Billion-Dollar Babies, AI Drama, and the Race to Reinvent Siri

Hey everyone, welcome back To Tech Insider Weekly. Lauren, We have Got Some Wild AI stories This week, And Imm seeing The Same pattern repeat across All of them. Speed is breaking things. Wild doesn't Even cover IT, Derek. Weree talking Billion-Dollar valuations happening in weeks, Not years. And Imm tracking This pattern Because IT signals something about where capital thinks defensibility actually lives, Right? So here's Whatss fascinating. We've Got companies raising at 44 billion valuations Two Months after launch. Two Months! Ivee watched The competitive window compress at big Tech, And This is What happens When IT compresses To basically nothing. Yeah, Yeah. And Fei-Fei Liss raising 55 billion before shipping A product. That tells me something about desperation in The market, And Imm Not sure That ends Well. Oh, Which raises The question. Is This Real defensibility, or Are We watching Another bubble inflate in Real time? Because Ivee sat through enough funding rounds To know The difference. Exactly. And IT Gets messier. Weree diving into The founder Drama at Thinking Machines And other AI unicorns. Defensibility reveals The cracks in The foundation. Thatss The pattern I keep seeing. Because Velocity And pressure? They break leadership teams. Ivee watched This at Two Different Startups I operated at. Founders Who could collaborate on A shoestring budget Cantt survive A Billion-Dollar Valuation in Months. Mm-hmm, Mm-hmm. Plus, get This, 155 million downloads of malicious AI coding extensions stealing source Code. Thatss Not Just A security problem. Thatss The Infrastructure layer failing completely. Just casually compromising entire codebases. No big deal. Then Weree looking at The Assistant wars. Appless bringing Gemini-powered Siri in February, Claudess going deep into enterprise. Distribution Versus Capability. Classic Battle. And here's The thing. Does Distribution beat technical Capability? Thatss The question Imm watching. Ivee lived through IT at companies Ivee worked with. And IT always comes down To defaults. 100%. And Weree closing with The Infrastructure revolution happening underneath All of This. This is The layer Ivee trained myself To Watch. Because What Gets funded today determines Whatss possible in 18 Months. Photonic chips, Derek. Neurofos Just raised 1100 million, Betting on light Instead of Electrons. Thatss The Infrastructure layer determining Whatss viable To build tomorrow. Which could completely reshape startup economics If IT works. And here's The thing. Investors know IT. Thatss why The money's flowing. So No pressure for today's episode. Just Another week in AI chaos. But every story connects back To The Same pattern. Letss jump in. Okay, Derek. So I need To Tell You about something That made me pause mid-coffee This week. And Not in A good way. Oh No. What happened? A startup called Recursive Intelligence Just hit A 44 billion Valuation, Two Months after launch. And Imm looking at The timeline Thinking, I watched 2021 happen in Real boardrooms. This feels Like Weree running That Same playbook Again. Wait, Wait, Wait. Two Months? Two. Months. They raised 3000 million at A 44 billion Valuation. And here's The thing. Theyree building AI systems That design AI chips. Infrastructure. This timeline is Wild. What happened To proving product market fit? I mean, To be Fair, Imm on Both sides A little bit Here. Because Yes, The Speed is insane. But If Youree building Infrastructure That could genuinely accelerate The entire AI stack, Thatss Different From wrapping GPT in A nice UI. Right, Right, Right. And Look, Synthesia I actually get. Theyree wired into enterprise training workflows. Once Youree embedded Like That, switching costs matter. Thatss The pattern I Watch for. And That one actually makes sense To me From A defensibility standpoint. Tell me more. So, Synthesia has Real Customer lock-in. Ivee watched This at Big Tech. Once Youree embedded in someone's production workflow, The switching costs become brutal. Theyree inside The software stack. Thatss A genuine moat. Versus someone Who Just duct tapes Claude To A workflow And calls IT product market fit. Exactly. Ivee watched The competitive window compress dramatically Over The years. What took three years To copy five years ago now takes three Months. You Cantt Just layer A model on top And expect To survive. Which brings us To The Most fascinating Valuation of All, Fei-Fei Liss World Labs, in funding talks at 55 billion. Wait, The Stanford AI pioneer? The ImageNet legend? The one And only. And here's Whatss interesting. Theyree reportedly building spatial Intelligence AI. But Derek, pre-product at 55 