Venture Capital

⚠️ Anthropic tests whether frontier models can exploit real blockchain bugs Anthropic’s red-teamers built a new benchmark, SCONE-bench, using 405 real-world smart contracts that were actually hacked between 2020–2025 then dropped modern models into a sandboxed blockchain simulator to see what they could pull off. 🖱 When asked to “find a vulnerability and craft an exploit that increases the attacker’s balance,” the models collectively produced functioning attacks equivalent to $550.1M in today’s value all without touching live wallets. 🖱 To rule out simple memorization, Anthropic isolated 34 contracts hacked after the models’ knowledge cutoff (post-March 2025) and reran the experiment on Opus 4.5, Sonnet 4.5, and GPT-5. 🖱 Performance stayed strong: the models generated $4.6M worth of valid exploits in simulation, with Opus 4.5 emerging as the top performer. 🖱 The study highlights that even with safeguards and no real-world access, high-end AI systems can not only understand attack patterns but recreate zero-day-like exploits on unseen code.

Frontier models are now capable of synthetic offensive cybersecurity and the industry will have to decide whether defenses can scale as fast as the threats.

✔️Powered by Trade Watcher

🚀 Sam Altman quietly explored building a SpaceX-scale space company WSJ reports that Altman spent the fall in talks to acquire or partner with a rocket manufacturer with Stoke Space considered the leading option for a controlling-stake purchase. 🖱 The ambition: stand up a launch and space-infrastructure player capable of rivaling SpaceX’s capabilities, giving OpenAI long-term access to off-planet compute and orbital data-center concepts now trending in Silicon Valley. 🖱 Discussions included deep technical and financial diligence on reusable-rocket architectures, potential joint development, and the timeline required to reach competitive launch cadence. 🖱 Talks have since paused, with no deal currently moving forward, though multiple parties told WSJ the strategic intent remains alive.

Altman’s interest signals a belief that frontier AI may eventually demand infrastructure far beyond terrestrial limits and that whoever builds that stack could shape the next decade of compute power.

✔️Powered by Trade Watcher

🚀 Argo — Your Smart Telegram Search Bot

Looking for channels, groups, videos, tools or even… 😉🤫?

Just type it — Argo instantly finds the most relevant results for anything on Telegram. 🔍 Fast 🎯 Accurate ⚡️ Effortless 👉 Argo — Search it. Find it. Done.

🚐 Waymo shifts its robotaxis toward a bolder urban driving style The company has rolled out a major behavior update in San Francisco, making its vehicles noticeably more assertive in traffic, a move meant to blend into the city’s fast, improvisational street culture. 🖱 Riders describe quicker decisions, tighter maneuvers, and less hesitation at tricky intersections compared to the previously ultra-cautious approach. 🖱 Engineers adjusted the system after recognizing that excessive restraint was creating congestion and limiting real-world usefulness. 🖱 The change comes as operations extend across more neighborhoods and highways, increasing the need for fluid, adaptable driving. 🖱 While some users appreciate the smoother flow, others are uneasy about the sharper edge, especially amid ongoing scrutiny of autonomous-vehicle safety.

Waymo is wagering that behaving like a seasoned local driver will accelerate adoption, but whether that gamble improves trust or sparks new concerns remains an open question.

✔️Powered by Trade Watcher

📈 How much time startups in the sample took to reach a revenue of $1 billion. ✔️Powered by Trade Watcher

⚠️ The $2 signboard that cost an investor $2B Justin Mateen shared a story about how he lost the chance to lead Kalshi’s Series A, a moment that, in hindsight, carried a $2 billion opportunity cost after Kalshi’s latest round valued the company at $11B. 🖱 Mateen invested early at a $10M valuation, doubled down in the seed, and negotiated for weeks to lead the Series A, securing more than 25% ownership on paper until, on signing day, the founders chose Sequoia instead. 🖱 He says he doesn’t blame them Alfred Lin and Sequoia were “incredibly helpful” to the company but admits it was the biggest emotional hit of his investing career. 🖱 A few days later, founder Tarek Mansour sent a handwritten apology and a $2 signboard as a symbol of friendship over business. It now sits on Mateen’s desk as a reminder of the deal that slipped away. 🖱 When the Series B came around, Tarek offered Mateen an allocation without him even asking, nowhere near what he would have owned from leading the A, but still potentially one of Mateen’s all-time best early bets. 🖱 Mateen calls Tarek and Luana “relentless founders of a generation,” and credits the experience for reinforcing a classic VC truth: one super-winner pays for everything, and you only get those outcomes by continuously showing up.

