Iron Foundry

Geoff Cook, CEO of Noom, started microdosing GLP-1 to put himself in the user's shoes. Noom's combined behavior-change-plus-medication GLP-1 programs are now 70% of revenue.

He joins Kyriakos Eleftheriou, CEO of Terra API, to talk about the $500 million Meet Group exit, why most people regain the weight in 18 months, why he wouldn't be surprised if Oura and Whoop start prescribing, why a closed loop is hard for the frontier labs to copy, and a peptide market he'd call a multi-billion-dollar black market.

0:00 I Started Microdosing GLP-1
1:39 The $500M Exit — Then Breakfast Made Him CEO
4:17 Why He Ignores Every Banker-Led Deal
6:15 The Only App Where More Screen Time Makes You Healthier
8:52 70% of Noom's Revenue Now Comes From GLP-1 Programs
11:34 Most People Regain All the Weight in 18 Months
15:57 Noom's Next Target After Weight Loss
20:08 I Wouldn't Be Surprised If Oura and Whoop Start Prescribing
22:58 AI Will Know Things About Your Body No Doctor Could
25:06 Why the Frontier Labs Can't Copy This
27:35 Every Noom Coach Now Has an AI Ghostwriter
29:04 You Only Get Paid If the Biomarkers Move
36:10 The One Peptide Noom Sells Besides GLP-1s
38:20 Peptides Are a Multi-Billion-Dollar Black Market
44:14 You're About to Enter a Golden Age of Health

KJ Dhaliwal raised $41M to build Lotus, an AI doctor backed by Kleiner Perkins and CRV.

He joins Kyriakos Eleftheriou, CEO of Terra API, to discuss why San Francisco still wins,” why Lotus doesn’t bill insurance, and what doctors will actually do in five years.

His prediction: doctors become the backup, and not using AI could eventually become medical malpractice.

00:00 Pizza, Sneakers and Options Trading

04:33 Why San Francisco Still Wins

15:14 Building Dil Mil Where Tinder Fell Short

22:25 Raising $41M for Lotus

27:17 No Insurance. Straight to the Patient.

32:36 “Healthcare Investors Are the Worst”

38:19 Could Avoiding AI Become Malpractice?

41:20 How Often Doctors Reject the AI

46:09 Medical Benchmarks Miss the Real World

52:15 What If the Doctor Is Wrong?

57:02 Doctors Become the Backup

1:00:10 Why OpenAI Won’t Run a Medical Practice

Vinod Khosla’s conclusion: “It’s over for human doctors. AI’s just going to be better.”

Full episode, transcript and key takeaways: https://tryterra.co/podcast/Vinod-Khosla-Its-Over-for-Human-Doctors

Vinod Khosla joins Kyriakos Eleftheriou in front of hundreds of YC AI Startup School founders to explain what the future of healthcare actually looks like.

0:00 Introduction

1:51 When “Do What Users Want” Is Wrong

2:26 Nobody Asked for AI or Fusion

3:51 You Don’t Need a Noble Reason to Start

4:58 Workers, Not Tools for Workers

5:43 Humans Get in the Way of AI

7:52 The Most Important Founder Decision: Who to Ignore

8:15 70% of Investors Add Negative Value

9:40 Most VCs Haven’t Earned the Right to Advise

10:01 Great Founder References Can Be a Bad Sign

10:45 Four Years Destroyed by Polite Investors

12:07 Sam Altman’s Hardest Problem

13:27 Disagreement Is Always Good

15:39 Follow the Talent, Not Today’s Failures

16:55 When Every Telecom Company Bet Against TCP/IP

19:33 “My Willingness to Fail Allows Me to Succeed”

20:57 Judge the Learning Curve, Not the Founder Today

21:13 European Investors Turn Great Ideas Into Decent Ones

22:57 No Doctor Can Keep Up With Medicine

23:41 Drug Interactions Kill More Than Breast Cancer

23:54 Why Vinod Diagnoses Himself With ChatGPT

25:12 Taking a Drug Approved Only for Women

26:44 Three Surgeons. Three Answers. One MRI.

28:28 It’s Over for Human Doctors

30:27 Doctors 73%. AI 88%. Doctors + AI 76%.

31:52 AI Mental Health Diagnosis: 93%

32:52 AI Rated as a Top-10% Therapist

33:19 A Frontier Model Is Not a Medical System

34:32 Human Error Is More Acceptable Than AI Error

36:08 $1T of Healthcare Expertise Should Be Free

38:20 Doctors Will Not Use AI If It Costs Them Fees

38:54 Why the AMA Opposes AI Doctors

39:13 Billing Decides Which Medical Technology Wins

40:34 Administrative AI Is the Smaller Opportunity

41:39 Going Founder Mode on Cancer

42:25 N=1 Medicine Can Route Around Drug Trials

43:06 Founders Go Around Obstacles. Big Companies Don’t.

