Founders and Empanadas

He shot 500 shots a day, 7 to 8 in the morning, on a shooting machine in his high school gym. If he skipped it, the anxiety going into the weekend was unbearable. He still remembers his friend Adam offering to pick him up on a Friday night and having to say no because he had shots to get up. Looking back he calls it a little toxic. He also credits it for the one thing he thinks every founder has to do out loud, which is name the trade-offs.

Tonner Jackson is 35 and the founder of Rare Days, a roughly 60-person firm that has built product with more high-impact creators than almost anyone. Doctor Becky came to him six months into the business with 80,000 Instagram followers and an idea for a community, and he told her she was actually building a media company. Before that there was a frozen smoothie subscription he grew in his own backyard at UBC, three years matching former entrepreneurs to science-based startups, and a MasterClass-for-sports concept that got as far as pitching Sean McVay before the deal collapsed weeks ahead of the pandemic. Rare Days now does AI transformation work for enterprises. He was in Medellín for this recording because his bathhouse, Madre, opened the next day.

In this episode of Founders & Empanadas he sits down with Joshua Eidelman to argue that judgment is the scarcest thing in business, scarcer than AI talent or design talent, and that the way you manufacture more of it is by tolerating the question most founders find maddening: why are we doing this. He is equally blunt that remote is a worse cultural environment and that remote companies should stop fooling themselves about it, which is why Rare Days spends real money on one annual summit, and he points at a year of zero regrettable churn as the return. Along the way, the co-founder conversation he handled with his ego, voice agents replacing a call center nobody wanted to staff, and Walt Disney turning a prop around to check the back.

Subscribe for more founder conversations over empanadas. New episodes on YouTube, Spotify, and Apple Podcasts.

Timestamps:

00:00 Intro

01:36 What college basketball taught him about trade-offs

03:37 500 shots a day, and the Friday nights he gave up

05:03 “I'm not sure this is the happiest path for me”

05:59 The frozen smoothie subscription

06:53 Why frozen vegetables were the thesis

08:52 Growing the first harvest in the backyard at UBC

11:11 The co-founder he looked up to

11:33 “I think I should be the CEO”

12:21 Leaving, the sale, and a dollar an hour

15:36 Pitching Sean McVay on a MasterClass for sports

16:19 The deal collapses, then the pandemic hits

17:24 Meeting Doctor Becky at 80,000 followers

20:36 Why the creator economy got overhyped

21:44 Customer obsession, and reading Bezos

24:22 Remote is a worse culture, and the Feynman line

26:19 A year with zero regrettable churn

28:08 Chesky's 11-out-of-ten exercise

31:23 Voice agents and the shift-attendance call center

33:38 Madre, and the experience he's after

36:35 The Aufguss ritual, and no phones inside

40:30 Walt Disney and the unpainted prop

43:33 Why he wants employees to ask why

44:35 Judgment is in the shortest supply

Brian M. Green got laid off, and instead of going straight back into the market he asked himself what he was actually good at. He'd spent decades on the data side of things, a master's in urban planning he never practiced with, a PhD candidate in public policy and health, and a long habit of asking whether an intervention actually helped anyone. So he hung the consulting shingle back up and pointed it at the least fundable thing in enterprise AI: the layer underneath the thing everybody wants to buy.

The pitch is a question. Before you pilot anything, what is your replacement thesis? Not which jobs you're replacing, which operational work is being replaced. Ten back-end processes that eat time and generate errors and hold up vendor payments, name them. If you can't, you have no business piloting AI in your company. That's where the Gartner number comes from, the one everyone cites about 95% of AI projects never leaving pilot. Nobody defined what success looked like before they started.

Joshua Eidelman sits down with Brian on governance as infrastructure, which he means both literally and as a metaphor, and on why the word governance makes American executives stop listening the second they hear it.

They also get into data centers, where Brian's urban planning background actually shows up. He wrote the prospectus on how to think about siting them, built for all three stakeholders: the developers, the regulators, and the community groups who show up angry. His view is that nuclear power plants are the closest historical analogue we have, that a data center doesn't have to be one gigantic footprint, and that we don't yet have good enough data on water, sound, or vibration to make the decisions we're making right now.

Highlights:

• Philly Builds AI, and the case that Philadelphia is badly underrated in tech

• Governance as infrastructure, and why he had to stop calling it governance

• Why nuclear power plants are the closest thing we have to the data center question

• A campus with trees and solar instead of one enormous building

• The data we don't have yet: water recycling, sound damping, vibration, soil type

• Why he thinks Las Vegas shouldn't exist as a city, and what that has to do with civic responsibility

• Flying versus driving, and why we over-amplify the risks we can't control

• The readiness assessment: is your data ready, your culture, your leadership, your workflow

• The replacement thesis, and the one question that gates every pilot

• Ambient scribes save doctors time. Brian wants to know if patients got better outcomes

• Health Rounds AI, and building guardrails that refuse to let a clinician diagnose with it

• Prompt injection that sits dormant for nine months before a bad actor turns it on

• Why he builds in public and believes in open source even when it costs him IP

• The belief most founders would disagree with: build the foundation before you raise

Subscribe for more founder conversations over empanadas. New episodes on YouTube, Spotify, and Apple Podcasts.