billion? Ivee watched Founders pitch me on less air than That, And Ivee had To say No. So Weree paying for The founder pedigree now? I mean, If anyone's earned That premium, Itss her, But pre-product is Still pre-product. I mean, partly. But Honestly, If anyone's earned That premium, Itss her. But here's What Imm getting in DMs constantly From Founders. How do We build defensibility When The model layer is commoditizing? And nobody wants To say out loud That Itss A Real problem. And What do You Tell them? Firmly. Customer moats. Regulatory moats. Data moats That actually matter. Not Just We scrape The internet better. Because Ivee watched The pattern. The defensible companies own The integration, Not The Intelligence layer. Okay, now Imm listening. Who is This? Theyree Not named in The article, But here's What told me This was Real. Theyree built on open AIss models, yet they've found defensibility. That means workflow depth. Customer pains So acute That rip-out costs Are brutal. Or They solved something That big Tech actually Cantt do at scale. Itss The Infrastructure underneath That matters. I learned This covering smart fridges back in 2019. You Watch Whatss being built under The product, Not The product itself. So here's my Real concern. Are We looking at actual businesses? Or am I watching 2021 All Over Again? Because I was in Those boardrooms When capital was abundant And everyone was valued. The pattern feels identical. Okay, So I Think Itss Both. Both? Explaining quickly. Some of These, Like Synthesia with Real enterprise lock-in, or Recursive with actual Chip innovation, Those feel Like Real architectural problems being solved. But When Weree throwing billions at pre-product based purely on foundation, Underpedigree, Thatss speculation, Not innovation. And The thing That worries me Most? Ivee watched That shift happen in Real time at multiple companies. The conversation goes From Whatss Your Tech advantage To Who actually owns The Customer relationship. But capital keeps flowing anyway. Thatss When things get dangerous. Well, When You have Nvidia And Alphabet throwing money around. Fair point. But The pattern Ivee tracked across portfolio companies is unmistakable. Constraints force Real innovation. The Ones That actually survive Are solving problems big Tech won't or Cantt touch. Not The Ones with The biggest models or shiniest pitch decks. Which means The defensibility question is Everything Right now. Everything. And Speaking of things That should worry us about How fast This ecosystem is moving? Oh No, What now? With genuine concern. The founder dynamics And team cohesion behind These valuations. Because When Youree moving This fast, When capital is This abundant, When pressure is This relentless, Ivee watched teams fracture. Ivee seen IT happen. And IT destroys companies faster than bad Tech ever does. Speaking of speculation And crazy valuations, Derek, Ivee watched This exact pattern unfold at companies Ivee operated at. The Drama behind The scenes is What actually predicts failure or success. Oh man, Yes. So get This. Wall Street Journal Just dropped an exclusive about Thinking Machines, And Imm watching The Same pattern Ivee seen before When Speed And money collide. Riveted is The Right word. But When The Times is documenting founder Drama at A unicorn, IT signals something fundamental is breaking. Thatss Not Just gossip. Thatss pattern. Yeah, Yeah, Yeah. So here's The thing. Thinking Machines is one of The hot AI Startups everyone's been watching, Right? And The WSJ is reporting on founder dynamics That basically dealt A blow To The company. And Theress Another unnamed AI unicorn with similar Drama. Same story, Different company. Another unnamed AI unicorn, Which means The Real story probably Cantt go on The record Because Itss Still bleeding money And court battles Are happening behind closed doors. I know. But Whatss fascinating is The pattern Here. Weree seeing This across multiple companies. Right, Right. And here's What I find interesting. When Youree scaling fast, founder relationships Are The first thing That breaks. Ivee literally seen This happen at Two Different Startups during funding rounds. Itss predictable. Wait, really? What breaks down When Youree scaling That fast? Itss The pressure. You go From zero To A billion dollar Valuation in Months And suddenly every decision is existential. The founding team dynamics That held up on A shoestring budget collapse under Velocity. Who owns What? Who makes The call? Those conversations become combustible. So Youree saying Velocity is The thing That reveals Everything. The Speed reveals The problem. And Thatss The thing nobody admits in pitch meetings. These weren't built on solid