🗣️Yann LeCun: “LLMs aren’t a bubble, the AGI hype is.”

Meta’s chief scientist Yann LeCun pushed back against claims that large language models are an investment bubble. In his view, the money flowing into AI isn’t misplaced but the expectations around AGI are.

🖱 LeCun argues that LLMs already have lasting, practical value and will remain useful across industries for years.
🖱 The real “bubble,” he says, is the belief that scaling current models alone will reach human-level intelligence.
🖱 True progress, in his view, requires scientific breakthroughs not just more data, parameters, or compute power.
🖱 “We’re missing something important,” LeCun warned, suggesting that the current deep learning paradigm needs fundamental innovation.

LeCun’s message lands as a reality check for the AI boom: LLMs may be profitable but not magical.

✔️Powered by Trade Watcher

💰 SoftBank cashes in Nvidia stake to fuel $6B+ AI push

SoftBank has sold its $5.8B stake in Nvidia, using the windfall to supercharge new bets across AI infrastructure, robotics, and data centers part of a sweeping shift from model investments to the physical backbone of AI.

🖱 The profit surge follows soaring valuations in AI portfolios held by Vision Fund and Arm, lifting quarterly results above expectations.
🖱 Proceeds are being redirected into AI data centers with OpenAI, robotics manufacturing in the U.S., and semiconductor ventures aligned with Japan’s tech strategy.
🖱 The sale also trims exposure to overheated GPU markets, signaling SoftBank’s pivot toward long-term ecosystem control instead of short-term chip gains.
🖱 Analysts note this marks SoftBank’s most decisive reallocation since its Alibaba exit, effectively turning Masayoshi Son’s empire into an AI-first holding company.

SoftBank’s Nvidia exit shows that in the next AI wave, the real power may lie not in who trains the biggest model but in who owns the compute, infrastructure, and robotics that make it possible.

✔️Powered by Trade Watcher

⚡️ Yann LeCun to leave Meta and build his own “world model” startup

According to the Financial Times, Meta’s chief AI scientist Yann LeCun is preparing to leave the company and is already in early talks with investors for a new venture.

🖱 The project will reportedly focus on “world models” AI systems that learn from video and spatial data to understand the real world, an idea LeCun has championed for years.
🖱 His exit follows internal tensions at Meta, where LeCun now reports to Alexander Wang, the 27-year-old former Scale AI CEO, a structure some say undercut his autonomy.
🖱 Meta has not commented, and LeCun himself remains silent, but the move would mark the end of an 11-year run leading FAIR and shaping Meta’s AI research vision.

A symbolic moment: the pioneer who helped invent deep learning now leaves the corporate labs to chase a new foundation for machine intelligence, one rooted not in text, but in reality.

✔️Powered by Trade Watcher

💰 Anthropic may turn profitable years before OpenAI

A new WSJ analysis shows Anthropic could reach break-even by 2028, while OpenAI faces deep losses projected at $74B that same year despite record growth.

🖱 Anthropic’s steady climb: Its revenue and costs are rising in sync, keeping spending disciplined and scalability intact.
🖱 Corporate-heavy model: Roughly 80% of Anthropic’s revenue comes from enterprise clients, giving it predictable, recurring income.
🖱 OpenAI’s spending spree: Plans to pour $1.4T into infrastructure over the next eight years, delaying profitability past 2030.
🖱 Strategic contrast: While OpenAI bets on massive consumer reach and platform dominance, Anthropic is quietly building a leaner, cash-efficient B2B engine.

In the race to monetize intelligence, Anthropic might prove that profitability not scale, is the smarter kind of power.

✔️Powered by Trade Watcher

📊 Top 10 most active VC investors in post-seed AI startups

PitchBook’s latest AI & ML report ranked the most active investors backing startups beyond the seed stage (2022–June 2024). While big names dominate, a few lesser-known firms stand out for their quiet consistency.

🧾 Top 10 by number of AI/ML deals:

🖱 Gaingels: 199 deals
🖱 Alumni Ventures: 122
🖱 Sequoia Capital: 113
🖱 Andreessen Horowitz: 94
🖱 Soma Capital: 91
🖱 Tiger Global: 83
🖱 Mana Ventures: 74
🖱 FJ Labs: 72
🖱 Khosla Ventures: 71
🖱 Bossa Invest: 69

🌍 Hidden gems in the ranking:

🖱 Mana Ventures (California): Early investor in Unity, with a portfolio spanning SpaceX, Groq, Canva, Figure, and Epic Games. The firm leans toward deep tech and frontier innovation.