44:07 Removing Human Drivers Could Save 30,000 Lives

45:38 Radiologists Without AI Are Killing People

46:11 MRI Scanners Should Output Diagnoses, Not Images

48:02 Epic Still Runs on a 1960s Language

49:34 How AI Agents Can Route Around Epic

51:02 Impact Over IRR. OpenAI Made No Sense.

51:48 Fusion Will Not Happen Until Someone Builds It

52:34 Entrepreneurs Make Forecasts Irrelevant

54:17 Light-Rail Capacity Inside a Bicycle Lane

Mo Elshenawy helped put driverless cars on San Francisco streets as President and CTO of Cruise.

Full episode and show notes: https://tryterra.co/podcast/Hims-Hers-CTO-Mo-Elshenawy

Now CTO of Hims & Hers, he joins Kyriakos to explain why rebuilding healthcare is harder.

They discuss closed-loop healthcare, clinical guidelines as code, when AI should remain silent, and whether AI will replace doctors.

Mo’s argument: models commoditize. Data and workflow compound.

At 25:13, Mo introduces Hims & Hers’ new AI weight loss companion.

CHAPTERS

(00:00) The Chessboard That Started It All

(02:22) AI Changed the Cost of Building

(04:17) Healthcare Is Harder Than Self-Driving

(06:13) Twice the Cost. 3.7 Fewer Years.

(09:17) Conviction Before Driverless Worked

(13:47) The Lunch That Changed His Career

(15:25) Healthcare Is Actually Sick Care

(18:03) Write Your Own Onboarding Plan

(19:43) AI Without a Platform Is Demoware

(23:06) Why OpenAI Can’t Close the Loop

(25:13) Hims’ AI Weight Loss Announcement

(28:42) Silence Is a Feature

(31:53) Clinical Guidelines as Code

(33:32) Most AI Benchmarks Are Theater

(36:12) Models Commoditize. Workflow Compounds.

(38:46) AI Will Not Replace Doctors

(44:37) AI Makes Bad Engineers Worse

(48:15) Healthcare Is Still Blockbuster

(52:00) Is Hims Getting Into Peptides?

(54:06) Concierge Care in Everyone’s Pocket

(58:46) AI Destroys the Excuse Not to Build

More episodes

George Hadjivarnava co-founded Foody as CTO with no VC money, made it profitable in under 2 years, and sold it to Delivery Hero in the biggest tech acquisition in Cyprus.

Full episode and show notes: https://tryterra.co/podcast/George-Hadjivarnava-Founding-Foody-and-building-AI-at-Terra

Scaling through COVID from restaurant menus to full supermarkets, total downtime: 45 minutes.

Now he leads engineering strategy at Terra and he believes the next infrastructure layer won't be built for developers at all.

• Why in two years 90% of Terra's customers will be agents instead of developers
• Why MCPs didn't invent anything new, and the missing layer that transfers intent instead of commands
• Why nothing performs well past 100k tokens despite 1M context windows
• The decision log: AI's most underrated asset inside a company
• Bootstrapping Foody from phone orders and cash payments to a national exit
• Compressing hiring from 3 weeks to 1 day with 30,000 applications a year
• The 24/7 AI health system: even a billionaire's team of doctors goes to sleep

(0:00) Intro

(1:30) Startup ecosystem in Cyprus: more incubators than startups
(7:20) Why ecosystems compound
(23:00) Flunking Imperial and building websites to pay rent
(29:00) Foody's first version: phone orders, cash only
(39:35) The Delivery Hero acquisition and COVID scaling
(43:05) 45 minutes of total downtime
(44:45) You can one-shot an app. You can't one-shot 100M users
(50:25) Joining Terra
(56:40) The missing agent layer: intent, not commands
(1:02:50) The 24/7 AI health system
(1:09:35) The 90% agents prediction
(1:11:05) Hiring compressed from 3 weeks to 1 day
(1:15:35) The decision log: building a company brain
(1:19:10) Why 1M context windows are marketing
(1:26:35) Three changes to become a real technology hub

⁠Geoff Ralston⁠⁠ built Rocket Mail before Yahoo Mail existed, built Lala before Spotify worked in America, sat in Steve Jobs' living room to get an acquisition approved, and ran Y Combinator as President. Now he says the most important thing he's ever worked on is making sure AI doesn't kill us.