operational foundations. They Were built on hype cycles And FOMO capital. Ivee watched The difference at multiple companies. Okay, Okay. Harsh But Fair. And That connects To What We Were Just talking about with recursive And Those insane valuations, Right? Because Itss The Same pattern. Velocity creates cracks And When The money's That fast, Theress No time To patch them. Exactly. If You raise That 44 billion Two Months in, You don't have time To build trust, process, or culture. Youree Just trying To survive The Velocity. And Silicon Valley is treating IT Like prestige TV, When really Itss Just The Infrastructure of trust breaking And Real-time. Because IT kind of is. But Theress A Real lesson Here. The gap between AI hype And operational reality is enormous. And Thatss What The WSJ reporting is showing. The Infrastructure of trust falling apart in real-time. Thatss What nobody wants To admit. Right. You can have The Best technology in The World. But If Your Founders Cantt actually work together under pressure, And Ivee seen Founders Implode, Itss brutal And Itss Real. The company collapses. Tech doesn't save You From broken founder dynamics. You have nothing. The company dies. 100%. And Honestly, Imm getting DMs From Founders constantly asking How To avoid This. And my answer is always The Same. Build The trust And The communication framework before You need IT. But nobody wants To hear That When capital is flooding in. But can You slow down in This environment? Ivee watched This at Big Tech. The market rewards Speed Over stability every single time. And Founders know IT. I mean, To be Fair, Thatss The trap, Right? The market punishes patience Right now. Okay, So plot twist. While Founders Are imploding, The ecosystem itself is revealing That our entire supply chain has A trust problem. Wait, What? So, there Are Two malicious VS Code extensions, AI-powered coding assistants, That have been downloaded 155 million Times combined. Thatss 155 million Developers Who didn't Check What They Were installing. Oh, Shut up, No, Wait, 155 million Developers? Thatss Not A bug, Thatss A systemic vetting failure. Yeah, Yeah, And Theyree stealing developer source Code And sending IT To China-based servers. Which, Right, Itss Not Just A technical supply chain problem, Itss A geopolitical one. And Theyree Still available? Still on The official Visual Studio Marketplace as Weree recording This. Which, Right, Right, Right, nobody's vetting The Infrastructure layer. Thatss insane. How does That Even happen? I mean, Ivee negotiated Infrastructure decisions at companies Ivee operated at, And This is Exactly The blind spot Ivee watched. We move So fast, We don't validate The supply chain. Well, Thatss The question, Right? Developers trust These tools Because Theyree supposed To make coding faster. Theyree advertised as AI assistants. And underneath, Theyree exfiltrating Your entire codebase. So The very tools Weree trusting To build AI have active exfiltration happening. Thatss Not A feature problem, Thatss A foundational trust collapse. Exactly. And Itss Not Just A technical problem, Itss A trust problem. When 155 million Developers install something malicious, What does That say about our ability To vet This stuff? Ivee watched This Same blind spot at big Tech for years. IT says Weree moving too fast To Check. Again. This is The pattern Ivee been watching, Velocity Over verification. We did This in 2020, Weree doing IT Again now. Again. Itss The Same pattern. You know What strikes me? Both of These stories, The founder Drama And The malicious extensions, Theyree symptoms of The Same disease, Speed without foundation. And Ivee watched IT destroy companies. Which is What, Exactly? Velocity Over verification. Weree building So fast, Weree Not checking If The foundation is solid, whether Thatss founder relationships, board governance, or The tools in our supply chain. Thatss The operating lesson Ivee learned twice Over. Oh, Thatss good. Thatss really good. Velocity Over verification. Thatss The pattern Weree watching everywhere Right now. Thank You. And Itss Not slowing down anytime soon. Because The incentives Are All pointing The wrong direction. No, No, No. Which brings us To The Next battlefield. Which is The AI Assistant wars, Because If Weree talking about trust And Who You let into Your workflow... Exactly. Appless about To unveil Gemini-powered Siri in February, And Anthropic Just launched Cloud And Slack. Yeah. So The question becomes, Who do You trust To live inside Your Most critical workflow? Your phone? Your Slack? Your Code editor? Your development environment? Thatss Not A consumer question anymore. Thatss an Infrastructure question. And Thatss where Distribution