Timestamps:

00:00 Intro0

2:44 Governance as infrastructure, literally and metaphorically

05:40 Data centers, and why nuclear plants are the closest analogue

16:27 Why Las Vegas shouldn't exist as a city

18:54 How we measure risk: flying versus driving

22:58 The layoff that started the company

25:39 The AI readiness assessment, bucket by bucket

27:47 Why 95% of pilots die, and the replacement thesis

30:57 Ambient scribes save doctors time. Is that the right metric?

50:31 Build first, then go to investor

In 1993 the CEO of a paper mill in Portland, Maine came down to IBM Research and asked for help. Alok Aggarwal’s team told him the truth, which was that they had no idea how paper gets made. They spent three months learning, then built a scheduling system out of agents and meta-learning, words that would not become fashionable for another thirty years. The mill improved by about 1.5 percent, worth roughly nine million dollars a year. Nobody at IBM Global Services understood the thing well enough to run it, so the client ran it themselves for the next twenty-four years.

He calls himself an accidental entrepreneur. Sixteen years at IBM Watson Research, then he opened IBM’s India Research Lab in 1998, then co-founded Evalueserve in 2001 and coined the term knowledge process outsourcing on the way to 4,500 people. Scry AI is the company he started to get back to building products. From the paper mill he took the four parameters he still judges every project by: is it timely, is it high quality, does it make money, is it robust. His view is that generative AI scores well on the first one and is terrible on the third.

In this episode of Founders & Empanadas he sits down with Joshua Eidelman to argue something most of LinkedIn will resist: generative AI is completely real, and several of the companies building it will not survive the decade. His case is about funding, debt, and the revenue required to service both at a 15 percent return, and he has already moved his own date for the shakeup forward twice, most recently to September 2027, largely because Chinese labs are far more financially disciplined than American ones. Along the way, the nine months of zero revenue that ended with 9/11, why a product company needs three co-founders when he only has one of himself, and the reason COBOL is still running your bank.

Highlights:

- Nine months into Evalueserve with no revenue at all, and the two-month freeze after 9/11 that forced the pivot

- Why nobody was selling to small and mid-sized companies in 2001, and how thirty-to-fifty-thousand-dollar projects got them profitable

- Doubling revenue every year for four years, and the moment Deutsche Bank found them

- The 2004 prediction about data management services that he made and then failed to act on, which he still calls his error

- Person-to-person offshoring, called in 2008, now going by the name gig economy

- Two co-founders for a services company, three for a product company, and why running solo makes him “handicapped”

- The four parameters from the paper mill, and the one where AI keeps failing

- Why 5 percent edge cases in a demo become 12 percent in production, and CFOs stop signing

- COBOL was created in 1960 and still runs most banks, insurers, and government offices

- 400 million jobs gone by 2050, and the population decline that could have us begging robots to work

- Why he would rather build after a crash than during a boom

Subscribe for more founder conversations over empanadas. New episodes on YouTube, Spotify, and Apple Podcasts.

Timestamps:

00:00 Intro

01:50 The Mountain View billboards remind him of 1999

02:37 Johns Hopkins, then 16 years at IBM Research

03:52 The paper mill that ran on agents for 24 years

05:19 Founding Evalueserve and coining KPO

08:01 Nine months of zero revenue, then 9/11

09:10 The $30k projects that got them profitable

11:25 Calling person-to-person offshoring in 2008

16:21 Two co-founders for services, three for product

19:38 Why the big AI labs may not survive

22:40 The funding, debt, and 15 percent return math

24:29 The four parameters from the paper mill

26:03 Moving the shakeup date to September 2027

29:43 COBOL is still running your bank

31:43 400 million jobs gone by 2050

33:17 The population decline nobody prices in

35:04 His real worry is that we forget how to talk

36:11 Why he would rather build after a crash

38:08 The belief 99 percent of founders would disagree with

Almost everyone who climbs Kilimanjaro stops at Stella Point and calls it the summit. It isn't. The actual highest point in Africa is Uhuru Peak, roughly ten blocks further on, and Rishi Bhatnagar says that ten-block walk was the hardest forty minutes of his life. What kept him going was picturing his grandkids asking whether he made it, and having to tell them grandpa quit early.

