Welcome to the very first episode of This Week in AI! Host Oliver Korzen kicks things off with an exclusive interview featuring Aaron Walls (CEO) and Andrius Useckas (CTO), the founders of ZioSec. Together, they dive deep into the world of OpenClaw security—exposing the vulnerabilities they've discovered in the wild, breaking down real attack vectors, and revealing the essential steps you need to take to protect your agent from exploitation.🔒 Key Takeaways:1. Why Mac Mini might be your safest local setup2. The "Lethal Trifecta" that makes AI agents vulnerable3. How indirect prompt injections hijack your agent through email & documents4. Why 26% of OpenClaw skills contain vulnerabilities5. Input sanitation & judge models as defensive layersZioSec: AI agent assisting in penetration testing | https://ziosec.comTimestamps:00:00 - Why Mac Mini isn't the right setup for everyone12:30 - The right way to host OpenClaw (hint: not AWS)16:11 - Understanding your attack surface27:33 - Why third-party skills are a security risk36:30 - Start with sandboxes & read-only permissions40:03 - Career advice for college students

Google released Gemini 3.0 this week, and I’ve been hands-on with it every day since launch. After diving into all the new capabilities, performance upgrades, and quality-of-life improvements, I’ve distilled the most important updates you should know. In this breakdown, I walk through the biggest feature changes, what they mean in real-world use, and how Gemini 3.0 stacks up against the other frontier models.

Jason Calacanis sits down with three CEOs building at the bleeding edge of the agentic revolution: Mitesh Agrawal (Positron AI), Alex Elias (Qloo), and Kash Ali (TaxGPT). They discuss the rapid rise of Open Claw, an open-source platform that is fundamentally changing work by allowing users to create autonomous agents. "Saving 30 minutes a day... that’s like getting three more weeks a year back." "I don't need a technical recruiter today." We explore how autonomous agents are moving beyond "Copilots" to become independent teammates:

• The AI Executive Assistant: How Mitesh Agrawal uses agents to automate inbox filtering and Slack drafts, recapturing 6% of his year.
• The Hiring Hiatus: Kash Ali screened 1,000 job applications in two hours with "flawless" accuracy, replacing 40+ hours of manual labor.
• The $1M One-Person Firm: How AI is solving the shortage of 340,000 accountants by allowing a single person to command an "army of agents".
• Democratizing Luxury: Scaling the "chauffeur and concierge" experience to everyone through taste-based AI and structured inference.
• Autonomous Media: A demo of an agent that transcribes, clips, and captions viral content in five minutes, a process that previously took an hour.The agent era is here. Are you building your replicant, or are you being left behind by a 10x more efficient workforceThis Week In Startups is made possible by:Notion - https://www.notion.com/twistQuadratic - https://www.Quadratic.ai/twistTimestamps: 00:00 Welcome to This Week in AI!03:17 How are CEO’s using OpenClaw?09:25 Notion - Notion brings all your notes, docs, and projects into one connected space that just works with AI built right in. Try Notion, with Notion Agent, at notion.com/twist00:19:09 — Automating the "Great Hiring Hiatus”00:19:20 Quadratic - Bringing the productivity boost of AI into your spreadsheets. Visit [quadratic.ai/twist](https://www.quadratic.ai/twist) to sign up and use the code TWIST to get one free month of their pro tier subscription.00:27:11 — Saving Independent Film with AI Extras00:33:04 — Solving the Global Accounting Crisis00:36:36 — The $1 Million One-Person Tax Advisory00:43:05 — Demo: The Autonomous Content Clipper0:00 Making Clips with OpenClaw01:07:39 — Micro-Agents vs. "Ultron”01:10:23 — The Developer, the CEO, and the AI Joke (Outro)Subscribe to This Week in AI on Apple: https://podcasts.apple.com/us/podcast...Thanks for watching!🤖 If you want to stay ahead of the curve on all things AI, make sure to join our community across all platforms:📩 Get the Weekly Newsletter: https://thisweekinai.ai/📺 Subscribe on YouTube: [   / @thisweekinaipodcast  ](   / @thisweekinaipodcast  )📸 Instagram: [www.instagram.com/thisweekinaipodcast](  / thisweekinaipodcast  )📱 TikTok: [www.tiktok.com/@thisweekinaipodcast](  / thisweekinaipodcast  )✖️X:https://x.com/ThisWeeknAIFollow Jason:X:   / jason  LinkedIn: [  / jasoncalacanis  ](  / jasoncalacanis  )*Thank you to our partners:09:25 Notion - Notion brings all your notes, docs, and projects into one connected space that just works with AI built right in. Try Notion, with Notion Agent, at notion.com/twist00:19:20 Quadratic - Bringing the productivity boost of AI into your spreadsheets. Visit [quadratic.ai/twist](https://www.quadratic.ai/twist) to sign up and use the code TWIST to get one free month of their pro tier subscription.Check out all our partner offers: https://partners.launch.co/

This week we sit down with three founders building at the frontier of AI infrastructure, evaluation, and hardware: Chase Lochmiller (Crusoe), Naveen Rao (Unconventional AI), and Anastasios Angelopoulos (Arena). We dig into the real bottlenecks slowing down the AI buildout, how to standout in the job market as AI matures, and the explosive government standoff between Anthropic and the Pentagon.

We explore how AI infrastructure, open source models, and policy are reshaping the industry from the ground up.

• The Anthropic-Pentagon Standoff: The U.S. government blacklisted a leading domestic AI company for the first time. Who sets the rules for AI in warfare?
• Jevons Paradox & AI: Every efficiency gain in intelligence-per-watt creates more demand, not less. AI infra is positioned to capture the largest market humanity has ever built.
• OpenClaw & Open Source Mania: The open-source AI agent that just became GitHub's most-starred project. It now runs locally on your phone.
• Should You Learn to Code?: Yes, but not as a trade skill. Bootcamp-style training is dead. Understanding how machines work is what separates leaders from everyone else.
• Rethinking the Computer from Scratch: Naveen's team is building chips that mimic biology, targeting 1,000x efficiency gains.
• The Coming Labor Crisis: Anastasios argues most valuable work is just training humans for repeatable tasks, and that work is disappearing. Chase counters that digital labor is a new variable in the GDP equation.

This Week In Startups is made possible by:

Notion - https://www.notion.com/twist

Quadratic - https://www.Quadratic.ai/twist

Timestamps:

00:38 — Welcome to This Week in AI

01:21 — Show intro and guest introductions (Chase from Crusoe and Anastasius from Arena)

02:22 — Crusoe's Stargate partnership: 1.2GW campus for Oracle and OpenAI in Abilene, Texas

03:09 — Rethinking computing from first principles for 1,000x efficiency

06:53 — How Arena went from a Berkeley side project to half a billion conversations

09:28 — Quadratic - Bringing the productivity boost of AI into your spreadsheets. Visit https://www.quadratic.ai/twist to sign up and use the code TWIST to get one free month of their pro tier subscription.

12:01 — Open source AI explosion: On-device models, token costs, and adoption

15:31 — Defining AGI: Clay Math millennium problems, moving goalposts, and what AI still can't do

22:08 — Notion - Notion brings all your notes, docs, and projects into one connected space that just works with AI built right in. Try Notion, with Notion Agent, at https://www.notion.com/twist

23:07 — System 1 vs. System 2 thinking and where compute actually happens

26:31 — Crusoe Spark: Modular data centers that go where the power is

34:33 — Model compression is moving faster than anyone predicted

36:51 — Jevons Paradox, the employment question, and the coming labor crisis

45:53 — Abundance, advice for young people, and the importance of learning to code

64:41 — AI in warfare: Corporate responsibility, inevitability, and national obligations

Subscribe to This Week in AI on Apple: https://thisweekinai.ai/spotify

Subscribe to This Week in AI on Spotify: https://thisweekinai.ai/apple

Thanks for watching!

🤖 If you want to stay ahead of the curve on all things AI, make sure to join our community across all platforms:

📩 Get the Weekly Newsletter: https://thisweekinai.ai/

📺 Subscribe on YouTube: https://www.youtube.com/@ThisWeekinAIPodcast

📸 Instagram: https://www.instagram.com/thisweekinaipodcast

📱 TikTok: https://www.tiktok.com/@thisweekinaipodcast✖️X: https://x.com/ThisWeeknAI

Follow Jason:

X: https://twitter.com/Jason

LinkedIn: https://www.linkedin.com/in/jasoncalacanis

Follow Oliver:

https://x.com/oliverkorzen

Check out all our partner offers: https://partners.launch.co/

Jason Calacanis sits down with two founders building at the frontier of AI products and infrastructure: Tanay Kothari (Wispr Flow) and Richard Socher (You.com & AIX Ventures). They dig into the Citrini "doom post" that rattled markets, the real economics of automation waves, and how AI-first teams are being rebuilt from the ground up.