🖱 Bossa Invest (Brazil): A micro VC powerhouse with 1,500+ LATAM investments and over 100 exits. Their Airtable-based portfolio shows at least three unicorns: Enjoy, Swvl, and Amprius.

While the giants still write the biggest checks, regional and specialized funds like Mana and Bossa are quietly becoming power brokers in the AI ecosystem.

✔️Powered by Trade Watcher

💼 The 70/20/10 rule, how McKinsey’s innovation model shapes startup portfolios

The 70/20/10 rule, or McKinsey’s innovation portfolio, offers a clear framework for distributing capital and risk across startups at different maturity levels, balancing stability, growth, and moonshots.

🖱 70% — Core: Proven startups with clear business models and profit. Focused on stable income and risk minimization.
🖱 20% — Adjacent: Startups expanding current business directions. Moderate risk, solid growth potential, and diversification.
🖱 10% — Transformational: Radical, high-risk innovations, AI, biotech, quantum tech, with a chance for x50 returns.

Applied to a $1M portfolio, this means: $700K in late-stage, profitable companies (x2–3 returns), $200K in early-stage, adjacent startups (x5–10), and $100K in pre-seed, breakthrough bets (x50+ potential).

The logic is simple: core holdings fund your stability, adjacent bets fuel growth, and transformational ideas create the upside. Venture success isn’t luck, it’s a system where reliable startups sustain bold innovation.

✔️Powered by Trade Watcher

💰 Robert Kiyosaki says he’s $1.2B in debt and calls it his best financial move

The Rich Dad Poor Dad author revealed he currently holds $1.2 billion in debt, but insists he’s not worried at all.

🖱 Kiyosaki explained that the debt is “good debt”, borrowed money used to acquire cash-flowing assets like real estate and businesses, not liabilities.
🖱 His philosophy: the rich use debt to get richer, while the poor and middle class use it to buy things that lose value.
🖱 He emphasizes that the key is who controls the debt, if your assets pay for it, you own the system; if your labor pays for it, the system owns you.

His takeaway: don’t fear debt, learn to make it work for you.

✔️Powered by Trade Watcher

⚡️ Pavel Durov’s French travel ban is officially lifted as probe continues

French investigators have fully lifted the travel restrictions previously imposed on Telegram founder Pavel Durov, even as their inquiry into potential criminal activity facilitated via the platform remains active.

🖱 The ban, which initially confined Durov to France and later allowed limited travel to Dubai, has now been completely removed, restoring his freedom of movement while the investigation proceeds.
🖱 Prosecutors are examining whether Telegram’s design and policies enabled criminal offences, though Durov has not been charged. The platform’s encryption-first philosophy has long drawn scrutiny from European regulators.
🖱 The decision does not signal the end of the case: investigators continue probing Telegram’s moderation practices and possible lapses tied to illicit activity coordinated on the app.

A telling moment: the restrictions fall away, but the regulatory spotlight remains firmly on Telegram, a reminder that privacy-first platforms are becoming central battlegrounds in Europe’s clash between encryption and oversight.

✔️Powered by Trade Watcher

🔔 Cursor raises a massive $2.3B just five months after its previous round

According to TechCrunch, AI coding assistant Cursor has closed a new $2.3 billion round, catapulting its valuation to $29.3B, barely five months after raising $900M at a $10B valuation.

🖱 The round was co-led by Accel and Coatue, with strategic backing from Nvidia and Google, a signal that AI-powered developer tools have become core infrastructure.
🖱 Cursor’s team plans to pour the new capital into “Composer,” its in-house AI model, aiming to reduce reliance on external LLMs and build a fully proprietary coding stack.
🖱 The valuation tripling in under half a year underscores hyperspeed adoption: investors see AI-native coding environments as the next major productivity platform for developers.

Cursor isn’t just raising capital, it’s becoming the flagship bet that AI-first software engineering could rewrite how the world builds software.

✔️Powered by Trade Watcher

❓ Should young founders learn etiquette or just build?

Slow Ventures’ Sam Lessin has launched an unusual initiative in San Francisco: an “Etiquette School” for young founders. What started as a joke turned into a full program and even sparked a mini-drama with Y Combinator’s CEO.

🖱 Lessin says many first-time founders drop out of school early and jump straight into startups, never learning basic etiquette or how to present themselves. When Slow Ventures opened applications, hundreds applied for a free class with only 50 spots.

🖱 The lesson took place at the Four Seasons: three hours on handshakes, investor communication, business attire, how to eat caviar, how to pick wine and why the era of wrinkled T-shirts might be ending.

🖱 But Y Combinator CEO Harry Tang wasn’t amused. He warned founders not to attend, arguing that great companies come from building things people want, not from polished manners or fancy signals. Real wealth, he argued, is built on value not vibes.

