Michael Marsella

Meta's new magic number is 49%. First it wrote a $14B check for 49% of Scale AI, poaching CEO Alex Wang. Now the same playbook hits venture land: Meta is offering to buy up to 49% of Nat Friedman & Daniel Gross's $1.1B NFDG fund - and hiring the duo to run its shiny new AI lab - letting early LPs cash out at a 4x markup while still calling the fund "independent". Antitrust alarms? Snoozing. All thanks to that cute little missing 1%.

25

1 Comment

Scaling LLMs isn't just a modeling challenge—it's a massive engineering one. 🛠️ In their latest tech blog, Netflix dives into how they’ve built a robust Post-Training Framework to move beyond simple scripts and into production-grade AI. They cover: ✅ The shift from SFT to complex RL workflows (like DeepSeek-R1 style GRPO). ✅ Solving "silent" training-serving skew by staying Hugging Face-centric. ✅ Optimizing throughput by 4.7x using on-the-fly sequence packing. A must-read for anyone building infrastructure to support the next generation of GenAI. Check out the full breakdown here: https://lnkd.in/gayzH_fq #AI #MachineLearning #LLM #NetflixTech #GenerativeAI #Engineering

3

Zynga sold for $12.7 billion. Its founder says startups fail due to this: They try to innovate too early (unless its deep tech). Mark Pincus calls it the “All new fails” rule: 1/ Don’t start with “new”: Start with what’s already proven to work. Copy it legally, study it deeply. 2/ Proven means 10 out of 10 users say: “Yeah, that works.” If they don’t, it’s not proven. 3/ Founders fail because they skip the proven phase: They chase novelty before they’ve earned the right to innovate. 4/ The Zynga story began with poker: Same table, same felt, same gameplay, nothing original. 5/ Then they made it better: No downloads. Just click and play. Friction cut in half. 6/ “Better” means 10 out of 10 users agree: Not your team, not your investors, real users. 7/ Only after proven and better do you earn “new”: That spark that surprises and delights users emotionally. 8/ Zynga’s “new” was showing your friends’ faces: It made games social, not solitary. 9/ Forget MVP, build a minimum viable idea: A concept that hits emotional resonance before metrics. 10/ Seek true signal: You’ll feel it viscerally before you can measure it. 11/ Silicon Valley worships originality: But originality without resonance is noise. 12/ Better founders are great students first: They copy what works before they try to change it. 13/ Most products die from founder ego: They’re “new” before they’re useful. 14/ Users don’t care about novelty: They care about ease, trust, and delight. 15/ “Proven → Better → New just works. And it built a $12.7B company. 📩 Get ai + vc intel in your inbox via my newsletter Startup Riders, link under my name ☝️ 📌 Source: Mark Pincus on Why Most Startups Fail a16z #ai #llms #agent #startups #founder #vc

40

4 Comments

Armis Security has secured $435 million in a pre-IPO funding round, valuing the company at approximately $6.1 billion. The round was led by Goldman Sachs’ Growth Equity arm, with participation from investors including CapitalG and Evolution Equity Partners. Armis specialises in cybersecurity for connected devices across enterprise, industrial and public-sector networks. With the new capital, the company positions itself for a potential IPO in the late-2026 to early-2027 timeframe. Source: TechCrunch Article (https://lnkd.in/ehB7suen) #armis #goldmansachs #secondaries

11

Uber founder Travis Kalanick has a new company called Atoms, which will roll up his current company CloudKitchens, and also tackle mining and transport. The transport piece will be about making a "wheelbase for robots" -- and Kalanick does not seem to be going after humanoids. Atoms has a website that has.... a lot of words about Kalanick's vision of the future, but not much detail on what comes next. If you have thoughts or ideas, reach out. More on TechCrunch: https://lnkd.in/eSzT6wXF

16

1 Comment

OpenAI dips its toes in consulting. Youssef Ben Mahmoud argues thats the biggest opportunity for OpenAI lies in moving up the value chain from selling access to an increasingly commoditized product (LLMs) to selling digital transformation services that support the edge cases: He says: "AI models are becoming commoditized. The value is in making them work inside enterprise chaos. When you realize that selling software isn't enough, you need to sell digital transformation." My take: Meta is driving up the cost of AI engineers in a massive bidding war. What if the next great talent crunch and opportunity were the people that could successfully apply AI to the edge cases. One other thing -- what if the biggest opportunity here for the big AI companies was figuring out how to support these edge cases by improving feedback from their consulting arms? Not sure if they can pull it off with all the hyperfocus on AGI. Companies like Salesforce seem to be in a better position for this particular feedback loop.