Full episode and show notes: https://tryterra.co/podcast/Early-to-Every-Wave-Former-Y-Combinator-President-Geoff-Ralston

In this episode with ⁠⁠Kyriakos Eleftheriou⁠⁠, Geoff speaks about YC, SAIF, the Yahoo acquisition, and the top learnings after decades in silicon valley

00:00 Introduction

03:45 The $70K bet that built Sand Hill Road

06:07 Early to every wave

06:59 "The most important thing I've ever worked on"

08:18 Timing is luck. Here's how to get lucky anyway

10:46 Quitting HP the day he saw the Mosaic browser

15:39 Paul Graham's "packets of cash" acquisition rule

19:35 The Yahoo revolt that almost killed the deal

33:45 Zuck killed Facebook Music in one sentence

39:05 Why Steve Jobs offered him a job — and why he said no

45:40 Running Y Combinator

50:37 Steve Jobs' real superpower

54:06 Why AI is more dangerous than people think

01:01:15 Peter Thiel is wrong about competition

01:03:31 Marc Andreessen is wrong about introspection

--

Full podcast and writeup → Geoff Ralston on early YC, AI risk, and why Thiel is wrong about competition

By the age of 25, David Lee survived stage 4 lymphoma. Not only did he not let it stop him, he worked early in Google, became a partner at SV Angel during the Airbnb / Twitter / Dropbox / Snapchat era, and now leads Samsung Next. In the latest discussion with Kyriakos Eleftheriou in Los Angeles, they discussed:

Full episode and show notes: https://tryterra.co/podcast/Head-of-Samsung-Next-David-Lee-Live-in-LA

CHAPTERS

(00:00) Introduction

(02:30) Why digital health still hasn't had its "reusable rocket"

(07:00) "Law school is coding for your mind"

(08:30) How Elon, Dorsey and Belichick boil any problem down to 3 things

(11:00) Telling Yo-Yo Ma from a college cellist — how to spot real talent

(14:00) The NFL gets the #1 pick wrong 50% of the time — so how do you pick founders?

(16:30) Most pro athletes are scared in big games

(19:30) Without a killer product, you have no strategy

(22:00) Vanity beats health — why no one wears the clunky device

(29:00) Your kids will never do what you tell them — they copy what you do

(32:30) Jensen Huang, Saturdays, and why the top person sets everything

(37:30) The "red zone of intelligence" AI will never touch

(38:30) Why the great fitness instructor gets paid more in the AI era

(40:30) Why VC is the wrong financing model for the next decade

(41:30) Mike Ovitz and CAA — the playbook for winning in AI

Full podcast and writeup → David Lee on Samsung Next, picking founders, and building inside a giant

42,000 tickets for one New York event. 15,000 gym affiliates. $0 spent on influencers. HYROX CGO Douglas Gremmen joins Kyriakos, CEO of Terra API , at the Connected Health Summit in Stanford to talk through what actually drove that — and what didn't work.

Full episode and show notes: https://tryterra.co/podcast/hyrox-cgo-douglas-gremmen

The US was their first market outside Germany and their hardest. Three years of losses. A single Lance Armstrong weekend changed the momentum — not because they paid him, but because he just turned up. The gym affiliate program started as inbounds with no structure; it's now the foundation of a bet to reach 100 million people. The CrossFit comparison gets a real answer, not a PR one.

But the most interesting parts are the early decisions: rejecting the $2 race t-shirt, the photo package that created the organic flywheel, and a simple principle — invest in the event experience and don't run paid marketing.

00:00 Welcome to Stanford

01:46 Running HYROX at a Conference & the "Bubble" Moment

03:48 The Sweaty Barrier to Participation Problem

05:44 Douglas's Backstory: Sports Marketing & How HYROX Was Born

08:59 From First Event in Hamburg to Leading Global Expansion

10:29 Early Growth: The Van, T-Shirts, Red Bull & the Photo Flywheel

12:36 Global Expansion: Japan, Brazil, India & Getting on the Calendar

14:56 Why the US Was the Hardest Market to Crack

17:28 Losing Money in America — Keeping Belief When the P&L Doesn't Work

19:09 The Lance Armstrong New York Bet: $1M & 30M Impressions

20:27 Influencers at HYROX: Always Organic, Never Paid

21:52 Scaling to 42,000 Tickets: 9 Days of Racing Bigger Than NYC Marathon

23:45 Building Operations: Christian Tutske's Eye for Detail

25:39 Why Athletes Keep Coming Back: Community & Identity

28:00 HYROX 365: From Events Company to 15,000-Gym Affiliate Business

30:54 Accelerating Gym Growth: Big Box Operators & Chicken-and-Egg Dynamics

34:27 HYROX as a Gym Retention Play: Coach Education & Member Stickiness

36:49 Fragmented Gym Tech & the Data Opportunity

38:23 HYROX vs. CrossFit: Honest Comparison & 1,500 Shared Affiliates

43:10 The Pickleball Analogy: Low Barrier + Social + Price = Rocket Ship

46:49 The Path to 100 Million People: Events, Gyms & Knowing Who Trains

49:50 HYROX x Puma: Building a Fashion Brand & the First Dedicated Shoe

54:14 Contrarian Take on the Future of Fitness: Back to Fundamentals

56:35 Physical Events vs. Social Media: Why Credibility Is Harder to Fake

59:11 Post-COVID Perfect Storm & Tickets Selling Out as the New Challenge

01:01:09 Where Will HYROX Be in 5 Years? Formula One, Olympics 2032 & 100M People

Full podcast and writeup → Douglas Gremmen on selling 42,000 tickets in a day and scaling HYROX to 15,000 gyms

In the latest episode, Kyriakos Eleftheriou sat down with Roman Bugaev , CTO of Flo and Vlad Nedosekin Director of AI Platform, at the Terra API HQ in London, to discuss how they built the top health AI platform globally for women's health.

