating the repetitive work done by junior investment bankers, from building financial models to drafting IPO and restructuring materials.

🖱 The company has reportedly hired over 100 former bankers from Goldman Sachs, JPMorgan, and Morgan Stanley to help train its AI systems.
🖱 These experts are paid about $150 an hour to create prompts and test model outputs for accuracy in high-stakes finance scenarios.
🖱 The goal is to teach AI to handle spreadsheet modeling, pitch decks, and data analysis, tasks that normally occupy thousands of analyst hours across Wall Street.

By embedding domain experts into AI training, OpenAI is positioning itself not just as a general-purpose platform, but as an automation layer for the knowledge economy — starting with finance.

✔️Powered by Trade Watcher

🚖 Uber and Nebius bet $375M on robotaxis with Avride

Bloomberg reports that Uber and Nebius will invest up to $375 million in Avride, a startup developing autonomous ride-hailing and delivery robots. The deal builds on their 2024 partnership and aims to launch robotaxi services in Dallas by late 2025 through the Uber app.

🖱 Avride already operates delivery robots in Jersey City, Austin, and Dallas via Uber Eats, the new funding will expand its fleet into full-scale autonomous vehicles.
🖱 The investment includes commercial commitments, suggesting Uber will integrate Avride’s vehicles directly into its mobility network.
🖱 Nebius, an AI infrastructure company, is backing Avride as part of a broader autonomous systems and robotics strategy.

With this deal, Uber is reviving its self-driving ambitions, shifting from in-house R&D to strategic partnerships that bring autonomy back under its platform, but without the capital burden.

✔️Powered by Trade Watcher

🔘 Sam Altman’s Orb wants to prove you’re human in the age of AI bots

Reports show that Tools for Humanity, the company behind Worldcoin, has open-sourced its eye-scanning Orb, a device designed to verify that users are real humans, not bots, as AI-generated activity floods the internet.

🖱 The Orb scans a person’s iris to create a unique digital identity while keeping personal data private through zero-knowledge proofs.
🖱 The hardware, firmware, and software are now open source, inviting public scrutiny and developer contributions.
🖱 Companies like Match Group and EventPop are exploring integrations to fight bot abuse and fake accounts.

By turning biometric identity into an open standard, Altman is betting that proof-of-personhood, not just CAPTCHA upgrades, will become a core layer of the AI-era internet.

✔️Powered by Trade Watcher

💸 Startup round sizes hit new highs in 2025

Carta analyzed 2,200+ venture rounds and found that early-stage startups are once again raising record amounts, even beyond 2021’s boom levels.

🖱 Median Seed and Series A rounds have surpassed their previous peaks, signaling renewed investor appetite for early-stage risk.
🖱 The largest rounds (95–99th percentile) are almost exclusively happening in San Francisco, reinforcing its dominance despite global expansion.
🖱 Bigger checks don’t ensure follow-ons, while the correlation is weak, smaller rounds statistically face higher mortality rates.
🖱 Round size is often driven by fund dynamics, not company need as VCs deploy larger funds, they push founders to take more capital.

Venture capital may be more selective in 2025, but when conviction hits, the check sizes show that the era of “mega-early rounds” is back.

✔️Powered by Trade Watcher

⚠️ Trump pardons Binance founder Changpeng Zhao

Former President Donald Trump has granted a full pardon to Changpeng Zhao (CZ), the founder of Binance, less than a year after Zhao pleaded guilty to violating U.S. anti–money-laundering laws and served a four-month sentence.

🖱 The White House framed the move as a correction to “over-enforcement” of crypto laws by the previous administration, signaling a friendlier stance toward digital assets.
🖱 Binance had paid $4.3 billion in fines in 2023, while Zhao resigned as CEO, now the pardon could open the door for his potential return to crypto leadership.
🖱 Critics argue the decision undermines accountability, given Zhao’s conviction and Binance’s history of compliance failures tied to illicit transactions.
🖱 Supporters see it as a pro-innovation gesture, aligning with Trump’s broader plan to position the U.S. as a global crypto hub and attract blockchain investment.

Trump’s pardon of one of crypto’s most controversial figures cements his administration’s pivot toward regulatory forgiveness as economic strategy, abet that leniency will fuel another bull run.

✔️Powered by Trade Watcher

⚛️ Google achieves verifiable quantum advantage with the Willow chip

Google’s Quantum AI team claims a major milestone, its new Willow processor ran the Quantum Echoes algorithm, outperforming supercomputers by up to 13,000× while producing results that can be independently verified.

🖱 The experiment used ~103 qubits to simulate quantum “echoes”’ reversible evolutions that reveal how information spreads in quantum systems.
🖱 Unlike earlier “quantum supremacy” demos, this result is verifiable, meaning other quantum systems can check it, a step toward real scientific utility.
🖱 The Willow chip achieved record gate fidelity and coherence, enabling stable performance at scale for complex physical simulations.
🖱 Researchers replicated nuclear spin echoes similar to NMR signals in 15–28-atom molecules, hinting at future use in chemistry and materials modeling.