We explore how the agentic shift is reshaping companies, costs, and careers:

• The AI Doom Post Debunked: Why the viral Citrini Substack predicting 10% unemployment and an S&P crash is completely wrong.
• AI-First Hiring: How Wispr Flow shipped their #1 Android voice app with one engineer running ten Claude agents simultaneously — and stopped hiring junior developers entirely.
• Customer Support at Scale: How Wispr went from needing 200 support agents to 4 by automating 15 distinct support loops, from bug tickets to product feedback.
• The Cognitive Surplus: Why automation doesn't destroy jobs, it creates the next wave of businesses nobody has predicted yet.
• Open Source vs. Closed Models: Why DeepSeek forced every major lab to innovate faster, and how Wispr built its own 90% gross margin voice model by starting on open source.
• Taiwan & the Chip Risk: Apple's Mac Mini onshoring, the 2027 CCP military warning, and the five contingency plans the world is quietly building around TSMC.
• OpenClaw Mania: Producer Oliver joins the show to talk about how he is using OpenClaw to do his tasks for him.

This Week In Startups is made possible by:

Notion - https://www.notion.com/twist

Quadratic - https://www.Quadratic.ai/twist

Timestamps:

00:01:16 - Welcome & Guest Introductions: Tanay Kothari (Wispr Flow) and Richard Socher (You.com)

00:03:33 - Why Voice Dictation Finally Works: How Wispr turns rambling speech into send-ready messages

00:05:43 - Foot Pedals, Rings & the Future of Voice Input Modalities

00:09:32 - The Citrini "Doom Post": The viral Substack predicting 10% unemployment and an S&P crash

00:11:52 - Quadratic - Bringing the productivity boost of AI into your spreadsheets. Visit https://www.quadratic.ai/twist to sign up and use the code TWIST to get one free month of their pro tier subscription.

00:14:28 - Debunking the Doom: Cognitive Surplus, the Lump of Labor Fallacy & Why People Adapt

00:22:08 Notion ****- Notion brings all your notes, docs, and projects into one connected space that just works with AI built right in. Try Notion, with Notion Agent, at https://www.notion.com/twist

00:26:18 - Open Source vs. Closed Models: How DeepSeek Forced Innovation & Who Wins the Cost Race

00:38:05 - Building AI-First Teams: Certifications, Agent Management & the New Skill Stack

00:43:42 - Customer Support Reinvented: How Wispr Went from 200 Agents to 4

01:16:29 - Breaking News: Apple Onshores Mac Mini to Houston & The Taiwan Chip Risk Explained

Subscribe to This Week in AI on Apple: https://podcasts.apple.com/us/podcast/this-week-in-startups/id315114957

Subscribe to This Week in AI on Spotify:

https://open.spotify.com/show/6cPSOft5WRm3zJq7gchacA?si=3b86d10c10434b21

Thanks for watching!

🤖 If you want to stay ahead of the curve on all things AI, make sure to join our community across all platforms:

📩 Get the Weekly Newsletter: https://thisweekinai.ai/

📺 Subscribe on YouTube: www.youtube.com/@ThisWeekinAIPodcast

📸 Instagram: www.instagram.com/thisweekinaipodcast

📱 TikTok: www.tiktok.com/@thisweekinaipodcast✖️X: https://x.com/ThisWeeknAI

Follow Jason:

X: https://twitter.com/Jason

LinkedIn: https://www.linkedin.com/in/jasoncalacanis

Follow Oliver:

https://x.com/oliverkorzen

Check out all our partner offers: https://partners.launch.co/

This week we sit down with three founders building at the frontier of AI in space, autonomous hardware, and software reliability: Philip Johnston (StarCloud), Boris Sofman (Bedrock Robotics), and Spiros Xanthos (Resolve AI). We dig into why data centers are heading to orbit, how AI is taking the wheel on construction sites, and whether the industry is moving too fast for its own good.

We explore how physical infrastructure, real-world autonomy, and AI trust are reshaping the industry from the ground up.

• Sam Altman vs. Space Data Centers: He said it won't matter "this decade." But Philip breaks down why the economics of space compute are about to flip.
• The Construction Labor Crisis: Half the skilled workforce is retiring in the next seven years while data center construction spend hits $700B this year alone. Boris explains why autonomous excavators aren't replacing workers, they're the only way to keep up.
• AI Broke Amazon's Code: Generative AI is shipping software faster than engineers can understand it. Spiros warns that accelerating code velocity without upgrading reliability is a disaster waiting to happen.
• Automation Bias & Skill Degradation: What aviation taught us about over-trusting autopilot, and what developers need to learn fast.
• Why Americans Don't Trust AI: A KPMG survey puts AI's favorability between ICE and Iran. The panel debates whether healthcare, education, and construction are the three wins that could turn the tide.
• Karpathy's AI Research Agent: A weekend project that fires off entire teams of AI grad students running experiments overnight. Is this what automating the scientific method actually looks like?
• Anthropic vs. the Pentagon: The U.S. government threatened to label a leading domestic AI company a supply chain risk. Who gets to decide how AI is used in warfare?

Timestamps:

00:00 — Welcome to This Week in AI: Episode 4

01:22 — Philip Johnston (Star Cloud): Why data centers in space will dominate

03:35 — Boris Sofman (Bedrock Robotics): Autonomizing construction equipment with AI

08:44 — Spiros Santos (Resolve AI): Intro & AI-generated code's "high blast radius" problem at Amazon

09:31 — Squarespace: Turn your idea into a beautiful website! Go to http://www.squarespace.com/twist for a free trial. When you’re ready to launch, use offer code TWIST to save 10% off your first purchase of a website or domain.

10:35 — Deep dive: Amazon's code red meetings & trusting AI-generated code too much

13:20 — Automation bias & skill degradation: the aviation autopilot analogy

17:09 — Why Americans distrust AI: and what the industry should do about it

20:05 — Notion ad

21:29 — KPMG survey results: only 9% of large companies plan job cuts from AI, 55% plan to hire more; panel debates job displacement fears

31:43 — AI's PR fix: healthcare, education & construction as the big three wins

39:22 — World models & humanoid robots: LeCun's $1B bet & the Figure video controversy

47:59 — Star Cloud's 88,000-satellite constellation: economics of space compute

56:05 — Andrej Karpathy's AI research agent: automating the scientific method

1:01:04 — Anthropic vs. the military: supply chain risk, Dario's dilemma & the Iran strike claim

1:07:50 — Hiring pitches & outro: Star Cloud, Bedrock Robotics, Resolve AI

Subscribe to This Week in AI on Apple: https://thisweekinai.ai/spotify

Subscribe to This Week in AI on Spotify: https://thisweekinai.ai/apple

🤖 If you want to stay ahead of the curve on all things AI, make sure to join our community across all platforms:

📩 Get the Weekly Newsletter: https://thisweekinai.ai/

📺 Subscribe on YouTube: https://www.youtube.com/@ThisWeekinAIPodcast

📸 Instagram: https://www.instagram.com/thisweekinaipodcast

📱 TikTok: https://www.tiktok.com/@thisweekinaipodcast

✖️X: https://x.com/ThisWeeknAI

Follow Jason:

X: https://twitter.com/Jason

LinkedIn: https://www.linkedin.com/in/jasoncalacanis

Follow Oliver:

https://x.com/oliverkorzen

Check out all our partner offers: https://partners.launch.co/

This week, JCal sat down with Shiv Rao at LAUNCH Fest 2026. Shiv is a cardiologist, CEO, and co-founder of Abridge, a $5B+ company that's raised close to $1B to build vertical large language models for healthcare. Shiv breaks down why the US healthcare system is broken, how AI is already outperforming doctors on average, and what it took to build a unicorn in one of the most regulated industries on the planet.

We explore how AI is transforming doctor-patient conversations, eliminating clinician burnout, and reshaping how care is delivered and paid for.