1

$1B+ bet on the next frontier of AI: Physical & Spatial Intelligence Today, Yann LeCun’s new startup, AMI - Advanced Machine Intelligence, announced a massive $1.03 billion funding round to challenge the dominance of Large Language Models (LLMs). Current AI is "trapped" in a digital box of text. To reach human-level autonomy, AI needs to understand the physical world. It needs to know that if you push an object, it falls; it needs to navigate a room without a map; it needs to plan complex tasks in real-time. From Digital to Physical: AI is moving into robotics, smart glasses (like Ray-Ban Meta), and industrial automation. From Prediction to Planning: Moving away from probabilistic guessing toward goal-oriented reasoning. The "Ami" (Friend) Approach: Building AI that is controllable, safe, and grounded in reality. The next frontier isn't just "generative"—it’s spatial and physical. We are moving from AI that talks to us, to AI that works alongside us in the physical world. Congrats to the AMI team on this milestone. The era of "World Models" is here. https://lnkd.in/gE99YfGJ #AI #Robotics #MachineLearning #YannLeCun #AMI #Innovation #SpatialIntelligence #TechNews

2

The 𝗣𝗮𝗹𝗮𝗻𝘁𝗶𝗿 "𝗠𝗮𝗳𝗶𝗮" is quietly building again — this time inside 𝗬 𝗖𝗼𝗺𝗯𝗶𝗻𝗮𝘁𝗼𝗿 Former Palantir engineers and operators are launching startups that touch AI infra, developer tools, and next-gen data intelligence. I pulled the data using Harmonic — my go-to startup discovery engine as a BA - and here are 6 new YC-backed startups founded by ex-Palantir talent 👇 ▫️ Clidey (YC X25) – AI-powered infra platform with containerized workflows and real-time analytics ▫️ Overstand Labs (YC W25) – AI insights from Slack, email, WhatsApp to surface product signals and churn risk ▫️ Laminar (YC S24) – LLM tracing + evaluation with blazing-fast Rust pipelines and graph-based workflows ▫️ Zeit AI (YC S24) – Natural language queries across messy enterprise data, with GDPR-safe profiling ▫️ Kenley– AI copilots for consulting workflows like RFPs, due diligence, and internal search ▫️ Parsewise – AI agents that extract structured data from messy documents, with traceability All 6 were discovered through Y Combinator’s latest batch. All have roots at Palantir. And all are solving real data problems at the edge of what's possible. Which one is your favourite? 

329

19 Comments

🚨 BREAKING NEWS: Capital One just announced it’s acquiring Brex in a $5.15 billion cash-and-stock deal — first reported in The Wall Street Journal. Brex — a fintech giant built as the next-gen corporate spend — is being brought under the wing of a very legacy player. Capital One isn't just agreeing to buy a fintech. It is buying its way into $13B+ in Brex-managed deposits, a massive startup client base, and one of the slickest B2B UX stacks in fintech. This comes just months after Capital One’s $35B acquisition of Discover — giving the bank its own payments network to compete with Visa and Mastercard. What’s clear? Capital One isn’t playing catch-up. It’s buying its future — and bringing Brex’s tech and customer base with it. Fintech M&A isn’t slowing down in 2026. If anything, we’re seeing the convergence of speed, scale, and survival strategies across the board. 📌 More thoughts in tonight's Fintech Is Femme newsletter — but currently thinking about: – Why this deal isn’t just about credit cards — it’s about data, deposits, and distribution – What this means for Brex’s legacy as a “startup for startups” – And how the M&A narrative is shifting from survival to strategy Sign up for the Fintech Is Femme newsletter -- and I'll see you in your inboxes tonight! What’s your read on this one? Let me know your thoughts in the comments.