Quantum computing is edging past headline “speed-ups” and into useful, checkable science, marking Google’s strongest signal yet that practical quantum advantage is finally within reach.

✔️Powered by Trade Watcher

🤖 OpenAI lays out a $1T plan to become the next tech superpower

OpenAI has unveiled a 5-year master plan to evolve from an AI lab into a trillion-dollar technology empire, with ambitions spanning chips, hardware, data centers, and self-operating AI organizations.

🖱 The company “commits” to investing $1 trillion, focusing on data centers, GPU infrastructure, and the Stargate super-cluster project built with Microsoft.
🖱 It’s also partnering with Broadcom to design its own chips and with Jony Ive to build an “AI iPhone”, a consumer hardware device for natural AI interaction.
🖱 OpenAI’s new 5-level roadmap defines its evolution: from chatbots → solvers → agents → innovators → autonomous organizations (Level 5).
🖱 Today, OpenAI earns from ChatGPT subscriptions, API access, and enterprise deals, but plans to add new revenue streams like e-commerce commissions, agent transactions, advertising, and social features around Sora and ChatGPT.

The vision is audacious: turn today’s conversational AI into a self-running digital economy, where OpenAI powers the infrastructure, interface, and intelligence behind it all.

✔️Powered by Trade Watcher

🧠 OpenAI buys Sky, bringing AI directly into macOS

OpenAI has acquired Software Applications Inc., the startup behind Sky, an AI-powered interface that lets users control their Mac through natural language.

🖱 Sky runs on top of macOS, seeing what’s on your screen and performing tasks in apps, from writing and organizing to planning and coding.
🖱 The entire Sky team, including former Apple engineers who built Workflow (now Shortcuts), will join OpenAI to deepen native system integration.
🖱 The move follows OpenAI’s rollout of the Atlas browser and its push into hardware, signaling a strategy to embed ChatGPT beyond the web.
🖱 While the deal’s value wasn’t disclosed, it positions OpenAI to compete directly in the OS layer, where AI doesn’t just chat but acts.

With Sky, OpenAI is inching closer to its goal of building a truly hands-free computing experience — where your computer understands context, not clicks.

✔️Powered by Trade Watcher

⚠️ Microsoft clones OpenAI’s Atlas browser features in Edge

Just two days after OpenAI unveiled Atlas, Microsoft rolled out a nearly identical update to its Edge browser, powered by GPT-5.

🖱 The new Edge lets Copilot analyze open tabs, summarize pages, access browsing history, and chat contextually with the user.
🖱 In the U.S., an experimental AI agent mode also performs actions directly in the browser, booking, filling forms, and completing tasks on behalf of users.
🖱 The update is live across all regions where Copilot operates, marking Microsoft’s fastest product response yet to an OpenAI release.
🖱 With Edge now mirroring Atlas, the lines between OpenAI’s and Microsoft’s ecosystems are blurring and competition between partners may be turning into quiet duplication.

The browser wars are back but this time, it’s AI assistants fighting for control of the web.

✔️Powered by Trade Watcher

🛒 Walmart partners with OpenAI to bring shopping directly into ChatGPT

Walmart has teamed up with OpenAI to let users browse, plan, and buy products directly through ChatGPT, transforming the chatbot into a full retail assistant.

🖱 ChatGPT users can now ask for meal plans, back-to-school lists, or budget shopping, and the AI will suggest items, fill the cart, and complete checkout via Walmart.
🖱 The feature, called Instant Checkout, marks the first phase of what Walmart calls “agentic commerce” AI agents that anticipate needs and transact on behalf of users.
🖱 For OpenAI, the deal expands ChatGPT from a conversational platform to a commerce engine, capturing real-world data and retail transactions.
🖱 Analysts say the move could reshape product discovery, shifting power from search and ads to conversational shopping powered by APIs and personalization.

By embedding retail into ChatGPT, OpenAI and Walmart aren’t just changing how people shop, they’re redefining the interface of commerce itself.

✔️Powered by Trade Watcher

🤖 Tesla targets mass production for Optimus humanoid robot by 2026

Tesla plans to unveil its Optimus V3 prototype in Q1 2026 and is building production lines designed to scale up to 1 million humanoid units per year.

🖱 The V3 version is a production-intent prototype meant to resemble the final product, Musk says it should “feel more like a person in a robot suit” than a machine.
🖱 Tesla is installing its first dedicated Optimus manufacturing lines and expects to evolve the design through “rolling changes” during ramp-up rather than waiting for model freezes.
🖱 Musk highlighted the robot’s hands and forearms as the hardest components to perfect, describing dexterity as “more complex than the rest of the robot combined.”
🖱 Long term, Tesla envisions Optimus as a mass-market platform, projecting tens of millions of units across future generations and a new supply chain built from scratch.