• Doctors Need 30 Hours a Day: An American Journal of General Internal Medicine study found physicians can't complete their daily workload in 24 hours. Shiv explains where all that time goes anAbridge is clawing it back.
• AI Has Better Bedside Manner Than Your Doctor: Research from Harvard and other institutions now shows AI delivers more consistent answers and patients actually prefer it. Shiv explains why that's not as scary as it sounds.
• The Healthcare Stakeholder Problem: Shiv invokes Conway's Law to explain why every player in the system, doctors, insurers & hospitals are pointing guns at each other instead of aligning around the patient.
• A Real-Time AI Copilot for Doctors: Abridge's live assistant listens to conversations, pulls context from years of patient data, and prompts doctors with follow-up questions mid-visit to unlock insurance approvals on the spot.
• The Self-Directed Patient Explosion: Nearly every patient walks in armed with ChatGPT research. Shiv explains why this is compounding the burnout crisis, not solving it.
• Go-To-Market in Regulated Industries: Why "start small and swim upstream" can be a death trap in healthcare, and how Abridge's decision to target large health systems early became a core part of their moat.
• Being Early Isn't Being Wrong: Abridge started three months after "Attention Is All You Need" and raised a $3M seed on a $12M pre. Shiv shares what it was like waiting years for the market to catch up.
• Robotics in Healthcare: From AI-powered suturing at Johns Hopkins to the case for humanoid EMTs, Shiv lays out what's coming and what's still far off.

🔗 Learn more about Abridge: https://www.abridge.com

This Week In AI is made possible by:

Quadratic - https://www.Quadratic.ai/twist

Timestamps:

00:00 Welcome & intro to Shiv Rao, CEO & co-founder of Abridge

01:20 Rural hospital closures & the necessity of AI agents

03:33 Designing an AI intake process for primary care

06:01 Solving the 30-hour workday & clerical burnout

08:25 Conway’s Law & misaligned stakeholders in healthcare

10:07 Quadratic - Bringing the productivity boost of AI into your spreadsheets. Visit https://www.quadratic.ai/twist to sign up and use the code TWIST to get one free month of their pro tier subscription.

11:50 Why patients prefer AI models over average clinicians

14:27 Moving past the charade of the all-knowing doctor

17:01 How automated notes restore human clinical presence

21:00 The Joe Rogan crowd & consumer-driven healthcare

27:29 Advances in surgical robotics & AI precision

32:44 Surviving the Bert era to reach the AI inflection point

Subscribe to This Week in AI on Apple: https://thisweekinai.ai/spotify

Subscribe to This Week in AI on Spotify: https://thisweekinai.ai/apple

Thanks for watching!

🤖 If you want to stay ahead of the curve on all things AI, make sure to join our community across all platforms:

📩 Get the Weekly Newsletter: https://thisweekinai.ai/

📺 Subscribe on YouTube: / @thisweekinaipodcast

📸 Instagram: / thisweekinaipodcast

📱 TikTok: / thisweekinaipodcast

✖️X: https://x.com/ThisWeeknAI

Follow Jason:

X: / jason

LinkedIn: / jasoncalacanis

Follow Oliver:

https://x.com/oliverkorzen

Thank you to our partner:

Quadratic - Bringing the productivity boost of AI into your spreadsheets. Visit https://www.quadratic.ai/twist to sign up and use the code TWIST to get one free month of their pro tier subscription.

This week we sat down with Aravind Srinivas and Edwin Chen. Aravind is the co-founder and CEO of Perplexity, whose revenue has grown from $100M to $500M on the back of Perplexity Computer. Edwin is the co-founder and CEO of Surge AI, the data training company teaching frontier models how to think, which quietly bootstrapped past $1B in revenue without ever raising a dollar.

This Week In AI is made possible by:

PayPal Open - One Platform for all Business: https://paypalopen.com/

Timestamps:

00:00 Cold open

01:00 Welcome & intro to Aravind Srinivas and Edwin Chen

02:40 Why Perplexity Comet and Perplexity Max are taking off

05:25 Edwin on Surge AI and why "data labeling" is the wrong term

10:47 Tim Cook steps down — what Apple's new CEO should do

13:56 Owning your agent loops: why Apple wins

21:20 "The iPhone is not getting disrupted by AI at all"

23:09 $242B raised in Q1 2026 and the late-stage capital flood

23:55 Edwin on bootstrapping Surge past $1B without raising

25:09 The ChatGPT "one weird trick" story and clickbait models

30:58 Claude Code as a loss leader to dominate token collection

33:30 Are we in the endgame for coding?

35:35 Autocomplete to auto-diff to auto-outcomes

38:54 The death of the no-code movement

41:34 30% headcount growth, 5x revenue, the efficiency playbook

45:41 AI in Hollywood, Gal Gadot, and $70M movies that should cost $200M

50:29 "People don't buy models, they buy products"

57:20 Specialization vs commoditization — what actually accrues value58:00 "LM Arena is a cancer on AI"

63:34 Perplexity's heuristics for measuring user intent65:41 Model Council, Jensen Huang, and orchestrating frontier models

71:23 Most impressive AI experiences — WhisperFlow, Grok on X, Claude Design, and WHOOP

78:29 Hiring at Surge AI and Perplexity

Subscribe to This Week in AI on Apple:https://thisweekinai.ai/apple

Subscribe to This Week in AI on Spotify:https://thisweekinai.ai/spotify

Follow Jason:

X: @jason

LinkedIn: /jasoncalacanis

Check out all our partner offers:

https://partners.launch.co/

Links Mentioned on the Show:

Perplexity: https://perplexity.ai/

Surge AI: https://surge.ai/

Perplexity Comet: https://comet.perplexity.ai/

Perplexity Max: https://perplexity.ai/max

Claude Code: https://claude.com/product/claude-code

Claude Design: https://claude.com/claude-design

OpenAI Codex: https://openai.com/codex

GitHub Copilot: https://github.com/features/copilot

Cursor: https://cursor.com/

LM Arena: https://lmarena.ai/

Kimi K2: https://www.kimi.com/

DeepSeek: https://www.deepseek.com/

Apple Silicon: https://www.apple.com/mac/m4/

WhisperFlow: https://wisprflow.ai/

WHOOP: https://www.whoop.com/

This week we sit down with Lin Qiao and Demi Guo on This Week in AI Episode 9. Lin is the co-founder and CEO of Fireworks AI, a frontier inference platform processing tens of trillions of tokens per day, built by seven ex-Meta engineers who created PyTorch. Demi is the co-founder and CEO of Pika, building humanized AI agents for creative work, agents you interact with like a person, not a prompt box.

This Week In AI is made possible by:

PayPalOpen - One Platform for all Business: https://paypalopen.com/

Timestamps:

00:00 Welcome & intro to Lin Qiao and Demi Guo

02:38 Lin's journey from Meta to Fireworks AI

05:13 Building PyTorch and bootstrapping AI infrastructure

06:58 Who Fireworks competes with and why enterprises need it

08:43 Activating the 95% of private data locked in enterprises

11:31 Demi's journey building Pika and the pivot to humanized agents

16:03 The best interface for creation is a human-like agent

16:42 The AI layoff trap — a prisoner's dilemma for firms

19:00 Cambrian explosion of startups and the hobbyist-to-inventor pipeline

22:07 Flattening organizations and the death of middle management

28:03 Taste, judgment, and why "slop" is the real risk

31:22 Why agents drift and the case for constant iteration

33:40 Rethinking agents: not tools, but children you raise

39:36 How close are open-source models to frontier?

