Founders in Arms

Jensen Huang is out at Nvidia, so Immad and Raj recorded an emergency pod the morning the news broke.
They start with what his exit means for Nvidia and the AI trade, the lines a CEO can't cross, and whether Travis Kalanick was done dirty at Uber. From there they get into interest rates, Anthropic's push to become an everything app, and how Tribe now runs with AI on the inside, with agents handling design, engineering, analytics and moderation.
The back half is for founders: what to build right now, why the marketing co-founder is the new technical co-founder, skipping VCs with an LLC model, what the Anthropic and OpenAI IPOs could mean for venture, and whether you should walk away from a non-AI startup. They wrap with how they're investing, from buying Anthropic to the S&P vs QQQ, and why robotics could be the big trade of the next decade.
What you'll learn:
Why Jensen's exit is a long-term risk for Nvidia even if the stock holds up short term
The lines a public company CEO can't cross, and why an undisclosed relationship with a direct report means an instant fire
Whether this pops the AI bubble, and how rising interest rates could change the AI trade
How Tribe runs with humans on the outside and AI on the inside
How AI agents are starting to run sales, ops and support at small companies
Where founders should build right now: AI infrastructure, regulated spaces or go to market
Why an LLC funded by friends and family could become a more common startup model
What the Anthropic and OpenAI IPOs could mean for VC liquidity
How to think about quitting, spinning off or handing off a non-AI startup
Why Immad and Raj are bullish on robotics over the next decade

Timestamps:
(00:00) Intro
(01:00) Jensen's out. What happens to Nvidia?
(03:15) The lines a CEO can't cross
(07:24) Travis Kalanick and founders getting pushed out
(09:34) Does this pop the AI bubble?
(11:38) Interest rates vs the AI trade
(14:09) Anthropic wants to be the everything app
(15:04) How Tribe runs with AI on the inside
(16:31) AI agents running sales, ops and support
(20:01) AI moderation and a comeback for anonymous social?
(23:03) The marketing co-founder is the new technical co-founder
(24:02) What should founders build right now?
(27:14) Skip the VCs: the LLC model
(28:23) The VC market: hot AI vs everyone else
(29:57) What the Anthropic and OpenAI IPOs mean for VC
(32:44) Should you quit your non-AI startup?
(36:51) Spin-offs, selling and handing off to a co-founder
(39:24) Buying Anthropic, S&P vs QQQ
(41:37) Robotics and the next decade

Jean-Denis Greze (CEO @ Town), joins Immad and Raj live to break down the pivot from AI tax prep to a work assistant, why horizontal AI beats vertical AI, and how to hire in the age of agents. Before Town, he was head of engineering, CTO and CPO at Plaid.

In this episode, you'll learn:
- Why Town walked away from AI tax prep after its first tax season, and how the team handled telling customers and employees
- How a blank homepage with a "connect Gmail" button turned into Town's first users and product-led growth
- Why Town bets that horizontal AI beats vertical AI, and where it won't work (legal, payroll, HR)
- Why being neutral across AI labs resonates with companies of 500 to 2,000 employees
- How Town hires engineers: a 2.5 to 3 hour in-person build with AI agents, judged on when you read the code
- Why product advantage now lasts months, not years, and what to build instead (network effects, brand)

Rahul Vohra is the founder of Superhuman, the email client that was acquired by Grammarly. Grammarly has since taken on the Superhuman name, and Rahul now leads Superhuman Mail. Before Superhuman, he built Rapportive, the first Gmail extension to scale to millions of users. LinkedIn acquired it 20 months after its first line of code. Rahul studied computer science at Cambridge, ran Cambridge University Entrepreneurs, and went through Y Combinator in Summer 2010.
In this episode, Immad Akhund and Raj Suri talk with Rahul about the origin of the Superhuman name, how a leaked demo link took Rapportive from 10 to 10,000 users overnight, and why being early to a new platform still pays off. Most of the conversation covers M&A: how to read a buyer, how to protect your leverage, and what a banker is actually for.
What you'll learn:
The four things an acquirer needs to be buying for a company to sell at a top valuation
How Rahul sold Rapportive at a premium with two weeks of runway left, including letting a no-shop period expire
Why building multiplayer features moved Superhuman's team net dollar retention from about 70% to 122%
Why founders should be early to every new platform, from browser extensions to ChatGPT apps and MCP
How the founder "bell curve" changes how much the idea matters
What M&A advisors actually contribute, and how they pace hot and cold buyers

Chapters:
(00:00) The four things acquirers are really buying
(01:07) How Grammarly became Superhuman
(04:15) Superhuman vs. superintelligence
(07:04) From the BBC Micro to a Cambridge PhD
(09:48) Fundraising on good vibes at Cambridge University Entrepreneurs
(12:22) Building Rapportive
(14:14) The leak that brought 10,000 users overnight
(18:19) Be early to every platform
(20:11) Email's trillion-hour problem
(22:33) Raising before Demo Day and meeting LinkedIn
(23:56) The LinkedIn API deal
(29:21) Does the idea matter? The founder bell curve
(31:33) Going multiplayer: net dollar retention from 70% to 122%
(33:45) Selling Rapportive with two weeks of runway
(44:20) Selling Superhuman: 16 buyers and the banker's real job