Tesla isn’t just building robots, it’s betting its next trillion-dollar business will come from manufacturing artificial workers at automotive scale.

✔️Powered by Trade Watcher

💠 Takeda bets $11B on Chinese cancer drugs

Takeda has signed a massive licensing deal with Innovent Biologics, securing global rights (outside Greater China) to two late-stage oncology drugs and an option for a third, amove that could total $11.4 billion.

🖱 The first asset, IBI363, is a bispecific antibody targeting both PD-1 and a modified IL-2, showing promise in solid tumors unresponsive to standard immunotherapy.
🖱 The second, IBI343, is an antibody-drug conjugate (ADC) aimed at Claudin 18.2, with strong early data in gastrointestinal cancers.
🖱 The third, IBI3001, is another ADC (EGFR + B7H3) under Takeda’s option. Innovent receives $1.2B upfront, $100M in equity at a 20% premium, and milestone payments tied to Takeda’s development and commercial success.
🖱 The partnership highlights how pharma now relies on China’s maturing biotech ecosystem, fast trials, deep talent, and strong government support, to source cutting-edge therapies.
🖱 In H1 2025 alone, US pharma signed over $18B in licensing deals with Chinese biotechs, including Novartis–Argo and Pfizer–3SBio.

China is no longer just the world’s pharma factory, it’s becoming the R&D powerhouse shaping the next generation of global cancer drugs.

✔️Powered by Trade Watcher

🌍 IMF: Global economy enters 2026 on shaky ground

The IMF’s latest outlook paints a cautious picture for 2026, no major surprises, but clear warning signs.

🖱 Growth slowdown: Global expansion is losing momentum amid economic fragmentation, rising trade barriers, and protectionist policies.
🖱 Market risks: Overvalued tech stocks could threaten financial stability if corrections hit overleveraged investors.
🖱 Power balance: The ranking of the world’s largest economies is expected to remain unchanged from 2025, stability at the top, stagnation underneath.

The message is simple: the world isn’t heading for crisis, but for a long, uneven plateau.

✔️Powered by Trade Watcher

📉 Crunchbase: 2025 feels a lot like 2021’s market peak

Crunchbase data shows that the current funding climate mirrors the late-2021 boom, rich valuations, fast rounds, and growing signs of overheating.

🖱 Global venture investment reached $303B in the first three quarters of 2025, not far off 2021’s record pace, with AI and defense tech driving most of the capital.
🖱 Startups are again raising follow-on rounds within months and at higher valuations, echoing the frenzy of the pandemic-era bull run.
🖱 Public market exits remain shaky, several high-profile IPOs underperformed, hinting that private valuations may be detached from fundamentals.
🖱 Despite similar exuberance, investors today are more selective, demanding clear revenue traction and defensibility, especially in crowded AI sectors.

The funding party isn’t over, it’s just happening with tighter guest lists and higher expectations.

✔️Powered by Trade Watcher

🎵 OpenAI prepares to launch AI music generator, taking aim at Suno

OpenAI is developing a new system for generating music through AI, positioning itself to compete directly with Suno and other startups building tools for AI-composed songs.

🖱 The project marks OpenAI’s next big expansion beyond text and image generation, entering a new creative frontier: music and sound.
🖱 By leveraging its scale and existing infrastructure, OpenAI could integrate music generation directly into ChatGPT or its upcoming creator tools.
🖱 Suno, currently leading the AI-music space, may soon face an uphill battle as OpenAI’s entry could redefine user expectations and distribution power.
🖱 The music domain presents heavier IP and copyright risks than text or images, suggesting OpenAI will need new data partnerships or licensing models.

OpenAI isn’t just adding another feature, it’s moving to own the entire creative workflow: text, image, video, and now sound.

✔️Powered by Trade Watcher

Top 100 Most Promising Fintech Startups of 2025 According to CB Insights.

✔️Powered by Trade Watcher

🧠 Sam Altman backs non-invasive Neuralink rival — Merge Labs

OpenAI CEO Sam Altman is co-founding Merge Labs, a brain–computer interface startup aiming to rival Elon Musk’s Neuralink, but without brain surgery or implants.

🖱 Merge Labs is reportedly raising around $250 million at an $850 million valuation, with support expected from OpenAI’s venture arm.
🖱 The startup’s core tech uses ultrasound-based interfaces and gene therapy to let neurons respond to sound waves avoiding invasive brain implants.
🖱 Caltech bioengineer Mikhail Shapiro is joining as a key technical leader, known for pioneering research in acoustic neuro-modulation.
🖱 Altman has stated he “would definitely not sew something to [his] brain,” preferring a “read-only” interface that lets users think and receive AI responses.
🖱 The company aims to build a consumer-friendly BCI, potentially integrating directly with AI systems like ChatGPT.

Altman isn’t trying to cut into the brain, he’s trying to make AI part of how we think, turning mind-machine links into a mainstream, non-surgical reality.

✔️Powered by Trade Watcher