43:20 Token usage and the economics of running agents

46:15 Toys becoming tools — the hobbyist signal

48:59 Public perception of AI: America vs. China vs. Silicon Valley

53:04 Agents as self-expression and identity creation

57:06 The expert vs. public perception gap on AI's impact

63:17 Meta's Muse model and the open-source debate

68:26 Data drought, synthetic data, and the next architecture leap

72:06 Hiring at Fireworks AI and Pika

Subscribe to This Week in AI on Apple: https://thisweekinai.ai/spotify

Subscribe to This Week in AI on Spotify: https://thisweekinai.ai/apple

Follow Jason:

X: https://twitter.com/Jason

LinkedIn: https://www.linkedin.com/in/jasoncalacanis

Follow Oliver:

https://x.com/oliverkorzen

Check out all our partner offers: https://partners.launch.co/

Links Mentioned on the Show:

Fireworks AI: https://fireworks.ai/

Pika: https://pika.art/

PyTorch: https://pytorch.org/

"The AI Layoff Trap" paper (UPenn & Boston University): https://arxiv.org/abs/2603.20617

Fortune: CFOs say AI cuts will be 9x bigger than reported: https://fortune.com/2026/03/24/cfo-survey-ai-job-cuts-productivity-paradox-2026/

HBR: Companies firing based on AI potential, not performance: https://hbr.org/2026/01/companies-are-laying-off-workers-because-of-ais-potential-not-its-performance

New Yorker profile on Sam Altman (Ronan Farrow): https://techcrunch.com/2026/04/11/sam-altman-responds-to-incendiary-new-yorker-article-after-attack-on-his-home/

Sam Altman home attack: https://www.cnbc.com/2026/04/10/sam-altman-house-hit-with-molotov-cocktail-openai-office-threatened.html

Block layoffs — Jack Dorsey cites AI: https://fortune.com/2026/02/27/block-jack-dorsey-ceo-xyz-stock-square-4000-ai-layoffs/

Pew Research — AI experts vs. public perception: https://www.pewresearch.org/internet/2025/04/03/how-the-us-public-and-ai-experts-view-artificial-intelligence/

Meta Muse Spark announcement: https://ai.meta.com/blog/introducing-muse-spark-msl/

DeepSeek V3: https://github.com/deepseek-ai/DeepSeek-V3

Bubble Pal AI toy (Shenzhen): https://www.prnewswire.com/news-releases/hugging-every-fun-thought-haivivi-unveils-the-worlds-first-aigc-toy-bubblepal-302209714.html

Cursor (Composer): https://cursor.sh/

NVIDIA Nemotron: https://developer.nvidia.com/nemotron

Qwen (Alibaba): https://qwenlm.github.io/

Mistral AI: https://mistral.ai/

This week on TWiAI, Jason sits down with three founders at the center of the AI stack: Kanjun Qiu (CEO of Imbue), Carina Hong (CEO of Axiom), and Jonathan Siddharth (CEO of Turing). They break down Anthropic's explosive $30B run rate, why it just overtook OpenAI in revenue, Meta's bizarre internal token-burning leaderboard, and what happens when every person on Earth has 100 AI agents running for them.

• Anthropic's Revenue Explosion: Anthropic hit a $30B run rate, up from $9B just six months ago, and appears to have overtaken OpenAI in token sales. The panel breaks down where the money is coming from and why Meta might be the biggest customer.
• Meta's Token-Burning Leaderboard: Reports surfaced of an internal Meta contest rewarding teams for burning the most tokens, with employees building bots that loop just to rack up usage. Is it brilliant adoption strategy or pure waste?
• Open Source Agents vs. Lock-In: Kanjun argues that handing your entire business, memories, and workflows to closed AI platforms is a recipe for lock-in. She's building open agent infrastructure so users can swap models freely and own their data.
• OpenAI's Identity Crisis: The panel dissects OpenAI's dropped Disney deal, its $100B+ raise, internal CFO tensions, and whether the company is doing too many side quests while Anthropic eats its lunch.
• The Commodification of Developers: A year ago, the debate was whether AI makes developers 5% or 15% more efficient. Now the question is whether everyone is a developer, and what that means for the industry.
• Why Open Source Must Win: From Apple Silicon to local models, the case for owning your AI stack and why Jason is going all-in on open source hardware and software.

Learn more about Imbue: https://imbue.com

Learn more about Axiom: https://axiommath.ai

Learn more about Turing: https://www.turing.com

This Week In AI is made possible by:

PayPalOpen - One Platform for all Business: https://paypalopen.com/

Timestamps:

00:00 Cold open

01:03 Welcome and intro to this week's panel

01:31 Kanjun Qiu on open source agents and Imbue

05:10 Carina Hong on building an AI mathematician at Axiom

06:30 Formal verification and why superintelligence needs proof

09:52 Jonathan Siddharth on Turing's superintelligence accelerator

15:53 Anthropic's $30B run rate and overtaking OpenAI

22:27 OpenAI's strategy, Disney deal, and $100B raise

28:27 Meta's internal token-burning leaderboard

44:08 The commodification of developers

50:31 God mode for code: formal verification in practice

58:27 Superintelligence in 36 months: what happens next

01:08:41 Apple, Siri, and why Big Tech is failing at AI products

Subscribe to This Week in AI on Apple: https://thisweekinai.ai/spotify

Subscribe to This Week in AI on Spotify: https://thisweekinai.ai/apple

Thanks for watching!

🤖 If you want to stay ahead of the curve on all things AI, make sure to join our community across all platforms:

📩 Get the Weekly Newsletter: https://thisweekinai.ai/

📺 Subscribe on YouTube: https://www.youtube.com/channel/UCZo46NiqoWMnuJo1SxCnX0g

📸 Instagram: https://www.instagram.com/thisweekinaipodcast

📱 TikTok: https://www.tiktok.com/@thisweekinaipodcast

✖️X: https://x.com/ThisWeeknAI

Follow Jason:

https://x.com/jason

LinkedIn: https://www.linkedin.com/in/jasoncalacanis/

Follow Oliver:

https://x.com/oliverkorzen

Check out all our partner offers: https://partners.launch.co/

This Week in AI, JCal sits down with three CEOs building the infrastructure, intelligence, and interfaces for the next era of AI: Jeremy Fraenkel (CEO, Fundamental), Victor Riparbelli (CEO, Synthesia), and Nick Harris (CEO, Lightmatter). We break down what's actually happening beneath the AI hype: the data modality LLMs completely missed, why copper is the real bottleneck in AI data centers, OpenAI shutting down Sora, the build vs. buy debate for AI tools, and how close we really are to AGI.

• AI's Biggest Blind Spot, Tabular Data: LLMs transformed text, images, and code, but 70-80% of enterprise data lives in rows and columns.
• Copper Can't Keep Up: Nick explains why AI data centers are hitting a wall. GPUs compute faster than they can communicate. Lightmatter's photonic chips push 1.6 terabits per fiber and can 3x training speed.
• Why OpenAI Killed Sora & Anthropic's Focus is Winning: Victor breaks down why even OpenAI had to learn the lesson of focus, and why Claude Code has every founder talking.
• Vibe Coding Your Own CRM vs. Buying Salesforce: Jeremy reveals Fundamental built their own internal CRM using vibe coding. The panel debates when building beats buying and when it's a distraction.
• The Omnipresent CEO: Jason shares how he's using AI agents for root access to Slack, Gmail, and Notion, resurrecting former employees as AI personas, automating SDR workflows, and summarizing employee inboxes while they're on vacation.
• Are We Already at AGI?: Nick says the rate of progress is a double exponential. Jeremy argues AGI is a moving goalpost. Victor warns of "Future Shock" and societal disruption.

🔗 Learn more about Fundamental: https://fundamental.tech🔗 Learn more about Synthesia: https://www.synthesia.io🔗 Learn more about Lightmatter: https://lightmatter.coThis Week In AI is made possible by:PayPal Open - One Platform for all Business: paypalopen.comTimestamps:00:00 Welcome & intro to Jeremy Fraenkel, Victor Riparbelli, and Nick Harris01:47 What is Fundamental? Large tabular models explained07:01 Victor Riparbelli on Synthesia & why OpenAI killed Sora11:09 Claude Code dominance & the Lightspeed founder retreat12:08 Nick Harris on Lightmatter, photonics & the new Moore's Law14:38 Copper vs. fiber: why AI data centers are hitting a wall18:44 Reinventing video: interactive, real-time, personalized21:32 The economics of a custom AI movie23:55 Why Amazon, Google & Meta are building their own chips28:06 Tables have a bandwidth problem too32:27 When will compute be as cheap as storage?36:10 The future of software: every company gets a custom stack38:06 Vibe coding your own CRM vs. buying Salesforce45:57 Jason's quest for root access to Slack50:18 The omnipresent CEO: Doctor Manhattan meets Jesus CEO52:13 Resurrecting former employees as AI personas53:25 Victor's executive changelog for a 650-person company55:07 Whisper Flow & the Plaud Pin1:00:03 AGI: is it already here?1:03:37 Jeremy: we've only solved half the brain1:06:30 70% of Americans fear AI will impact jobs1:08:47 Future Shock & keeping the rope tight*Mentioned in the show:*

Wisper Flow: https://wisperflow.ai

• Plaud Pin: https://www.plaud.ai
• Athena Executive Assistants: https://www.athenawow.com
• WHOOP: https://www.whoop.com
• "Future Shock" by Alvin Toffler: https://www.amazon.com/Future-ShockAlvin-Toffler/dp/0394425863
• Victor on TWiST, E1776: https://youtu.be/jxET4fq_2eA
• Nick on TWiST, E1787: https://youtu.be/FPW2nnEqfMsSubscribe to This Week in AI on Apple: https://thisweekinai.ai/spotifySubscribe to This Week in AI on Spotify: https://thisweekinai.ai/appleThanks for watching!🤖 If you want to stay ahead of the curve on all things AI, make sure to join our community across all platforms:📩 Get the Weekly Newsletter: https://thisweekinai.ai/📺 Subscribe on YouTube: https://www.youtube.com/@ThisWeekinAIPodcast📸 Instagram: https://www.instagram.com/thisweekinaipodcast

This week Jason sat down with Jake Loosararian and Chris Lattner on Episode 6 of This Week in AI. Jake is the CEO and co-founder of Gecko Robotics, a company deploying purpose-built robots and AI for mission-critical infrastructure inspection across energy, defense, and manufacturing. Chris is the CEO and co-founder of Modular, building a universal software layer that lets developers run AI models across Nvidia, AMD, and Apple silicon without being locked into any single hardware vendor.