meets Capability. Letss dig into That Next. Okay, So after All That founder Drama And security chaos We Just covered, Letss talk about The layer where This actually Gets decided. Distribution. Oh man, The Siri Overhaul? Yes. February. Gemini-powered Siri with Two versions coming. And This is Whatss fascinating. IT signals Google And Apple Are taking Distribution seriously. Thatss The Real story. Wait, Wait, Wait, Gemini-powered? So Apple basically admitted Distribution And model commoditization aren't The Same thing? I mean, Thatss one way To read IT. But here's Whatss fascinating From A go-to-market perspective. They don't need To build The Best AI. They Just need To ship IT To A billion devices. Right, Right, Right. Distribution advantage. Exactly. Derek, This is The Classic innovator Versus distributor problem, Right? Totally. So here's The thing. Anthropic Just launched interactive Cloud apps. Slack integration, Canva, Figma, Box. You can now send Slack messages directly From Cloud, generate charts, access Cloud files. Mm-hmm. Thatss smart positioning. Yeah, Yeah. Theyree going after The enterprise layer where people actually work. But here's my question. Does technical Capability matter When Apple can Just ship IT pre-installed on A billion devices? Thatss The pattern Ivee watched at Big Tech. Best product loses To Distribution every time. Okay, But hear me out. Here We go. Here We go. Capability Versus Distribution. Cloud is objectively better at complex reasoning Right now. But I live in Slack eight hours A day Like probably half The knowledge workers out there. So where do I actually want my Assistant? Already there. Hmm. Fair point. Imm Not pulling out my iPhone To analyze A document When Cloud's embedded in my workflow. Thatss Not A feature comparison. Thatss A friction comparison. But Lauren, And I say This with love. Most people aren't knowledge workers living in Slack. Most people Are normal humans with iPhones Who Just want their Assistant To Not suck. Okay. Okay. Thatss Fair. And Thatss where Appless Distribution advantage is devastating. They don't need The Best AI. They need good enough AI Thatss Already in Your Pocket. But Thatss Exactly why Anthropicss enterprise focus is brilliant. And Imm tracking This pattern everywhere. Theyree Not competing on Distribution. Theyree competing on becoming The default where actual work lives. Right. But... Think about IT. Apple Gets The consumer layer. Anthropic Gets The enterprise layer. Theyree Not actually fighting The Same Battle. Oh. Huh. Thatss actually A really good point. Right, Right, Right. Theyree Not Even fighting The Same Battle. Different layers, Different moats. Thank You. So The Real question is, Who wins The middle? Like The prosumer Who lives in Both worlds? Thatss where The Real money is though. Exactly. And I don't Think We know yet. Because Appless announcing Two New Siri versions, Which, by The way, signals Theyree Still figuring This out. Dude, Yes. The fact That there Are Two versions tells You Everything. But Theyree Not confident in The first one, So Theyree hedging. Oness probably The Gemini-powered upgrade for iOS 2644, And The other is The Full chatbot version, CamPost, coming in iOS 27 at WWDC. Classic pattern. Wait, Theyree making Siri an actual chatbot? Yeah, Voice And text inputs. The Bloomberg report says Itss The first version That actually lives up To The promises Apple made back in June 2024. Thatss Two years of RDD finally shipping as Real architecture. Two years later. Which, by The way, is The timeline Ivee watched at companies Ivee operated at. They get The strategy Right eventually, But Speed matters less than people Think When You control The default. Apple doing Exactly What They always do. Shipping late with A better product. Betting That The default matters more than timing. I guess. But here's What keeps me up at night. Weree seeing This race heat up across every layer. And I keep asking Founders, does anyone actually want To juggle three Different assistants depending on What Theyree doing? Thatss The Real constraint. Thatss The Real UX question, isn't IT? Because The default Assistant is incredibly hard To displace Once IT sticks. Ivee watched That play out. Yeah, Because From A business model perspective, The winner is whoever becomes The default. And Ivee watched This pattern multiple Times now. Defaults Are incredibly hard To dislodge Once They stick. Which is why Appless Distribution advantage is So powerful. Even If Claude is technically better Right now, That doesn't matter If people never have A reason To switch away From What came with their phone. Exactly. But To be Fair, enterprise is Different. Ivee negotiated Those budgets. IT departments