Full episode and show notes: https://tryterra.co/podcast/cto-director-of-ai-at-flo-health-roman-bugaev-vladislav-nedosekin

CHAPTERS

(0:00) Intro — Flo Health: From 20 People to 80 Million Users

(1:02) How Flo became the fastest-growing health company in the world

(1:48) Roman's early days: 20 employees, no revenue, product-market fit

(3:15) How did you know the product was a hit?

(3:25) The underserved women's health market — everyone was building Uber alternatives

(4:31) First ML: neural networks for cycle and symptom prediction

(5:31) Product evolution — from symptom tracking to AI-powered insights

(5:38) Building chatbots inspired by how doctors ask questions

(9:18) A/B testing at scale — Flo's custom experimentation platform

(11:43) Engineering structure: autonomous two-pizza teams

(13:51) Team mistakes — why separate mobile and backend teams failed

(15:29) Scaling from 4 servers to 600 services and petabytes of data

(17:53) "Whenever it's possible, we are NOT doing AI"

(23:15) Why temperature data is critical for ovulation prediction

(25:09) Why Flo is the most accurate period tracker — data diversity advantage

(28:04) Competition: "We don't really have real competitors"

(29:00) AI content creation — generating personalized medical articles

(31:01) Hallucinations vs. conflicting medical sources

(32:34) The three-person blind test: when AI disagrees with humans

(35:10) AI is more consistent than clinicians — but biased against women

(36:52) Fine-tuning open-source models on synthetic women's health data

(38:25) User profile: the foundation of Flo's personalization

(41:19) The digital avatar — your AI health twin that notices what you don't

(43:09) AI router: like a GP triage system for language models

(46:04) Router also controls tone of voice and remembers past conversations

(47:29) "Evaluation, evaluation, evaluation" — how Flo picks models

(48:40) Model stability: why proprietary model updates are dangerous for medical AI

(51:01) Anonymous mode: privacy that enables AI instead of blocking it

(53:49) On-device ML for the most sensitive health data

(56:03) Cloudflare outage — "when everyone is down, you're allowed to be down"

(56:58) Fine-tuning Llama 7B on Databricks — 10,000+ GPU hours per run

(58:07) Training vs. inference cost breakdown

(59:45) 100,000-token prompts: the hidden cost of medical AI

(1:01:05) Build vs. buy: "Build your competitive advantage, buy everything else"

(1:04:47) Value creation vs. value capture teams

(1:07:04) The future: AI that knows you better than you know yourself

(1:09:00) Time series models: the future of health prediction from wearables

(1:10:38) Q&A

Full podcast and writeup → Roman Bugaev and Vladislav Nedosekin on scaling Flo Health to 80M users and shipping AI in femtech

In this episode, Kyriakos Eleftheriou, Terra API CEO sits down with Sacha Michaud, co-founder of Glovo, one of the biggest tech successes in the European tech ecosystem,and Yellow.VC, an early-stage fund

Full episode and show notes: https://tryterra.co/podcast/glovo-and-yellow-vc-co-founder-sacha-michaud

Sacha shares the journey of building the on-demand delivery service from its inception in Barcelona to its expansion across 25 countries and a $2.3 billion acquisition.

He discusses the initial challenges of fundraising, competing against larger companies, and the importance of passion and culture in building a successful team. Michaud reflects on the early days of Glovo, where they experimented with a simple app that allowed users to order anything from the city, leading to unexpected viral growth despite initial losses. He emphasizes the significance of adapting to market needs and the evolution of Glovo's business model over the years, particularly during the COVID-19 pandemic, which accelerated the demand for grocery delivery.

Chapters

00:00 Introduction

02:29 The Birth of Glovo: From Idea to Launch

05:59 Viral Growth and Initial Operations

12:06 Realizing the Potential of Glovo

21:39 Expanding into New Markets

38:55 Navigating Competition and Strategic Decisions

50:39 The $2.3B Acquisition

55:00 Yellow, and investing in new startups

01:04:11 AI and Future Trends in Delivery

Full podcast and writeup → Sacha Michaud on building Glovo into a $2B delivery giant and now backing founders at Yellow.vc