128

13 Comments

Wooow. Palantir, too, moved to Miami. Very exciting news and another important milestone for Miami's AI & Tech evolution. 4A Labs Erden Alpan

15

1 Comment

CHAOS Industries just detonated a signal flare across the defense tech landscape with a $510M Series D that feels less like fundraising and more like the market admitting it finally understands what this team has been building since 2022. Watching CHAOS Industries go from stealth to a $4.5B valuation in under 4 years is the kind of momentum you normally only see when a company stops talking about the future and starts delivering it in #fieldtests, #contract wins, and operational hours that stack up like a pilot's logbook. Valor Equity Partners stepped in to lead the round, 8VC and Accel came back for more, and Antonio Gracias joining the board tells you this is not momentum by accident. It is momentum by design, the kind that gets measured in TRL-9 hours and not PowerPoints. John T. has been guiding this company with the calm precision of someone who already sees around corners. Dr. Bo Marr, PhD turned coherent distributed radar from a theoretical advantage into the backbone of a real #sensingnetwork that improves every time another node hits the field. Gavin Hood brings a decade of intelligence and Palantir experience that lets him read the global landscape like a novel he has already underlined. Brett Cummings built the financial architecture that took CHAOS Industries from fewer than 12 people to 150+ while raising $1B in total capital with the tempo of a company that refuses to wait for the market to catch up. The milestones are not buzz; they are proof. Astria Therapeutics, Inc. advancing at Eglin Air Force Base with a $2M United States Air Force contract. A $10M House appropriation to strengthen #test and #training infrastructure. VANQUISH Group logging 1,000+ hours across targets from Group 1 UAS to #fighte jets with #antijamming resilience that speaks for itself. A Middle East allied partner signing on after seeing the system perform in real operational environments. The Forterra partnership linking VANQUISH with autonomous Smet vehicles gives CHAOS Industries a foothold in the future of ground-based sensing the same way their CDN platform is redefining #airdefense today. What makes all of this hit harder is the pace. A $275M Series C in May 2025 and a $510M Series D just 4 months later would look reckless if the demand were not real, validated, and accelerating. Manufacturing scale-up, ZIVA's wireless time sync woven into every product line, AI-driven #threatdetection, and expansion across NATO, the Middle East, and Indo-Pacific markets paint a picture of a company not chasing the defense tech wave but shaping it node by node. CHAOS Industries is proving that the future of sensing will not be built around a single exquisite system but around distributed intelligence that grows stronger as it spreads. #Startups #StartupFunding #VentureCapital #SeriesD #AI #Defense #DefenseTech #Communication #DeepTech #Infrastructure #Technology #Innovation #TechEcosystem #StartupEcosystem #Hiring #TechHiring

58

5 Comments

Same as it ever was, in reverse: this time around NON-sports is subsidizing sports. 1. Amazon Prime's NFL rights are supported by revenue from Prime shipping and entertainment content 2. YouTube TV's sports offerings are backed by YouTube's massive ad business 3. Apple TV+ uses device margins to fund MLB/MLS rights Key implications: - The true cost of sports rights is being masked by diverse revenue streams - Traditional sports networks can't compete with this model - We're likely to see more tech platforms enter sports rights bidding The numbers are striking: A streaming platform can justify paying 30-40% more for sports rights compared to traditional broadcasters due to these alternative revenue streams. This isn't just about sports - it's about the future of content monetization. The winners will be those who can leverage multiple revenue streams to support premium content acquisition.