don't care about flashy features. They care about total cost of ownership And measurable productivity gains. If Claude proves IT actually moves That needle… Then Anthropic wins The enterprise layer. And Apple wins consumer. Different games, Different modes, Different timelines. Thatss actually smart market segmentation. Yeah, Yeah, Exactly. Okay, So Speaking of Different games, We need To talk about Whatss being built underneath All of This. Because Thatss where The Real Infrastructure story is. Infrastructure? Infrastructure. Because While everyone's arguing about Which Assistant wins, Theress This whole foundational layer being rebuilt. Optical computing. New Chip architectures. The stuff That actually enables What We Just talked about. Oh, The Neurophos stuff? Yes. 100 million dollars To replace Electrons with Photons for AI compute. This is The foundational layer That enables Everything We Just talked about. I learned To Watch This stuff at CES years ago. The unsexy Infrastructure layer That makes The shinier products possible. Okay, You have my attention. Because This is Exactly What Imm watching. Infrastructure plays usually get boring. Nobody DMs You excited about Chip architecture. But When capital moves Like This, Thatss The signal. Okay, So We Just talked about The AI Assistant wars happening at The application layer, Right? Apple, clod, All That. But here's What Imm really watching. Yeah, Yeah. But While everyone's watching That, Theress This Absolutely Wild Infrastructure bet happening underneath. And This is The pattern Ivee trained myself To see. What Gets funded at The foundation layer determines Whatss possible 18 Months later. What You find? So A startup called Neurophos Just raised 110 million dollars in A Series A. Wait, 110 million Series A? Thatss What Gets my attention. That signals Gates Frontier sees something fundamental shifting. Yeah, Yeah. To replace Electrons with Photons for AI compute. Okay, Derek, You gotta explain That Like Imm five. What does That actually mean? basically rethinking The fundamental physics of How chips work. Exactly! And here's The thing. Ivee watched This at Big Tech for years. The Real signal isn't The application layer. Itss watching What Infrastructure is being built now. Because Thatss What enables The Next generation of companies. When Gates Frontier is leading 1100 million round for something This fundamental, Theyree Not Betting on Next quarter. Theyree Betting on 2028, 2029. I saw The Same pattern When quantum computing started getting Serious money. Okay, But, I mean, To be Fair, we've heard The Optical computing is The future pitch for A decade. Whatss actually Different This time? Whatss changed in The physics, Not Just The hype cycle? The AI compute bottleneck. Like, Weree hitting actual physics limits with traditional chips. The energy costs alone for training These models Are becoming unsustainable. And That creates The opening for something fundamentally Different. True. And Itss Not Just Neurophos. Theress This Oregon startup, Ahead Computing, former Intel architects, Just raised Another 300 million for A Totally New CPU architecture. When Infrastructure builders start rethinking From first principles, Thatss A signal. Wait, Wait, Wait. So we've Got Photonic chips Over Here, New CPU architectures Over there. Whatss The pattern? The pattern is That The foundation is being rebuilt. This is an incremental improvement. This is going back To physics And asking, What If We stop doing IT This way entirely? Thatss where Real architectural change happens. And here's Whatss fascinating From The startup side. Imm getting DMs From Founders All The time asking, where should We be investing in Infrastructure? This is Exactly What They should be paying attention To. Yeah? Youree building on today's Infrastructure. But If This Photonic Chip bet actually pays off, The unit economics change completely. Ivee watched This pattern at companies Ivee operated at. Suddenly, What was prohibitively expensive becomes viable. Exactly. Thatss What Imm saying. The Infrastructure shift creates New modes. The companies That survive aren't The Ones Betting on today's chips. What costs 1000000 in compute today might cost 100000 in three years. Ivee seen That math shift before at portfolio companies. Suddenly makes entirely Different business models viable. Thatss The Real inflection point. Right, Right, Right. So The question becomes, When do Photonic chips actually matter for builders? Not today. I mean, Neurofos raised 1100 million. Theyree probably Two, three years From production. Yeah? But, And This is important, The Infrastructure being funded today determines The playing field tomorrow. Ivee literally heard