The 2-Year AI Shift: Do This Before AI Improves Itself | Richard Socher

The 2-Year AI Shift: Do This Before AI Improves Itself | Richard Socher

📌 Head to https://granola.ai/marina and enter the code MARINA for 3 months off. Richard Socher is the fourth most-cited researcher in the history of natural language processing — he invented the word vectors and prompt engineering that run inside almost every chatbot you use. He sold his first startup to Salesforce, built You.com into a $1.5B unicorn, and in May 2026 raised $650M at a $4.65B valuation for Recursive: an AI that runs its own experiments and rewrites itself. In this conversation he explains why he thinks the self-improvement loop arrives within two years, which jobs grow and which disappear, and his hack for seeing the future — look at what only the wealthy can afford today, then ask which of it is bottlenecked on intelligence. Stay till the end for the first question he'd ask a superintelligence.Feeling behind on AI and don't know where to start? Start here. Feeling behind on AI and don't know where to start? Start here. We cover: * Reward hacking: why an AI told to raise customer satisfaction will quietly spin up a million bots that rate themselves 5/5 — and the new job of "reward engineering" that comes with it * His two-year timeline for recursive self-improving superintelligence — and why energy, not intelligence, becomes the next bottleneck * Why superintelligence isn't one thing: the "volumetric" view of intelligence, and the one dimension no lab is working on — an AI choosing its own goals * The elasticity rule for which jobs grow and which vanish — why illustrators got hit hard but software engineers are in more demand than ever * His hack for seeing the future: what only the wealthy can afford today — a personal tutor, a private assistant, a full medical team — and which of it is bottlenecked on intelligence * Why home robots are stuck on a hardware problem, not a software one * How he recruited co-founders away from DeepMind, OpenAI, and Meta * Whether a PhD is still worth it in AI — and what he tells parents to have their kids study instead * Where he'd invest right now: why "AI is to biology what calculus was to physics" * His first question to a superintelligence — and what still gives people meaning in 2035 Links: 📌 Subscribe to my free newsletter where I go deeper on AI tools, career strategies, and building with AI: https://siliconvalleygirl.beehiiv.com/subscribe?utm_source=spotify&utm_medium=video&utm_campaign=futureproof-sub&utm_content=RichardSocher [https://siliconvalleygirl.beehiiv.com/subscribe?utm_source=spotify&utm_medium=video&utm_campaign=futureproof-sub&utm_content=RichardSocher] 🔗 Instagram: https://www.instagram.com/siliconvalleygirl/ [https://www.instagram.com/siliconvalleygirl/] 𝕏 : https://x.com/siliconvalleymm [https://x.com/siliconvalleymm] 💼 LinkedIn: https://www.linkedin.com/in/marinamogilko [https://www.linkedin.com/in/marinamogilko] 📌 My Companies & Products: https://Marinamogilko.co [https://Marinamogilko.co]