We explore the GPU shortage, why China's chip smuggling reveals the stakes of the AI cold war, how purpose-built robotics are beating humanoids on ROI, the case for American reindustrialization, and why the next decade could be the best ever for private equity in capital-intensive industries.

• Purpose-Built Robots vs. Humanoids: Jake has been building mission-critical robots for 13 years. He explains why general-purpose humanoids still have too little ROI for industrial use, and why specialized robots that find and fix problems are winning in the field.
• The GPU Shortage Is Real: Chris breaks down why you can't just go buy 100 Blackwell chips today, why Nvidia's Cuda creates massive lock-in, and how Modular is building a unified software layer across all major chip architectures.
• Google TPUs Are the Sleeper: Chris ranks Google as the number one threat to Nvidia's dominance, ahead of Amazon's Trainium and AMD.
• China's Chip Smuggling & the AI Cold War: A Supermicro co-founder allegedly smuggled $2.5B in Nvidia chips to China using fake serial numbers and a hairdryer.
• The Best Decade for Private Equity: Jake makes the case that capital-intensive, commoditized infrastructure assets: waste-to-energy, water treatment, old power plants will all generate incredible returns.
• Self-Driving State of Play: Chris, a former Tesla Autopilot lead, gives his read on Waymo's lead, Tesla's small Austin pilot, and why the real signal is when Tesla starts filing for fully autonomous permits in California.

• Figure's New AI Lab, Hark: Breaking news mid-episode: Brett Adcock announces a new personal intelligence lab.

Learn more about Gecko Robotics: https://www.geckorobotics.com

Learn more about Modular: https://www.modular.com/

This Week In AI is made possible by:

*PayPalOpen* - One Platform for all Business: paypalopen.com

*Timestamps:*

00:00 Welcome & intro to Jake Lu (Gecko Robotics) and Chris Lattner (Modular)

01:34 Gecko's 13-year journey & the Cantilever platform

05:15 Chris Lattner on Modular: replacing Cuda & unifying AI hardware

11:10 Nvidia lock-in, AMD's Rock & why the software stack is broken

19:49 The GPU shortage: how real is it?

22:13 Who challenges Nvidia? Google TPUs, Amazon Trainium & AMD ranked

28:17 China chip smuggling: $2.5B in Nvidia GPUs & the AI cold war

37:43 Self-driving update: Waymo, Tesla's Austin pilot & Chris's Tesla history

42:20 Figure's humanoid package sorting — real or demo magic?

43:47 The best decade for private equity in capital-intensive assets

51:04 Reindustrialization, the trades boom & making manufacturing cool

58:39 Building tech companies outside Silicon Valley

1:06:46 Breaking news: Brett Adcock launches Hark from Figure

1:10:15 Closing thoughts: grit over hype, customers over valuations

Subscribe to This Week in AI on Apple: https://thisweekinai.ai/spotify

Subscribe to This Week in AI on Spotify: https://thisweekinai.ai/apple

Thanks for watching!

🤖 If you want to stay ahead of the curve on all things AI, make sure to join our community across all platforms:

📩 Get the Weekly Newsletter: https://thisweekinai.ai/

📺 Subscribe on YouTube: https://www.youtube.com/@ThisWeekinAIPodcast

📸 Instagram: https://www.instagram.com/thisweekinaipodcast

📱 TikTok: https://www.tiktok.com/@thisweekinaipodcast

✖️X: https://x.com/ThisWeeknAI

Follow Jason:

X: https://twitter.com/Jason

Follow Oliver:

https://x.com/oliverkorzen

Check out all our partner offers: https://partners.launch.co/

This week, JCal sits down with two builders working on opposite ends of the AI stack: Naveen Rao, co-founder and CEO of Unconventional AI (his $4.5B startup rethinking the computer from first principles), and Trey Holterman, co-founder and CEO of Tennr, the platform automating the messy paperwork that decides whether millions of patients ever get the care they need.

They break down why AI already beats ER doctors at diagnosis, why Epic is the most hated company in healthcare, what it takes to disrupt healthcare's $5 trillion incumbents, and whether $1.1 trillion in hyperscaler CapEx will pay off before the energy runs out.

Mentioned in the show:

• Unconventional AI: https://unconv.ai
• Tennr: https://tennr.com
• Abridge: https://abridge.com
• Ambience Healthcare: https://ambiencehealthcare.com
• Epic: https://epic.com
• Plaud (NotePin): https://plaud.ai
• Whoop: https://whoop.com
• Function Health: https://functionhealth.com
• Oura: https://ouraring.com
• Eight Sleep: https://eightsleep.com
• Perplexity: https://perplexity.ai
• Databricks: https://databricks.com
• Harvard / Beth Israel ER study: https://www.harvardmagazine.com/ai/ai-outperforms-doctors-diagnosis-harvard-study
• Jack Clark on recursive self-improvement: https://jack-clark.net

This Week In AI is made possible by:PayPal Open - One Platform for all Business: https://paypalopen.com

Timestamps:

00:00 Cold open

00:51 Welcome & intro to Naveen Rao (Unconventional AI) and Trey Holterman (Tennr)

05:43 Tennr's mission: fixing the U.S. healthcare "black hole"

06:23 Why primary care is changing in the AI era

09:09 The radiologist that wasn't replaced & what AI got right in 2015

15:31 Harvard study: AI beats ER doctors 67% to 55%

20:16 Plaud Pin & the always-on AI in the exam room

22:29 AI scribes, EMR privacy & how Abridge actually works

24:01 Whoop, Aura & when wearable data finally meets the clinic

25:20 Epic, FHIR & the 800-pound gorilla of healthcare

34:21 Disrupting Epic with an open-source EMR play

37:20 The Abridge vs Ambience scribe wars

42:06 Jack Clark's 60% odds on recursive self-improvement by 2028

53:41 Morgan Stanley's $1.1 trillion hyperscaler CapEx forecast

1:04:53 Anthropic's marketing machine & the AI religion thesis

Subscribe to This Week in AI on Apple: https://thisweekinai.ai/spotify

Subscribe to This Week in AI on Spotify: https://thisweekinai.ai/apple

Thanks for watching!

🤖 If you want to stay ahead of the curve on all things AI, make sure to join our community across all platforms:

📩 Get the Weekly Newsletter: https://thisweekinai.ai/

📺 Subscribe on YouTube: https://www.youtube.com/@ThisWeekinAIPodcast

📸 Instagram: https://www.instagram.com/thisweekinaipodcast

📱 TikTok: https://www.tiktok.com/@thisweekinaipodcast

✖️X: https://x.com/ThisWeeknAI

Follow Jason:

X: https://twitter.com/Jason

LinkedIn: https://www.linkedin.com/in/jasoncalacanis

Follow Oliver:

https://x.com/oliverkorzen

Check out all our partner offers: https://partners.launch.co/

This week Jason sits down with three founders at the frontier of AI infrastructure, software development, and vertical AI: Matan Grinberg, co-founder and CEO of Factory; Russ d'Sa, co-founder and CEO of LiveKit; and George Sivulka, founder and CEO of Hebbia.

They break down why AI coding agents are more powerful than vibe coding, how voice became the default interface for AI, why LLMs are becoming commoditized like RAM, and what it actually takes to build a moat in 2026.