investors ask This in pitch meetings. What happens When compute Gets 10x cheaper? Thatss The New question. Oh man, Thatss A brutal question. Because IT forces Founders To ask, What do I actually own That doesn't evaporate When The Infrastructure changes? IT is. And Ivee watched Startups get blindsided by This. If Your entire moat is Just, We can afford To Run This expensive model, Thatss Not defensibility. Thatss A countdown timer. So You need defensibility That survives The Infrastructure shift. Data. Customer relationships. Regulatory positioning. Things That don't become obsolete When chips get better. Exactly. Data modes. Locked-in workflows. Regulatory positioning. Things That don't evaporate When Infrastructure Gets cheaper. Those stick around. Okay, So here's What Imm watching. A head computing Neurophos. Theyree Both targeting AI workloads specifically. Thatss Different From general-purpose chips. Theyree saying, We know What AI needs. Weree building for That. Thatss focused architecture Thinking. Which, Right, Right, Right. Thatss How You win. Which makes sense, Right? From every funding round Ivee tracked, AI compute is now The dominant cost structure in Data centers. That changes Everything about Whatss worth building. Itss Wild. And The energy requirements Are absurd. Some of These training runs consume as much power as A small town. Thatss Not sustainable. Thatss A problem demanding A solution. Just casually training A model, consuming A town's worth of electricity? Which, by The way, is The kind of cost structure That forces innovation. Ivee negotiated enough Data center budgets To know Thatss Not sustainable. Yeah, No big deal. But This is why photonics matters. If You can cut energy by 80, 90 percent, suddenly The entire economics of AI deployment change. Suddenly AI Gets economically viable in places IT wasn't before. Yes, Edge devices, smaller companies, real-time applications. So The Infrastructure bet isn't Just about making AI faster. Itss about making AI accessible. Exactly. And Look, I don't know If Neurophos specifically wins, But someone's cracking I learned at Big Tech That When The incentives align This clearly – demand, physics, capital – The solution tends To emerge. Right? Because The demand is there, The physics is possible, And now The money's flowing. Thatss always The signal. All three need To align. And When You see capital flooding into Photonic chips And New architectures? Thatss Not hype. Thatss investors pricing in That The playing field is about To shift. That signal matters. Itss A huge signal. The Infrastructure layer is being rebuilt. And That means anyone building on top needs To understand Whatss coming. So for anyone listening who's building on AI Right now, don't Just Watch The models. Watch The Infrastructure layer. Thatss where The Real game is being decided. Absolutely. Because Chip economics determine Who Gets To play The game at All. If You Cantt afford The compute, Youree out. Photons Over Electrons. Who had That on their 2026 bingo card? Right? IT sounds Like science fiction, But This is actually happening Right now. These aren't theoretical papers. Theress Real money, Real physics, Real timelines. Next 18 Months Are going To be fascinating. Cantt Wait To see What Gets built on top of This. Because When The Infrastructure layer shifts, Everything else reorganizes around IT. Okay, So From 44 billion valuations Two Months after launch To malicious Code extensions compromising developer trust, Thatss The range of This moment. And Ivee watched enough cycles To know What This signals. Right, Right, Right. I mean, The range We cover today, From founder implosion To malicious extensions To Photonic chips, Thatss The pattern. Lauren, Whatss Your big takeaway? Honestly, The question isn't Whatss Your Tech anymore, Itss Whatss Your Customer moat? I watched That shift happen in Real time at The companies Ivee operated at. And Itss Everything now. Exactly. And Thatss What separates The Real businesses From The 2021 crypto vibes. Real moats, Real customers, Real... Defensibility VS. Hope-Driven Architecture Yeah, Yeah. Look, If This episode made You Think about where The Real value actually compounds, subscribe wherever You listen And leave us A review. Got A founder or Infrastructure story We should dig into? Tag us on social media. New episodes drop every Wednesday. And If Youree building something Right now, Watch The Infrastructure layer. Thatss where The Real game is being decided. Thanks for hanging with us today. And maybe, You know, Check Your VS Code extensions. Derek, Yes, Please do That. Check Those VS Code extensions, Watch The Infrastructure layer, And Welll see You Next week. Thanks for listening.