In a world where founders need both authenticity and influence, is etiquette a distraction from the grind or a hidden advantage in high-stakes rooms?

✔️Powered by Trade Watcher

🥸 OpenAI forced to hand over 20M user chats in NYT lawsuit

A federal judge has ordered OpenAI to provide 20 million anonymized user conversations to The New York Times as part of the newspaper’s copyright case, rejecting the company’s attempts to limit disclosure.

🖱 NYT argues ChatGPT reproduces protected articles, prompting a sweeping data request to examine how the model handles copyrighted text.
🖱 OpenAI pushed back, saying 99.99% of the chats are irrelevant and proposing a narrow, selective disclosure, the court rejected the plan.
🖱 The judge ruled that anonymization and a protective order are enough to safeguard user privacy, despite OpenAI’s warnings about overreach.
🖱 OpenAI has filed a motion to reconsider, but for now the company must comply and prepare the dataset for imminent handover.

If this stands, it could set a precedent where courts can demand vast AI usage logs, reshaping the boundaries of privacy, transparency, and copyright enforcement in the AI era.

✔️Powered by Trade Watcher

Virgin Galactic’s space dream crashes back to Earth

In 2020, Richard Branson’s Virgin Galactic vowed to kick off a new era of space tourism. Instead, the company stalled and its stock has since collapsed by 99%, leaving early believers with nothing but cosmic disappointment.

✔️Powered by Trade Watcher

🔔 The Barbell Venture Market: Why startups are splitting into two extremes

Parul Singh argues that today’s venture ecosystem is no longer a smooth curve, it’s a barbell, with winners clustering at two ends and almost nothing in the middle.

🖱 On one side: hyper-efficient, small teams building $10–20M ARR businesses with high margins, startups that don’t need much capital yet create huge personal outcomes for founders.

🖱 On the other: massive, high-upside, capital-hungry companies that justify big funds’ moonshot bets, the only place where large checks still make sense.

🖱 The “middle” (mid-burn, mid-scale, not fast enough to be a rocket) is becoming uninvestable, squeezed by both market discipline and shifting VC incentives.

The result: venture is polarizing, and founders must choose a lane, either build lean and profitable or swing big because the market increasingly rewards only the extremes.

✔️Powered by Trade Watcher

📊 Crypto Venture Weekly: November 10–14, 2025

$136M was raised across 12 projects this week, led by Lighter ($68M) and LISA ($12M). Meanwhile, Strive, VCI Global, Cypherpunk, Upexi, and Strategy collectively deployed $412M into digital asset treasuries across BTC, ZEC, and other holdings.

Here’s what the top 8 are building 👇

🖱 Lighter ($68M, Founders Fund, Ribbit, Haun, Robinhood)
Ethereum-based ZK rollup for perpetual futures, valued at $1.5B.

🖱 LISA ($12M, NGC, Signum, LongHash, UOB Ventures)
AI agent using LLMs to autonomously analyze and audit smart contracts.

🖱 Acurast ($11M, CoinList, Tezos Foundation, Web3 Foundation, Peaq)
Decentralized compute network leveraging smartphones as nodes.

🖱 Seismic ($10M, a16z, Polychain, Amber, dao5, LayerZero)
Encrypted blockchain platform rearchitecting open-source infrastructure around secure hardware.

🖱 Self ($9M, Sandeep Nailwal, Greenfield, Hart Lambur)
Private, verifiable identity protocol enabling self-sovereign digital access.

🖱 Curvance ($4M, Primal, F-Prime, GSR, Flowdesk)
DeFi lending platform simplifying institutional yield access.

🖱 Shodai Network ($2.5M, Consensys Mesh, Joseph Lubin)
Founder network and fundraising platform for early-stage Web3 startups.

🖱 Takadao ($1.5M, Draper, Adaverse, Wahed Ventures)
Bankless finance ecosystem operating from Riyadh and Singapore.

Investor focus this week circled around AI-integrated smart contract tools, ZK-powered infrastructure, and stablecoin-driven finance, while steady treasury allocations signaled continued institutional conviction amid a quieter funding climate.

✔️Powered by Trade Watcher

⚠️ South Korea plans a $35B AI-run megadata center, what could possibly go wrong?

South Korea is launching Project Concord, a $35 billion initiative to build a data center that will be fully designed, optimized, and operated by AI from start to finish.

🖱 AI will select the construction site, design engineering systems, plan cooling, and manage power optimization.
🖱 During operations, AI will oversee monitoring, detect failures, and handle troubleshooting with minimal human intervention.
🖱 The planned capacity is ~3 gigawatts, placing it among the largest hyperscale facilities ever attempted.
🖱 Construction is targeted for 2028, with the entire lifecycle design to daily ops delegated to automated decision-making.