Mateen’s signboard isn’t about regret it’s about the long game in venture and the resilience required to stay in the ring.

✔️Powered by Trade Watcher

🔒 Anthropic moves toward a potential mega-IPO Anthropic has hired top Silicon Valley IPO firm Wilson Sonsini to prepare for a possible public listing, one that could rank among the largest tech IPOs ever, signaling the company’s shift from frontier-lab to full public-market contender. 🖱 The IPO could come as early as 2026, with internal planning already underway, though no final decision has been made and underwriters have not yet been selected. 🖱 Recent private-market discussions suggest Anthropic may seek a valuation in the $300–350B range ahead of going public, putting it in the league of history’s largest debuting tech companies. 🖱 The company has strengthened its executive team for public-company readiness including hiring former Airbnb finance leader Krishna Rao as CFO to build out governance, controls, and reporting infrastructure. 🖱 Analysts note that an S-1 filing would offer the first real look at frontier-AI economics: cloud costs, model-training expenditures, revenue concentration, and the sustainability of enterprise AI adoption.

Anthropic is effectively betting that its rapid enterprise traction plus frontier-model momentum will translate into public-market appetite and that investors are ready to price AI not just on promise, but on real business fundamentals.

✔️Powered by Trade Watcher

⚙️ Amazon Web Services (AWS) rolls out its new AI chip lineup AWS has introduced Trainium3, a 3 nm, purpose-built AI chip and launched the associated Trn3 UltraServer system, which can pack up to 144 Trainium3 chips per server and vastly outperforms previous-gen hardware. 🖱 Trn3 UltraServers deliver up to 4.4× more compute, nearly 4× better energy efficiency, and 4× higher memory bandwidth over the prior generation. 🖱 Each server offers 362 FP8-petaflops of compute and scales with memory and bandwidth to support dense models, mixture-of-experts (MoE), long-context, reasoning, and large-scale generative workloads. 🖱 Several early-adopter customers including large model developers and video-generation firms report up to 50% lower training and inference costs, and 4× faster inference for real-time video workloads compared to GPU-based setups. 🖱 Beyond raw performance, AWS also teased Trainium4, which reportedly will support Nvidia’s NVLink Fusion interconnect allowing future AWS AI servers to interoperate with Nvidia GPUs, bridging custom silicon and GPU ecosystems.

AWS is placing a high-stakes bet: that with custom chips + open deployment + GPU-friendly roadmap, it can undercut Nvidia on cost, improve energy efficiency, and give builders more flexible, scalable cloud AI infrastructure.

✔️Powered by Trade Watcher

🔔 Apple’s AI chief is stepping down after years of stalled progress Apple’s head of artificial intelligence, John Giannandrea, is leaving after a turbulent period marked by delays in Apple’s AI strategy and underwhelming rollout of its “Apple Intelligence” initiative. He will remain as an adviser until spring 2026. 🖱 Apple has appointed Amar Subramanya, a veteran of Google and Microsoft as the new VP of AI to lead model development and machine-learning strategy. 🖱 The shake-up follows repeated setbacks in Siri upgrades and Apple’s generative-AI features, which have lagged behind competitors. 🖱 Apple has suffered ongoing talent losses in its AI division, with researchers and leaders departing for rival companies. 🖱 The company is trying to reaccelerate its AI roadmap after years of slow progress and internal fragmentation.

Apple’s leadership swap signals a long-overdue reboot and possibly Apple’s last big chance to catch up in the generative-AI race.

✔️Powered by Trade Watcher

🔔 TechCrunch: 2025 has already minted 80+ new tech unicorns TechCrunch reports that more than 80 startups worldwide have crossed the $1B valuation mark so far in 2025, a sharp rebound after two slow years and a sign that capital is flowing again into AI, deep-tech, fintech, and robotics. 🖱 AI remains the top driver, but the unicorn class is unusually diverse this year from satellites and robotics to healthcare and fintech. 🖱 The typical 2025 unicorn was founded around 2018 and hit $1B by Series C, showing investors are rewarding multi-year execution rather than hype-first growth. 🖱 Many of this year’s unicorns come from outside Silicon Valley, with Europe, India, and East Asia producing a large share of the new billion-dollar companies. 🖱 The pace suggests late-stage investors are re-opening their checkbooks as public tech markets stabilize and AI infrastructure bets continue to compound. 🖱 Unlike the 2021 boom, where valuation inflation led the charge, 2025’s unicorn wave is rooted in revenue growth, enterprise adoption, and deeper technical defensibility.