Stephen Balaban is co-founder and, as of a recent leadership change, CTO of Lambda, the AI cloud infrastructure company he and his twin brother started in 2012. It took five pivots — augmented reality, a facial-recognition contact book, a camera embedded in a baseball cap, the AI image app Dreamscope, then finally workstations and servers — before Lambda found a business that made money in 2017. Along the way, Stephen kept the company alive by consulting on the side, including projects with Airbus and the creators of South Park.
What you'll learn:
Why it took five pivots and 14 years for Lambda to find product-market fit
How side consulting work — for Airbus and even South Park's creators — funded the company through its leanest years
Why mainline Silicon Valley VCs kept passing on Lambda even while it was profitable, and why that didn't change after ChatGPT
How a Series D round pulled largely from Taiwanese manufacturers and family offices instead of traditional venture capital
What pushed Stephen to step down as CEO and bring in Michel Combes, a former CEO of Sprint and SoftBank International
Why Stephen thinks the disinformation around data centers — noise, water use — doesn't hold up

Chapters:
(01:10) Meet Stephen Balaban, co-founder and CTO of Lambda
(02:40) Palo Alto in 2012 and Lambda's earliest days
(03:42) Building Heads Up, a facial-recognition contact book for iOS
(05:06) The ImageNet moment and training neural nets on NVIDIA GPUs
(06:14) Five pivots: from augmented reality to Dreamscope
(08:55) Funding the company through consulting, including work with Airbus and South Park's creators
(12:13) Writing down the goal to IPO back in a 2012 notebook
(14:28) Raising a first $600K, including Austin Russell's $20K check at a $400K valuation
(24:29) Lambda's climb from $3M in revenue to a $1B run rate
(26:57) Why Silicon Valley VCs kept saying no — even after ChatGPT
(34:04) A Series D built largely on Taiwanese manufacturers and family offices
(37:13) Immad on Mercury Books, Mercury's new bookkeeping product
(44:12) Stepping down from CEO to CTO and bringing in Michel Combes
(52:34) Debunking data center disinformation, from water use to noise

What if the company students use to find their first internship became one of the most important players in training AI?
Garrett Lord, co-founder and CEO of Handshake, joins Immad Akhund and Raj Suri to break down Handshake's unlikely pivot. Handshake started as a way to help college students — regardless of where they went to school — find internships and jobs, and grew into a $200M+ ARR business used by most students in America. But over the last 18 months, Garrett has built a second business inside Handshake: using the company's network of 30 million students and alumni to help AI labs train their models on real, high-quality, professional-domain data — from oil and gas to finance to scientific research. That business alone has gone from zero to nearly $2 billion in revenue in about a year.
The conversation goes deep on how this actually works: recruiting domain experts, building task environments that function like video games, scoring model performance against expert-validated tasks, and why 70% of the money spent training a model today goes toward reinforcement learning rather than pre-training. Garrett, Immad, and Raj also cover the open-weight vs. frontier model debate, why China may already be ahead on robotics, and what jobs might look like in a world where AI models can eventually learn continuously, on the job.
The episode closes with a genuinely open-ended debate between Garrett and Immad about what humans will actually do for a living, and for meaning, if knowledge work is mostly automated — and how disruptive that transition might be along the way.
What you'll learn:
How Handshake used its network of 30 million students and alumni to build a second, multi-billion-dollar business training AI models
Why 70% of AI training spend now goes toward reinforcement learning, not pre-training
How AI labs identify gaps in their models and commission the specific data needed to close them
Why data, not algorithms, may be the real long-term moat for AI companies
Why computer use — AI navigating real software and websites — has recently gotten dramatically better
Why China may already be ahead of the U.S. in deploying real-world robotics
How enterprises like Mercury are likely to use a mix of frontier and open-source models going forward
What Handshake learned scaling a data business from zero to nearly $2B in under two years
Garret and Immad's differing views on what human work and meaning look like if knowledge work becomes automated

Timestamps:
(00:19) Introduction and Handshake's origin story
(01:39) Handshake's new business: training AI models on real-world data
(02:30) How Handshake's 30M-person network became a moat
(04:24) Inside the "video game" environments used to train agents
(05:00) Why 70% of AI training spend now goes to reinforcement learning
(08:18) How AI labs commission specific data from Handshake
(12:10) Why computer use has finally gotten good
(14:11) What models are still bad at, and why
(17:00) The "8 people can agree" test for what AI can be trained to do
(18:23) China's 2 million working robots, and why the US is behind
(22:12) Open-weight vs. frontier models, and how enterprises will use both
(31:15) Scaling from zero to $2B: what broke along the way
(34:02) Handshake's "Olympic pace" culture value
(39:41) Why continuous learning is the next frontier for AI
(42:07) Bill Gates' essay on AI, job loss, and taxing tokens
(45:23) Immad and Garret debate what jobs and meaning look like in an AI-driven future