Mentioned in the show:

• Factory: https://factory.ai
• LiveKit: https://livekit.io
• Hebbia: https://hebbia.com
• Cursor: https://cursor.com
• DeepSeek V4: https://deepseek.com
• SpaceX / xAI: https://x.ai
• Polymarket (Cursor acquisition odds): https://polymarket.com/event/will-spacex-acquire-cursor
• LM Arena (model benchmarks): https://lmarena.ai
• SWE-bench (coding benchmarks): https://swebench.com
• Silicon Valley HBO clip (Son of Anton): https://www.youtube.com/watch?v=m0b_D2JgZgY

This Week In AI is made possible by:

PayPal Open - One Platform for all Business: https/:/paypalopen.com/

Timestamps:

00:00 Intro & AGI debate: where are we really?

02:26 Meet the guests: Matan Grinberg, Russ d'Sa, George Sivulka

03:30 Factory's mission: bringing autonomy to software engineering

04:29 LiveKit's origin: open source WebRTC to ChatGPT voice backbone

07:40 Learned helplessness and the raptor fence: a founder story

10:31 Hebbia: financial superintelligence for capital markets

13:21 SpaceX acquiring Cursor for $60B: the deal breakdown

17:28 Why enterprises can't standardize on one model provider

21:50 The Silicon Valley clip: when AI deletes all the software

23:09 How AI coding tools have changed internal dev workflows

26:28 Moats in the age of vibe coding: what actually protects you

30:12 The relentless iteration promise as a company's core moat

33:40 Slack example: the buy vs. build calculation in 2026

38:10 Deterministic agents and encoding institutional expertise

41:17 LLMs as commoditized infrastructure: where value pools

45:56 DeepSeek V4 drops mid-episode: $3.48 vs. Claude's $25

50:13 AI in financial markets: arbitrage, meme stocks, and ASI

56:58 China blocks Metas acquisition, OpenAI-Microsoft deal redux

58:41 AGI definitions and who wins the US vs. China model race

01:02:13 US open source embarrassment and the talent war

01:05:23 OpenAI's spend vs. revenue: risk of ruin or name of the game?

01:09:14 Apple's new CEO, M-series compute flood, and space data centers

01:12:08 P-doom scores and final thoughts

Subscribe to This Week in AI on Apple: https://thisweekinai.ai/apple

Subscribe to This Week in AI on Spotify: https://thisweekinai.ai/spotify

Follow Jason:

X: https://twitter.com/Jason

LinkedIn: https://www.linkedin.com/in/jasoncalacanis

Follow Oliver:

https://x.com/oliverkorzen

Check out all our partner offers: https://partners.launch.co/

The frontier labs are coming for the application layer, and they're going to steal your idea. We dug into Andrej Karpathy's move to Anthropic, the new rules of AI etiquette, and why graduates are booing AI at commencement speeches.

This week's roundtable: Kanjun Qiu (CEO of Imbue, building open source agents), Jeremy Fraenkel (CEO of Fundamental, large tabular models for enterprise data), and Karri Saarinen (CEO of Linear, the product development system used by OpenAI, Coinbase, Cursor, Ramp, and Cash App).

Thank you to our exclusive sponsor:

PayPal Open, One Platform for All Business: http://paypalopen.com/

Timestamps:

0:00 Cold open

1:02 Welcome to Episode 14 of This Week in AI

1:45 Kanjun Qiu on Imbue's GPU cluster bet that funds the company

5:38 Jeremy Fraenkel on why LLMs choke on structured enterprise data

9:36 Karri Saarinen on Linear and product work for AI agents

12:50 Design in the age of AI: why early-stage product design is getting worse

21:18 Jason coins "AI etiquette": your output is your responsibility

33:30 Andre Karpathy joins Anthropic

40:00 The cult of each AI lab: SpaceX, Anthropic, OpenAI

47:00 Anthropic's pricing misstep and the "punk software" movement

50:00 The verticalization war coming for the application layer

55:00 "Claude knows more about me than my partner": power concentration in AI

58:30 Mac M5 clusters and confidential compute on AWS

1:07:30 Graduates boo AI at commencement: Schmidt, Borchetta, Cawfield

1:22:00 Four founders deliver their own commencement speeches

1:28:30 Reese Witherspoon vs the New York Times

1:33:00 Plugs and where to find the guests

🔗 Guests on X:

Kanjun Qiu, Imbue: https://imbue.com | https://x.com/kanjun

Jeremy Fraenkel, Fundamental: https://fundamental.tech | https://x.com/fraenkelj

Karri Saarinen, Linear: https://linear.app | https://x.com/karrisaarinen

🔗 Referenced in this episode:

Andre Karpathy joins Anthropic (announcement): https://x.com/karpathy/status/2056753169888334312?s=20

Imbue Blueprint (open source agent skill): https://imbue.com/blueprint

ExoLabs (daisy-chained Mac clusters): https://exolabs.net

Abacus (on-prem language models): https://goabacus.co

Whisper Flow (voice-to-text): https://wisprflow.ai

Perplexity Model Council: https://perplexity.ai

NYT editorial on Reese Witherspoon and AI: https://www.nytimes.com/2026/05/15/opinion/reese-witherspoon-mel-robbins-girlboss-ai.html

🔗 Subscribe and follow:

Newsletter and all platforms: https://thisweekinai.ai

#ThisWeekInAI #AI #Imbue #Fundamental #Linear #Anthropic #OpenAI #Karpathy #AIEtiquette #VibeCoding #PunkSoftware #AICommencement #OpenSourceAI

The future of AI isn't a smarter chatbot. It's a model that watches your screen, listens to the room, and acts on what it sees. We dug into Thinking Machines' new interaction model, what it means for compute, and the layoff wave that's already here.This week's roundtable: Anastasios Angelopoulos (CEO of Arena, formerly LMArena), Nick Harris (CEO of Lightmatter, photonic computing chips), and Philip Johnston (CEO of StarCloud, building megawatt data centers in space).

Thank you to our exclusive sponsor:

PayPal Open, One Platform for All Business: http://paypalopen.com/

Timestamps:

0:00 Cold open

1:21 Welcome to Episode 13

2:51 Is China closing the AI gap? Arena's data

5:16 Lightmatter and the photonic interconnect bottleneck

9:42 StarCloud 2, Nvidia Space Ruben 1, and orbital data centers

17:24 Thinking Machines' interaction model: what's actually new

28:22 Whisper Flow and the 3-pedal desk setup

33:48 Real-time desktop and camera awareness as the real unlock

40:25 Why this 100x's compute demand

42:43 The polarization of compute and $10M personal data centers

49:25 The layoff wave: Cloudflare, PayPal, Coinbase, Upwork

54:48 The 10x gap between AI-first and non-AI-first employees

59:52 Unlimited agency and the abundance future

1:00:46 Anthropic's Project Luna runs a retail store

1:03:45 Decoupling labor from value creation

1:05:03 P(doom) round

🔗 Guests:

Anastasios Angelopoulos, Arena: https://arena.ai | @ML_Angelopolous

Nick Harris, Lightmatter: https://lightmatter.co

Philip Johnston, StarCloud: https://starcloud.com | @philipjohnston

🔗 Referenced in this episode:

Thinking Machines, Introducing Interaction Models: https://thinkingmachines.ai/blog/interaction-models/

LMArena leaderboard: https://lmarena.ai

Lightmatter: https://lightmatter.co

Starcloud: https://www.starcloud.com

TechCrunch, Cloudflare says AI made 1,100 jobs obsolete: https://techcrunch.com/2026/05/08/cloudflare-says-ai-made-1100-jobs-obsolete-even-as-revenue-hit-a-record-high/

Fast Company, Tech layoffs this week due to AI (Cloudflare, PayPal, Coinbase, Upwork): https://www.fastcompany.com/91538995/tech-layoffs-due-to-ai-this-week-cloudflare-paypal-coinbase-upwork

Bloomberg, South Korea floats citizen dividend from AI profits: https://www.bloomberg.com/news/newsletters/2026-05-12/korea-s-massive-ai-boom-triggers-call-for-tech-tax-roiling-market

PYMNTS, Inside a retail store run entirely by AI (Andon Labs / Luna): https://www.pymnts.com/artificial-intelligence-2/2026/inside-a-retail-store-run-entirely-by-ai/

Wispr Flow (voice-to-text tool Jason uses): https://wisprflow.ai

Fermi Paradox: https://en.wikipedia.org/wiki/Fermi_paradox

🔗 Subscribe and follow:

Newsletter and all platforms: https://thisweekinai.ai

#ThisWeekInAI #AI #ThinkingMachines #Lightmatter #StarCloud #Arena #Anthropic #ProjectLuna #AIcompute #AIlayoffs #Superintelligence

The AI jobs panic is here. Meta cut 8,000. Intuit cut 3,000. CapEx went up. Nobody can agree if anyone has actually lost a job to AI, or just to the promise of it. We dug into the GPU squeeze, the new craft of "reward engineering," Pope Leo's call to disarm AI, and why Chinese open-source models just blew past American ones in token usage.