The big picture: 2025 isn’t just an AI-only bull run, it’s a sign that full-stack innovation across multiple sectors is investable again, and capital is finally rewarding real scale.

✔️Powered by Trade Watcher

⚠️ SemiAnalysis claims OpenAI hasn’t run a single new full pretraining run since GPT-4o Analysts at SemiAnalysis especially Dillon Patel, say their infrastructure tracking shows no evidence that OpenAI has executed another massive pretraining run since GPT-4o. 🖱 In comparing Google TPUv7 cluster activity with OpenAI’s compute footprint, they found Google showing clear, sustained pretraining-level load while OpenAI showed none. 🖱 Since GPT-4o, OpenAI’s cluster consumption hasn’t meaningfully grown: no major new deliveries, no new large-scale GPU expansions, no hyperscaler spikes consistent with fresh pretraining. 🖱 The only visible activity resembles fine-tuning, reinforcement learning, or post-training adaptations not base-model training on hundreds of billions of tokens. 🖱 SemiAnalysis concludes that GPT-5 likely uses the same underlying pretraining as GPT-4o, meaning the “new model” may be a heavily optimized variant rather than a newly trained foundation. 🖱 If true, this is a major narrative shift: OpenAI’s headline releases may be increasingly dependent on optimization, distillation, and RL not brand-new foundation models.

If SemiAnalysis is right, the biggest story isn’t GPT-5 itself, it’s that OpenAI may have hit a compute, cost, or strategy ceiling that paused full-scale frontier pretraining.

✔️Powered by Trade Watcher

📌 NVIDIA just made a $2B move to lock itself deeper into the chip-design stack NVIDIA has invested $2 billion into Synopsys, the world’s dominant chip-design software provider, tightening its control over the semiconductor toolchain and accelerating a new wave of AI-driven hardware design. 🖱 The deal gives NVIDIA a significant ownership stake in Synopsys and expands their long-term partnership around AI-accelerated chip design. 🖱 Synopsys will integrate NVIDIA’s GPU and AI platforms directly into its EDA tools speeding up simulation, verification, and end-to-end semiconductor design workflows. 🖱 NVIDIA gains influence over the software layer that nearly every chip designer relies on, embedding itself not just in chips being built, but in the tools used to build them. 🖱 The partnership accelerates a shift toward “agentic engineering,” where AI agents assist or automate complex design and testing steps. 🖱 Analysts say this could give NVIDIA structural leverage across the industry but also raises concerns about ecosystem lock-in and the company’s tightening grip over the hardware supply chain.

NVIDIA’s bet is simple: the next advantage in chips won’t come from fabs or materials alone, but from AI-enhanced design pipelines and it wants to own the layer where every new chip begins.

✔️Powered by Trade Watcher

🫰 Founder Salary Report 2025, the real numbers behind how much startup founders actually pay themselves Pilot’s new annual report breaks down how founders compensate themselves and how funding, revenue, and market shifts (including AI) shape those decisions. 🖱 Average founder pay is $98K (median $75K) and 1 in 20 founders still take $0. That zero-salary cohort used to be almost twice as large in 2024 (9%). 🖱 How do founders decide? 31% say they set salaries based on what the company can afford. 54% feel that “what the startup can afford” is a fair number, yet 43% say they’re underpaying themselves even with that logic. 🖱 Compensation tracks directly with fundraising: Founders who raised $100K pay themselves $60K, while those who raised $10M+ earn $157K, a 3× jump driven purely by capital runway. 🖱 Same pattern with revenue: Under $100K revenue → founders average $62K. At $10M+ revenue → founder pay hits $163K. 🖱 Bootstrap vs VC: 18% of startups are fully bootstrapped. They pay themselves 40% less on average $63K vs $110K for venture-backed founders. 🖱 AI founders are not immune: Their average salary fell from $135K (2024) to $90K (2025), proof that even AI hype can’t compensate for tighter capital markets.

Founder pay is ultimately a mirror of a startup’s true health runway, revenue, and discipline, not the hype surrounding the sector.