If this works, it sets a new bar for autonomous infrastructure but if it fails, we may learn why humans still like to keep a hand on the power switch.

✔️Powered by Trade Watcher

🔔 Amazon vs. Perplexity agents: the first real battle over AI autonomy

Everyone covered Amazon’s demand that Perplexity’s Comet browser stop its agents from making purchases on Amazon. Almost no one covered Perplexity’s sharp public response which raised a far bigger question: what actually changes when a human hands their authority to an AI agent instead of a human assistant?

🖱 If you ask a human assistant to buy something for you online using your accounts, your logins, even your cards, no platform objects. It’s normal, accepted, and even encouraged for high-value customers.

🖱 But when the same authority is transferred to a software agent, suddenly platforms want bans, restrictions, and legal intervention.

🖱 The reason: for decades, software was a tool. With agentic AI, software becomes labor, an employee acting on your behalf, making decisions and completing tasks.

🖱 And platforms hate this shift. An AI agent can’t be manipulated by ads, banners, fake urgency, or cross-sell tricks. It doesn’t click promos. It doesn’t “browse.” It simply executes the user’s intent and nothing more.

If user agents take over online commerce, they break the core business model of today’s platforms: capturing attention, nudging behavior, and monetizing distraction.

Expect a long and messy fight because AI agents work for the user, not the platform, and that threatens an entire advertising-driven internet economy.

✔️Powered by Trade Watcher

🚐 Global EV market: Chinese manufacturers tighten their grip on worldwide sales

Here’s the latest snapshot of the global electric-vehicle market by manufacturer, counting both BEVs and PHEVs and for Chinese brands, including both domestic and global deliveries.

🖱 BYD dominates with more than 3.8M EVs delivered last year, giving it over 22% of the global market, fueled by massive PHEV volume and fast-growing exports.
🖱 Tesla holds the BEV crown, but its overall EV share continues to slide as Chinese brands scale aggressively across both pure electric and hybrid segments.
🖱 GAC Aion, SAIC-GM-Wuling, Geely, and MG now sit firmly in the global top tier, collectively capturing double-digit market share and reshaping the competitive landscape.
🖱 Chinese automakers benefit from extreme manufacturing scale, vertically integrated supply chains, and lower battery costs giving them a pricing edge global rivals can’t match.

The center of gravity in the EV world has shifted decisively: China isn’t just competing in electric mobility, it’s setting the tempo for the entire global market.

✔️Powered by Trade Watcher

🔔 Peter Thiel dumps all his Nvidia shares as “AI bubble” fears intensify

Billionaire investor Peter Thiel has sold his entire Nvidia position, roughly 537,000 shares that made up 40% of his fund shifting capital into Microsoft, Apple, and partially Tesla.

🖱 Thiel has repeatedly warned that the current AI boom lacks real economic fundamentals, comparing Nvidia’s surge to the late-90s dot-com bubble.

🖱 SoftBank is also exiting its Nvidia stake, signaling growing caution among major institutions watching valuations stretch to extremes.

🖱 Michael Burry, famous for predicting the 2008 crisis, is shorting Nvidia with 14% of his portfolio, one of his largest macro bets in years.

When heavyweight investors start stepping back at the same time, it may be the first real sign that the AI trade is entering its “late-cycle psychology” phase.

✔️Powered by Trade Watcher

📌 Bezos returns to the trenches as co-CEO of $6.2B AI startup Project Prometheus

According to The New York Times, Jeff Bezos is taking on his first operational role since leaving Amazon, becoming co-CEO of Project Prometheus, an AI startup that has already secured a massive $6.2B in early funding.

🖱 Bezos will lead the company alongside Vik Bajaj, a physicist and former Google X executive, signaling that Prometheus aims to fuse deep science with frontier-model engineering.

🖱 The startup focuses on “AI for the physical economy,” building systems for aerospace, manufacturing, automotive, and advanced computing, areas tied tightly to Bezos’s long-term industrial and space ambitions.

🖱 Prometheus has quietly hired top talent from OpenAI, DeepMind, and Meta, assembling a 100-person team before even launching a product, an unusual scale that mirrors its outsized funding.

🖱 The company’s backers see industrial AI as the next trillion-dollar frontier, where software isn’t just generating text but redesigning machines, factories, and physical infrastructure.

Bezos isn’t merely backing another AI lab, he’s positioning Prometheus as a deep-tech powerhouse that could reshape how the physical world itself gets engineered.

✔️Powered by Trade Watcher