This week's roundtable: Erik Bernhardsson (CEO of Modal Labs, the serverless GPU cloud), Tanay Kothari (CEO of Wispr Flow, the voice dictation app every VC in the valley uses), and Richard Socher (CEO of Recursive Superintelligence and You.com).

Thank you to our exclusive sponsor:

PayPal Open, One Platform for All Business: http://paypalopen.com/

Timestamps:

0:00 Cold open

0:53 Welcome to Episode 15

5:53 Recursive's plan to build a self-improving "Eureka machine"

8:10 Token spend now exceeds headcount at the frontier

9:56 GPU crunch, Hopper prices, and the Anthropic-Colossus shockwave

12:23 Wispr Flow's 90% gross margin playbook

14:30 Running $100M in marketing with two humans and a swarm of agents

18:29 Reward hacking, paperclips, and the rise of the "reward engineer"

24:39 Why CEOs put one person in charge: multi-objective AI

28:17 Meta's 8,000 layoffs, $145B CapEx, Goldman vs. Stanford

33:23 "Nobody lost their job to AI, just the promise of AI"

37:57 Jevons paradox: software demand is infinite, illustrations aren't

39:48 Meta's keystroke monitoring and the back-channel reaction

42:55 Equity, Trump accounts, and rooting for your old employer

44:06 The Bloomberg/Indeed dev jobs chart

47:34 Jason's pitch: hiring a 22-year-old AI-native "software valet"

50:29 Pope Leo: "AI needs to be disarmed"

57:41 Chris Olah on AI displacement and the global poor

1:00:46 Chinese models hit 9 trillion tokens, DeepSeek V4 Flash goes #1

1:03:04 Strange biases, Tiananmen Square, and the US open-source vacuum

1:05:07 Who they're hiring

🔗 Guests:

Erik Bernhardsson, Modal Labs: https://modal.com | https://x.com/bernhardsson

Tanay Kothari, Wispr Flow: https://wisprflow.ai | https://x.com/tankots

Richard Socher, Recursive Superintelligence / You.com: https://recursive.com | https://x.com/RichardSocher

🔗 Referenced in this episode:

Modal Labs: https://modal.com

Wispr Flow: https://wisprflow.ai

Recursive Superintelligence launch: https://recursive.com

You.com: https://you.com

Anthropic / SpaceX Colossus 1 compute deal: https://www.anthropic.com/news/anthropic-spacex

TechCrunch on the $1.25B/month Anthropic-xAI compute deal: https://techcrunch.com/2026/05/20/anthropic-will-pay-xai-1-25-billion-per-month-for-compute/

Meta's 8,000 layoffs and 2026 AI CapEx of $145B: https://www.reuters.com

Intuit cuts 3,000 jobs to fund AI integration: https://www.reuters.com

Goldman Sachs CEO David Solomon NYT op-ed on AI job loss: https://www.nytimes.com

Stanford study on entry-level AI-exposed jobs (-16%): https://digitaleconomy.stanford.edu

Pope Leo XIV: "AI needs to be disarmed": https://www.vatican.va

Chris Olah (Anthropic) on AI and the global poor: https://www.anthropic.com

OpenRouter token usage leaderboard: https://openrouter.ai/rankings

DeepSeek V4 Flash: https://www.deepseek.com

Qwen 3 Max (Alibaba): https://qwenlm.ai

🔗 Subscribe and follow:

Newsletter and all platforms: https://thisweekinai.ai

#ThisWeekInAI #AI #ModalLabs #WisprFlow #Recursive #YouCom #DeepSeek #Qwen #MetaLayoffs #Anthropic #Colossus #PopeLeo #RewardHacking #AIJobs #OpenSource

Apple just paid a rival a billion dollars a year because it could not build Siri itself, and that tells you where the AI platform war actually stands: the edge has moved up the stack, from the model to the agent layer sitting on top of it.

This week's roundtable makes the case that whoever controls that layer controls the experience, and we brought in three founders building it from different angles: Hermes Agent, NotebookLM, and LiveKit. Their through-line is that the harness is temporary, the model keeps eating it, and the people building agents right now are designing for capabilities that do not exist yet.

This week's roundtable:

Jeffrey Quesnelle (Co-founder & CEO, Nous Research, the open source AI lab behind Hermes Agent)

Steven B Johnson (Editorial Director of NotebookLM & Google Labs, co-creator of NotebookLM)

Russ d'Sa (Co-founder & CEO, LiveKit, the open source real-time voice/video infrastructure behind ChatGPT voice mode)

Thank you to our exclusive sponsor: PayPal Open, One Platform for All Business: https://paypalopen.com

Timestamps:

0:00 Cold open

1:08 Welcome to Episode 17

1:40 Jeffrey Quesnelle on Hermes Agent and the open source agent race

5:03 Steven Johnson on NotebookLM, source-grounding, and going agentic

7:32 Why AI got booed at commencement: the first generation raised on ChatGPT

10:48 Russ d'Sa on LiveKit powering ChatGPT, Tesla, Grok, and Salesforce voice

27:50 Apple's $1B Gemini deal and the Siri UX problem

52:49 "Functional AGI, unevenly distributed": where the models actually are

57:24 The harness is temporary: why agent builders design for the next model

Guests:

Jeffrey Quesnelle, Nous Research: https://nousresearch.com | https://x.com/theemozilla

Steven Johnson, NotebookLM (Google Labs): https://notebooklm.google.com | https://x.com/stevenbjohnson

Russ d'Sa, LiveKit: https://livekit.io | https://x.com/dsa

Referenced in this episode:

Nous Research / Hermes Agent: https://nousresearch.com

NotebookLM: https://notebooklm.google.com

LiveKit: https://livekit.io

OpenAI ChatGPT: https://openai.com

Tesla: https://www.tesla.com

Grok (xAI): https://x.ai

Salesforce Agentforce: https://www.salesforce.com/agentforce

SAP Joule: https://www.sap.com/products/artificial-intelligence/ai-assistant.html

Apple Intelligence: https://www.apple.com/apple-intelligence

Google Gemini: https://gemini.google.com

Anthropic Claude Code: https://www.anthropic.com/claude-code

OpenRouter: https://openrouter.ai

Newsletter and all platforms: https://thisweekinai.ai

Apple Podcasts: https://thisweekinai.ai/apple

Spotify: https://thisweekinai.ai/spotify

YouTube: https://youtube.com/@ThisWeekinAIPodcast

Show X: https://x.com/ThisWeeknAI

#ThisWeekInAI #AI #NousResearch #NotebookLM #LiveKit #OpenSourceAI #AIAgents #Siri #Gemini #VoiceAI

The future of AI isn't about whether the model is smart enough. It's about whether we can afford to run it. We dug into the AI cost panic, the energy wall that's coming for compute, and why "developers are disappearing" gets the economics exactly backwards.This week's roundtable: Naveen Rao (CEO of Unconventional AI, building brain-inspired analog chips, formerly sold companies to Intel and Databricks) and Alex Finn (founder of Henry Intelligent Machines and Creator Buddy).Thank you to our exclusive sponsor: PayPal Open, One Platform for All Business: http://paypalopen.com/Timestamps:0:00 Cold open1:21 Welcome to Episode 162:44 Is the AI cost panic real, or just "token maxing"?5:08 It's not the intelligence, it's how people use AI9:40 Surgeons vs. shotguns: prompt discipline & matching models to tasks13:10 Naveen's path from Intel and Databricks to Unconventional15:08 Why developers aren't disappearing18:37 How energy overtook CapEx in the cost of compute21:34 The energy wall & getting to 3 orders of magnitude more efficient25:52 AI's PR problem & the data center backlash27:55 China's hunger vs. America's AI boogeyman30:38 Data center taxes, equity stakes & the politics of AI upside37:54 The Anthropic IPO & how these founders actually invest🔗 Guests:Naveen Rao, Unconventional AI: https://unconv.ai | https://x.com/AlexFinnAlex Finn, Henry Intelligent Machines: https://meethenry.ai | https://x.com/NaveenGRao🔗 Host:Alex Wilhelm, This Week in Startups: https://x.com/alex🔗 Referenced in this episode:Unconventional AI (analog chips for AI): https://unconv.aiMosaicML (acquired by Databricks): https://www.databricks.com/research/m...Nervana Systems (acquired by Intel)Creator Buddy: https://creatorbuddy.ioQwen (open-weight model Alex runs locally): https://qwenlm.aiNVIDIA DGX Spark: https://www.nvidia.com/en-us/products...Anthropic (filed to go public): https://www.anthropic.comHenry Intelligent Machines: https://meethenry.ai/🔗 Subscribe and follow:Newsletter and all platforms: https://thisweekinai.ai#ThisWeekInAI #AI #UnconventionalAI #NaveenRao #AlexFinn #AnalogComputing #AIcompute #AIenergy #TokenMaxing #AIlayoffs #AnthropicIPO