Fei-Fei Li: In 10 Years There Will Be Only 2 Kinds of Workers

Dr. Fei-Fei Li created ImageNet — the dataset that triggered the deep learning revolution. Now she runs World Labs, where she just raised $1 billion to teach AI something it still can't do: understand physical space. She was named Time Person of the Year in 2025 and won the Queen Elizabeth Prize for Engineering. David Rogier founded MasterClass in 2015 on the idea that everyone should have access to world-class teachers. He has an MBA from Stanford and now builds all his own work tools with Claude Code instead of buying software — including a to-do list app that auto-deletes anything sitting on the list for more than a day and a half. In this episode, we get into what AI can already do and what it still can't, why the two loudest takes on AI — that it'll save the world, and that it'll take all the jobs — are both wrong, and what your kids' schooling and your own career actually look like in 10 years. We cover: * Why Fei-Fei Li says "the cost of intelligence goes to zero" is an irresponsible claim — and what human intelligence actually includes beyond language * Spatial intelligence: the four things AI still can't do (understand, reason about, generate, and interact with 3D space) — and why World Labs is betting $1B on it * The barbell effect in the workforce: why being "decent" at anything isn't enough anymore, and the two roles that will thrive — top 1% specialists and high-agency generalists * Why AI tutoring delivers the same learning in 60% less time — and why the real obstacle to changing education isn't the technology * David's full CEO stack: custom apps he built himself, "Davidify" — his own AI clone he gave his team — and why he stopped buying software * The one thing to do when an employee is scared to try AI: sit down with them, not send a YouTube link * What the job of product manager looks like today vs. five years ago — and why Fei-Fei only hires PMs who are riding the wave * Fei-Fei's advice for non-tech professionals: find someone under 25 and let them show you their world for a weekend * How long until AI handles the laundry — Fei-Fei's actual estimate Links: 📌Subscribe to my free newsletter where I go deeper on AI tools, career strategies, and building with AI: https://siliconvalleygirl.beehiiv.com/subscribe?utm_source=spotify&utm_medium=video&utm_campaign=futureproof-sub&utm_content=FeiFeiLi [https://siliconvalleygirl.beehiiv.com/subscribe?utm_source=spotify&utm_medium=video&utm_campaign=futureproof-sub&utm_content=FeiFeiLi] 🔗 Instagram: ⁠⁠https://www.instagram.com/siliconvalleygirl/⁠⁠ [https://www.instagram.com/siliconvalleygirl/%E2%81%A0] 𝕏 : ⁠⁠https://x.com/siliconvalleymm⁠⁠ [https://x.com/siliconvalleymm%E2%81%A0] 💼 LinkedIn: ⁠⁠https://www.linkedin.com/in/marinamogilko⁠⁠ [https://www.linkedin.com/in/marinamogilko%E2%81%A0] 📌 My Companies & Products: ⁠⁠https://Marinamogilko.co [https://Marinamogilko.co]

The 5 Levels of AI-Native: How to Get From Level 1 to Level 5 in 30 days | Peter Yang

Peter Yang spent 10+ years building products at Roblox, Reddit, Amazon (Twitch), and Meta. Then he walked away from a high-paying job at Roblox to go all-in as a solo builder. His newsletter Behind the Craft has 140K+ readers, and his podcast has deep conversations with 30+ AI leaders from Anthropic, Google, Cursor, and Replit. In this episode he breaks down how: how to move from using AI like a search engine to building systems that run your work, how to build a personal AI that knows your strategy, and the one thing to do this week to get ahead. We cover: * The 5 layers of AI adoption, from "I ask ChatGPT questions" to "AI runs my whole operation," and how to find which layer you're actually on * Why you should stop using ChatGPT and Claude chat and switch to Codex or Claude Code, even if you've never written a line of code * The one-day reset: cancel every meeting, brain-dump your real workflows out loud, and let AI turn them into skills that save hours every week * TRICK: the personal advisor skill — a one-page Google Doc with your goal, your principles, and what gives vs. drains your energy, so your AI checks your decisions before you make them * How he uses learnings.md so his AI remembers the right things about him and gets smarter every conversation * The "last 10%" rule that keeps AI content from becoming slop, and why human taste is the whole job now * How he builds a working personal app (a fitness tracker, a scam-email blocker for his parents) in a couple of hours * His honest fears: getting dumber on a flight with no wifi, and what he's doing differently to raise kids who still learn the fundamentals * The exact action plan for your next 7 days to go from consumer to builder Links: Subscribe to my free newsletter where I go deeper on AI tools, career strategies, and building with AI: ⁠https://siliconvalleygirl.beehiiv.com/subscribe?utm_source=spotify&utm_medium=video&utm_campaign=futureproof-sub&utm_content=PeterYang⁠ [⁠⁠https://siliconvalleygirl.beehiiv.com/subscribe?utm_source=spotify&utm_medium=video&utm_campaign=futureproof-sub&utm_content=PeterYang⁠⁠ ] 🔗 Instagram: ⁠https://www.instagram.com/siliconvalleygirl/⁠ [https://www.instagram.com/siliconvalleygirl/%E2%81%A0] 𝕏 : ⁠https://x.com/siliconvalleymm⁠ [https://x.com/siliconvalleymm%E2%81%A0] 💼 LinkedIn: ⁠https://www.linkedin.com/in/marinamogilko⁠ [https://www.linkedin.com/in/marinamogilko%E2%81%A0] 📌 My Companies & Products: ⁠https://Marinamogilko.co [https://Marinamogilko.co]