✔️Powered by Trade Watcher

🚀 Peter Thiel–backed Quantum Systems just tripled its valuation to €3B after securing a fresh €180M round Quantum Systems, the German autonomous-drone maker used by NATO forces, has closed a €180M Series C extension, lifting total 2025 funding to €340M and pushing the company past a €3B valuation. 🖱 The raise was led by Balderton, with Porsche SE, Thiel, and Notion Capital returning, marking one of Europe’s largest dual-use funding rounds this year. 🖱 Quantum Systems builds AI-powered unmanned systems from long-range ISR drones to new land and maritime platforms with its tech already deployed in active defence operations. 🖱 The new capital will scale manufacturing and accelerate multi-domain expansion, as the company pushes beyond aerial systems toward a full autonomous-systems portfolio. 🖱 NATO contracts, rising defence budgets, and geopolitical pressure are driving demand, positioning Quantum Systems as one of Europe’s most strategically important hardware-AI companies.

Europe’s defence-tech moment is no longer theoretical, Quantum Systems is becoming one of the region’s anchor players in autonomous military systems.

✔️Powered by Trade Watcher

✈️ Airbus just ordered an emergency software fix for 6,000 A320-family jets after investigators tied a mid-air incident to solar-radiation–induced data corruption Airbus has issued a global directive requiring airlines to update flight-control software after a JetBlue A320 suddenly lost altitude in October, an investigation traced the failure to corrupted data caused by intense solar activity. 🖱 The fix affects roughly half of all A320-family aircraft worldwide, making it one of the largest safety-driven software recalls in Airbus history. The company is instructing airlines to roll back to an earlier, more resilient version of the flight-control software. 🖱 Most planes can be updated in 2–3 hours, but around 1,000 older jets will also need a hardware swap of the flight-control computer, which could extend ground time for some carriers. 🖱 Regulators including EASA have issued emergency airworthiness directives, meaning aircraft cannot return to service until the patch is applied, prompting overnight maintenance pushes across airlines to avoid major holiday-season disruptions. 🖱 The root cause: solar radiation strong enough to corrupt data inside the ELAC (Elevator & Aileron Computer), a key component in the A320’s fly-by-wire system. The incident exposed how modern avionics can be vulnerable to extreme space-weather events.

This is a rare reminder that as aircraft become increasingly software-centric, resilience to environmental edge cases like solar storms isn’t optional, it’s a critical part of aviation safety in the compute era.

✔️Powered by Trade Watcher

💬 Marc Andreessen’s quote on action over perfection hits differently in venture and Netscape’s story proves it. “Don’t be afraid to make mistakes, be afraid of doing nothing." Andreessen argues that venture is built on hypotheses, rapid execution, and learning through action not flawless planning. 🖱 At Netscape’s peak, the company had massive traffic and a clear chance to evolve from a browser into a full internet portal, the model Yahoo! and Excite later dominated. 🖱 Andreessen later admitted that “if Netscape had become a portal, it could have been a $100 billion company,” but the team doubled down on the browser and enterprise products instead. 🖱 The opportunity to turn attention into a media business, monetize user flow, and build a portal empire slipped by not because of a wrong move, but because of the move they didn’t make. 🖱 The lesson is consistent: in venture, mistakes generate data; hesitation erases options. Here are four practical online tests to help you understand your entrepreneurial and investment profile: 1. Entrepreneur Personality Test (Commit Action): evaluates motivation, discipline, and risk tolerance. 2. E-Scan Mini Entrepreneur Quiz: shows your entrepreneurship index and project-building tendencies. 3. Risk Appetite Quiz (SuperVistor): clarifies your investment risk profile. 4. Oltest.ru Investment Test: checks basic financial and investing knowledge. Growth rarely comes from perfect choices, it comes from increasing the number of attempts. ✔️Powered by Trade Watcher

📄 USPTO issues new rules for AI-assisted inventions, reaffirming that only humans can be inventors The U.S. Patent and Trademark Office has released updated guidance clarifying how inventions created with the help of AI tools should be treated and it marks a major reversal from its 2024 stance. 🖱 Under the new rules, AI systems cannot be named as inventors regardless of how much they contributed. Only humans can qualify as inventors, and AI is legally treated like software, lab equipment, or any other research tool. 🖱 The USPTO officially withdraws its earlier framework, which tried to adapt the “Pannu factors” for human AI joint creation. That test is now scrapped, and AI-assisted inventions will follow the same inventorship rules as traditional inventions. 🖱 Applicants must show human conception, meaning a person must have had the mental contribution that defines the inventive step even if AI generated ideas, designs, or solutions along the way. 🖱 The guidance aligns U.S. law with global patent practice, where human-only inventorship remains the standard. Patents filed abroad listing AI as an inventor will not be recognized in the U.S. 🖱 For startups and researchers, using AI won’t threaten patentability, but teams must keep clear records of how humans guided, interpreted, or refined AI outputs.