36

4 Comments

Rails, Not Roulette Everywhere you look in the collectibles space, companies are trying to build marketplaces. It’s the obvious play: aggregate demand, collect transaction fees, and ride the hype cycles. But marketplaces without rails collapse. What we see today is an industry propped up on roulette mechanics — breaks, repacks, and “chase” formats where $100 buys you a chance at a $1,000 grail card, but most of the time you walk away with a $50 floor. It extracts value from collectors without improving trust, liquidity, or the long-term health of the ecosystem. I understand gambling. It’s part of human behavior, and risk is part of collecting. But gambling cannot be the foundation of a $20B+ hobby. If the infrastructure remains broken, the economics never improve. The real opportunity isn’t roulette. It’s rails. Rails are the hard, unglamorous problems nobody wants to solve: • Fulfillment & Vaulting — reliable intake, scanning, storage, and shipping. • Unit Economics — making $1 cards, $50 cards, and $10K cards all tradeable with sustainable margins. • Shop Infrastructure — storefront SaaS, branded subsites, integrated payouts, and customer ownership tools for local card shops. • Market Intelligence — real pricing signals, player analytics, and automation that improve transparency and trust. These are the rails that will reduce fees, increase liquidity, and create a healthier ecosystem for both shops and collectors. At Pulltrader, we’re taking the contrarian approach: infrastructure first, marketplace second. By owning the rails, we improve the economics of every transaction — whether it’s a $1 Binz card or a $10K grail. The companies that last in this hobby won’t be the ones spinning the roulette wheel. They’ll be the ones laying down the tracks that everything else runs on. Rails, not roulette. #StartupGrowth #ScaleUp #MarketExpansion #UnitEconomics #GrowthStrategy #FutureOfCommerce #ProductMarketFit #SaaSgrowth

2

🚨 BREAKING: Apptronik raises $935M Series A at $5B+ valuation! University of Texas spinout Apptronik just announced it re-opened its Series A to raise a total of $935 million. Post-money valuation is now about $5.3 billion. The company previously announced a $350 million Series A a year ago, then expanded it to $415 million due to strong demand. Now it has raised another $520 million from earlier investors Google, Mercedes-Benz, and B Capital, alongside new investors. Investors have paid progressively more for shares in each extension, valuing the company at roughly triple the initial Series A valuation of around $1.75 billion. Why not call this a Series B? The company says it's still in early stages and wasn't actively seeking funding, rather dealing with inbound interest. Fair enough. Another $520 million at a higher valuation would be hard to turn away, especially for tech as expensive to build as bipedal robots. For context, closely watched competitor Figure AI raised nearly $2 billion total since its 2022 founding before announcing a further $1 billion round last fall. Apptronik has partnered with Google DeepMind, GXO, and Mercedes-Benz to deliver embodied AI, robots capable of perceiving their environment and taking physical action based on reasoning, rather than just following fixed instructions. The robot is being built for unloading trailers, picking warehouse inventory, and tending machinery. Apptronik is no newcomer. Its humanoid work dates back to 2013, three years before the company was formally founded, when members of the Human Centered Robotics Lab from the University of Texas competed in the NASA-DARPA Robotics Challenge with a robot called Valkyrie. Jeff Cardenas, congratulations to you and your awesome team! I hope to meet you soon again! 👏🏼 ~~ ♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com

319

10 Comments

I've been building AI companies since 2014. Node — an AI company using LLMs years before ChatGPT existed (acquired by SugarCRM). Fireside — a creator operating system co-founded with Mark Cuban. Now I'm General Partner at Digital Indies, and this week my AI agents built 5 autonomous systems while I was in meetings: → A company from scratch. No instructions, no check-ins. → A self-learning engine that monitors markets and teaches itself. → An outbound pipeline. Finds leads, writes emails, sends, follows up. 0% bounce. → A trading portfolio with autonomous signal detection. → A content system. One idea → native posts for 5 platforms. For the first time, it's faster to build than to talk about it. That's what we're building at Digital Indies — an operating system on top of AI agents that takes the technical guesswork out. So founders, creators, and operators can get value without worrying about the rest. The future isn't learning to code. It's learning to architect. What would you build if your AI agents could work while you sleep?

41

2 Comments

That’s a hot news. Palantir Technologies is moving its #hq to Miami-Dade County. #Miami more and more is not looking anymore like #Singapore 🇸🇬, where we have similar #GDP of $500B +, but it’s becoming its own city that will dictate trends. Go #305

4

Prediction for 2026: OpenAI will shrink to become a minor player (potentially even worse) Their first-to-market advantage is thinning each week, and the critical mass that is needed to snowball the shift will come quickly. Google will most likely be the big winner, ending 2026 as market leader. What do you predict? 👇

38

16 Comments