SpaceX bought Cursor for $60 billion. Satya Nadella says companies need to stop relying on third party AI models and build their own “token capital.” The focus is shifting from LLMs to the application layer sitting on top of them. Here to unpack what that means are guest experts Ali Ansari (Micro1) and Ryan Daniels (Crosby).Plus we get a sneak peek at their new contract redlining benchmark, a crucial eval for how well LLMs don’t just answer questions about the law, but demonstrate actual legal reasoning.Timestamps:0:00 SpaceX acquires Cursor7:36 Distillation vs. building your own model19:43 Nadella's "Frontier Without an Ecosystem"30:21 AI in the courtroom32:15 Ando: the intriguing new workplace tool1:05:38 Inside Micro1 and Crosby's new benchmark1:07:14Guests:Ali Ansari: https://x.com/aliansarinikMicro1: https://www.micro1.ai/Ryan Daniels: https://x.com/ryanjdanielsCrosby: https://crosby.ai/Relevant Links:Bloomberg: “SpaceX acquires Cursor for $60B”: https://www.bloomberg.com/news/articles/2026-06-16/spacex-cements-60-billion-deal-to-take-over-ai-startup-cursorQuinn Thompson “brilliant corporate finance” post: https://x.com/qthomp/status/2066859672749977988Business Insider: “Inside Cursor’s Wild Rise”: https://www.businessinsider.com/cursor-ceo-michael-truell-spacex-elon-musk-anthropic-2026-6Satya Nadella: “A frontier without an ecosystem is not stable”: https://x.com/satyanadella/article/2066182223213293753Joshua Browder’s Do Not Pay: https://donotpay.com/Harvard Magazine: “AI Outperforms Doctors in Emergency Room Tasks”: https://www.harvardmagazine.com/ai/ai-outperforms-doctors-diagnosis-harvard-studyJoshua Kushner “Long Humans” post: https://x.com/JoshuaKushner/status/2065093542809092465Ando: https://ando.so/Nim Ravid on X: https://x.com/Nim_Ravid1Subscribe to the TWiST500 newsletter: https://ticker.thisweekinstartups.comCheck out the TWIST500: https://www.twist500.comSubscribe to This Week in Startups on Apple: https://rb.gy/v19fcpFollow Lon:X: https://x.com/lonsFollow Alex:X: https://x.com/alexLinkedIn: ⁠https://www.linkedin.com/in/alexwilhelmFollow Jason:X: https://twitter.com/JasonLinkedIn: https://www.linkedin.com/in/jasoncalacanisCheck out all our partner offers: https://partners.launch.co/Great TWIST interviews: Will Guidara, Eoghan McCabe, Steve Huffman, Brian Chesky, Bob Moesta, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarlandCheck out Jason’s suite of newsletters: https://substack.com/@calacanisFollow TWiST:Twitter: https://twitter.com/TWiStartupsYouTube: https://www.youtube.com/thisweekinInstagram: https://www.instagram.com/thisweekinstartupsTikTok: https://www.tiktok.com/@thisweekinstartupsSubstack: https://twistartups.substack.com

This Week In Startups is made possible by:

Paypal

Today’s show:

*When we think of “Physical AI,” most of us conjure images of humanoid robots and self-driving cars. But there’s a lot more to the category than just breakdancing Chinese bots and Waymos, from excavators autonomously digging tomorrows worksites, to a robot the size of four pizza boxes deploying to the ISS.

Guest host Alex Wilhelm sits down with three founders in the trenches of the Physical AI Space:

Boris Sofman of Bedrock Robotics, which builds autonomous construction equipment

Jeff Hawke of Odyssey, a frontier AI lab specializing in world models for robotics and video games

Ethan Barajas of Icarus Robotics, designers of the free-flying Joy robot that will join the ISS crew in 2027

Guests:

Boris Sofman on X: https://x.com/bsofman

Bedrock Robotics: https://bedrockrobotics.com/

Jeff Hawke on X: https://x.com/jeffrey_hawke

Odyssey: https://odyssey.ml/

Ethan Barajas on X: https://x.com/ethanbarajas11

Icarus Robotics: https://www.icarusrobotics.com/

Relevant Links

Bedrock Excavators Remove 65,000 Cubic Yards of Dirt: https://www.enr.com/articles/61982-bedrock-robotics-excavators-remove-65-000-cubic-yards-of-dirt-on-southwest-project

Odyssey $310M Series B article: https://techcrunch.com/2026/06/17/world-model-maker-odyssey-nabs-1-45b-valuation-backed-by-amazon-and-other-big-names/

Icarus “Joyride” Mission announcement: https://thedebrief.org/icarus-is-building-the-robotic-labor-force-for-space-voyager-technologies-is-sending-a-next-generation-zero-gravity-robot-on-a-joyride-to-space/

PROWL: Prioritized Regret-Driven Optimization for World Model Learning: https://arxiv.org/abs/2605.18803

NVIDIA Cosmos 3 report: https://research.nvidia.com/labs/cosmos-lab/cosmos3/technical-report.pdf

ISS Columbus Laboratory Module: https://www.nasa.gov/international-space-station/columbus-laboratory-module/

NASA’s Astrobee flying robot: https://www.nasa.gov/astrobee/

Voyager Technologies: https://www.voyagerspace.com/

Vast Space: https://www.vastspace.com/

Blue Origin’s New Glenn: https://www.blueorigin.com/new-glenn

“Red Mars” by Kim Stanley Robinson: https://www.amazon.com/Red-Mars-Stanley-Robinson/dp/0553560735

“Rendezvous with Rama” by Arthur C. Clarke: https://www.amazon.com/Rendezvous-Rama-Arthur-Clarke/dp/0553287893

Theo Von “This Past Weekend” podcast: https://www.theovon.com/podcast

Anthropic statement on Fable/Mythos suspension: https://www.anthropic.com/news/fable-mythos-access

Timestamps:

0:00 The Year of Physical AI

1:28 Why everyone's bullish on world models

7:03 Physical AI is harder, but probably bigger

9:08 How to think about world models

16:00 When data becomes "adversarial"

25:00 How Bedrock powers autonomous excavators

30:48 Why teleoperation gets a bad rap

39:10 Meet Joy the space robot

41:33 Laser comms coming to next-gen stations

46:29 Why everyone hates data centers

57:33 When does Claude Fable come back?

Subscribe to the TWiST500 newsletter: https://ticker.thisweekinstartups.com

Check out the TWIST500: https://www.twist500.com

Subscribe to This Week in Startups on Apple: https://rb.gy/v19fcp

Follow Lon:

X: https://x.com/lons

Follow Alex:

X: https://x.com/alex

LinkedIn: ⁠https://www.linkedin.com/in/alexwilhelm

Follow Jason:

X: https://twitter.com/Jason

LinkedIn: https://www.linkedin.com/in/jasoncalacanis

Check out all our partner offers: https://partners.launch.co/

Great TWIST interviews: Will Guidara, Eoghan McCabe, Steve Huffman, Brian Chesky, Bob Moesta, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland

Check out Jason’s suite of newsletters: https://substack.com/@calacanis

Follow TWiST:

Twitter: https://twitter.com/TWiStartups

YouTube: https://www.youtube.com/thisweekin

Instagram: https://www.instagram.com/thisweekinstartups

TikTok: https://www.tiktok.com/@thisweekinstartups

Substack: https://twistartups.substack.com