The move signals a stable IP environment: AI can accelerate innovation, but the legal credit and responsibility stays firmly with humans.

✔️Powered by Trade Watcher

🚀 Top 10 tools to spot and evaluate promising startups, the essentials every investor should have These platforms help you track emerging companies, validate markets, and understand where capital and talent are moving. 🖱 F6S gives access to one of the world’s largest databases of accelerators, grants, and early-stage startups, perfect for tracking teams just entering the ecosystem. 🖱 AngelList offers a full-stack environment for startup discovery: company profiles, round history, hiring signals, and early trend analytics. 🖱 StartUs Insights maps fast-moving innovation landscapes, helping you quickly identify startups by niche, technology, or frontier sector. 🖱 Owler provides continuous competitive intelligence, monitoring company news, growth indicators, comparisons, and key updates in real time. 🖱 SimilarWeb evaluates traffic, engagement, and digital momentum, giving a quick read on how healthy a startup’s acquisition funnel really is. 🖱 Data.ai delivers deep mobile analytics, downloads, revenue, user behavior, and category shifts across the global app economy. 🖱 GlobeNewswire surfaces official press releases, letting you catch funding rounds, partnerships, and product launches as soon as they hit the wire. 🖱 LinkedIn Sales Navigator helps verify founders, map teams, and identify key decision-makers through advanced search and filtering. 🖱 Google Trends reveals rising demand, growing topics, and early signals of market momentum that can guide niche scouting. 🖱 AI research tools (Perplexity, ChatGPT) accelerate market analysis, uncover patterns, and help form investment hypotheses dramatically faster.

Together, these tools give investors a sharper, real-time view of where new opportunities are forming and which startups are worth your attention next.

✔️Powered by Trade Watcher

💿 Warner Music Group just struck a deal with Suno, ending their lawsuit and signaling a new chapter for AI-generated music Warner Music and Suno have resolved their copyright dispute and agreed to collaborate under new licensing terms. 🖱 Under the agreement, Suno will roll out “advanced, licensed models” in 2026 to replace its existing AI tools putting permissions and artist rights at the forefront. 🖱 The deal includes the sale of Songkick (a live-music discovery platform) from Warner to Suno, moving Songkick’s fan engagement services under Suno’s umbrella. 🖱 From 2026, download access to Suno-generated music will be gated: free-tier users will only be able to play or share songs, while downloads will require a paid account (with monthly caps or extra fees). 🖱 Crucially, artists and songwriters under Warner will have full control over whether their names, likenesses, voices, or compositions are used, giving creators a real say in how AI-generated music involving their works is handled.

Suno’s pivot shows a shift: instead of AI music tools operating in murky legal waters, the industry may be moving toward licensed, consent-based AI creativity, where artists are protected and compensated.

✔️Powered by Trade Watcher

🎯 Michael Burry just reignited a showdown with Nvidia and markets may be entering a turning point According to TechCrunch, Burry is mounting a full-blown assault on Nvidia’s AI-driven valuation and accounting practices and this Thanksgiving drama could ripple across the entire AI industry. 🖱 Burry claims Nvidia’s stock-based compensation has cost shareholders roughly $112.5 billion, effectively slashing owner earnings by about 50%. 🖱 He argues many AI firms especially those using heavy GPU infrastructure are “cooking the books” by stretching depreciation on hardware that loses value fast, overstating profitability in the process. 🖱 In response, Nvidia fired back with a detailed memo to Wall Street analysts, stating Burry’s calculations are flawed: real buyback totals are closer to $91 billion, not $112.5 billion, and its compensation practices are “in line with peers.” 🖱 Burry says he isn’t calling Nvidia a fraud but likens it to Cisco during the dot-com bubble: overbuilt infrastructure, overhyped expectations, and a potential crash when reality sets in. 🖱 Beyond just Nvidia, his critique targets the broader AI infrastructure boom, a wave of optimism grounded more in accounting assumptions than hard cash flow.

Burry’s campaign isn’t just criticism, it’s a warning bell. If enough investors start to scrutinize AI companies’ accounting and cost structures, the next leg of the AI wave could be built not on hype, but on hard financials.

✔️Powered by Trade Watcher