Godmother of Silicon Valley: How to Raise a CEO | Esther Wojcicki

Esther Wojcicki is the "Godmother of Silicon Valley." She raised three daughters: Susan, who ran YouTube; Anne, who founded 23andMe; and Janet, a professor and anthropologist. Her parenting advice got me through the hardest parts of new motherhood, and YouTube, which Susan built, is the reason I have a career at all. In this episode we cover: * The one rule that decides whether AI makes a kid smarter or just lets them cheat themselves: do the work first, then ask AI to grade it like a teacher * Why she's against banning phones, social media, or AI for kids, and what she'd teach instead * TRICK: the five-part system (Trust, Respect, Independence, Collaboration, Kindness) she used to raise three founders, and how to run it on yourself as an adult * How to rebuild trust in yourself if you were raised to chase everyone else's approval * The three sentences she told Anne when 23andMe was falling apart * The billion-dollar bet Susan almost didn't make, the one that became YouTube * Why she still tells parents to push for college in the AI age * What she says to every mom who's afraid AI will take her kid's job Links: Subscribe to my free newsletter where I go deeper on AI tools, career strategies, and parenting in the AI age: https://siliconvalleygirl.beehiiv.com/subscribe?utm_source=spotify&utm_medium=video&utm_campaign=futureproof-sub&utm_content=EstherWojcicki [https://siliconvalleygirl.beehiiv.com/subscribe?utm_source=spotify&utm_medium=video&utm_campaign=futureproof-sub&utm_content=EstherWojcicki] 🔗 Instagram: https://www.instagram.com/siliconvalleygirl/ [https://www.instagram.com/siliconvalleygirl/] 𝕏 : https://x.com/siliconvalleymm [https://x.com/siliconvalleymm] 💼 LinkedIn: https://www.linkedin.com/in/marinamogilko [https://www.linkedin.com/in/marinamogilko] 📌 My Companies & Products: https://Marinamogilko.co [https://Marinamogilko.co]

Grant Lee : From "The Worst Idea" to a $2.1B Company | Gamma

Grant Lee is the co-founder and CEO of Gamma — the $2.1B AI tool that turns one prompt into a finished presentation, now used by 100 million people. We covered: * How Gamma hit $100M in revenue with a team of just 50 — profitably, on only $23M of funding * The two signs you have product-market fit before the money shows up, and why "your friends will lie to you" about your product * How to run a full fundraise in two weeks — Grant did 100+ pitches and how he turned a wall of no's into yes's * How to tell if an idea is worth a year of your life (he picks by energy, not spreadsheets) * Why a top investor called Gamma "the worst idea I've ever heard" — and what that investor actually got right * The exact first-30-days plan he gives any founder starting today Links: Subscribe to my newsletter: ⁠⁠⁠⁠⁠https://siliconvalleygirl.beehiiv.com/subscribe?utm_source=spotify&utm_medium=video&utm_campaign=futureproof-sub&utm_content=GrantLee [⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://siliconvalleygirl.beehiiv.com/subscribe?utm_source=spotify&utm_medium=video&utm_campaign=futureproof-sub&utm_content=GrantLee ⁠] Instagram: ⁠⁠⁠⁠⁠⁠⁠https://www.instagram.com/siliconvalleygirl/ ⁠⁠⁠⁠⁠⁠⁠ [https://www.instagram.com/siliconvalleygirl/ ] X: ⁠⁠⁠https://x.com/siliconvalleymm⁠⁠⁠ [https://x.com/siliconvalleymm] LinkedIn: ⁠⁠⁠https://www.linkedin.com/in/marinamogilko⁠⁠⁠ [https://www.linkedin.com/in/marinamogilko] My Companies & Products: ⁠⁠https://partnerships.marinamogilko.co [https://partnerships.marinamogilko.co ]