This is his fourth startup. Two worked, one didn't. He started out as a credit rating analyst at an agency now part of S&P Global, covering steel and engineering, and touring compressor factories in India gave him the idea he still builds on: every one of those plants made the same widget, and every one of them had a completely different process for making it. His conclusion was that there is no monolithic AI coming to fit everybody. AI is an ingredient. What you do with it decides the outcome.

In this episode of Founders & Empanadas he sits down with Joshua Eidelman to make an argument that most founders will push back on: the AI labs are locked into a finite game they eventually have to lose, and everyone else should stop competing and start absorbing the people they'd normally treat as rivals. Along the way, a power plant with 17 years of data and only six usable failure events, arena staffing models for Tampa Bay Lightning games, and the VC advice he turned down and still regrets.

Highlights:

• The Kilimanjaro false summit, and why the last ten blocks are the ones that break people

• 17 years of plant data that contained six failure events, and the Monte Carlo model built on top of it

• Services versus product: showing up with three kinds of wood, or showing up with two tables and four chairs

• Why he thinks IT consulting stalled because clients froze, not because the work stopped mattering

• The chasm between your AP system and your contract system, and why nobody can tell if you're already a vendor

• Why mid-size and small companies are the biggest beneficiaries of AI, at a hundredth of the old cost

• Meeting customers where they are, including running an enterprise product through WhatsApp

• Finite games versus infinite games, and why he expects the labs to lose theirs

• The VC advice he turned down: build it just for Shopify sellers first, and don't boil the ocean

Subscribe for more founder conversations over empanadas. New episodes on YouTube, Spotify, and Apple Podcasts.

Timestamps:

00:00 Intro

01:17 Welcome to Founders & Empanadas

01:59 The Kilimanjaro question

03:16 Stella Point is not the summit

05:39 From credit rating analyst to founder

06:31 Four factories, one widget: AI is an ingredient

07:47 Forecasting staffing for Tampa Bay Lightning games

09:06 17 years of plant data, six failures

11:36 What services founders know that product founders don't

14:46 Why IT consulting stalled

16:28 The chasm between AP systems and contracts

18:44 Why small and mid-size companies benefit most

20:29 Empanada intermission

21:27 Data velocity is fast, decision velocity is not

25:27 The iPhone never came with a manual

27:49 CPG, manufacturing, and the Walmart jeans story

31:49 Finite games vs. infinite games

33:05 Why he expects the AI labs to lose

36:37 The advice he turned down and still regrets

38:40 Where to find Rishi and Quaeris

Marco Benitez was preparing to move his family from Mexico to Switzerland for the next chapter of a successful Big Pharma career when his wife asked him four words: “Why are you so sad?” He already knew the answer. The small fitness product he was building outside work made him feel more alive than the career he had spent years creating.

Marco left Roche and went all-in on RookMotion, a wearable and fitness platform that eventually served more than 150 gyms. His motivation was unusual for a founder: he had not identified the perfect problem. He wanted to prove that he could build a company from scratch.

In this episode of Founders & Empanadas, Marco and Joshua Eidelman also examine the personal weight behind the pivot. Marco explains why he sees a team of more than 30 employees as responsibility for more than 100 people once their families are included, why he considers entrepreneurship one of the best investments he has ever made, and how faith helped him accept the limits of what a founder can control.

Highlights:

• Why Marco left a successful Big Pharma career after one question from his wife

• Why he started building before he had identified the right problem

• How COVID forced RookMotion to leave fitness hardware behind

• The insurance conversation that revealed the API opportunity

• What 100 Techstars mentors saw inside the company

• How fundraising and founder risk differ between Latin America and the United States

• Why global hiring requires founders to balance money, talent, and culture

• Why integrating health data yourself can take months of partnerships and certifications

• Why Marco counts a 30-person team as more than 100 people

• How “God First. Do Hard Things. Serve Others.” shapes his approach to leadership

• Why Marco believes founders have far less control than they think

Timestamps:

00:00 - Building a company before finding the problem

01:21 - Marco, empanadas, and Latin America’s startup ecosystem

04:22 - Raising money in Latin America versus the United States

07:41 - Global talent, startup economics, and company culture

08:37 - “Why are you so sad?” and the decision to leave Big Pharma

14:26 - Building RookMotion to prove he could create something

15:56 - COVID closes 150-plus gyms and forces a major pivot

19:45 - Techstars Mentor Madness and 100 calls that changed the company

27:21 - Why wearable and health data integrations are so difficult

30:08 - What entrepreneurship has cost Marco

32:44 - The pressure of supporting more than 100 people

35:21 - Why founders cannot control every outcome

Subscribe for more founder conversations over empanadas. New episodes on YouTube, Spotify, and Apple Podcasts.

Matt Narciso started his career answering angry calls at a Dish Network help desk. Thirteen years later he was running the kind of company he once took calls for. In this episode of Founders & Empanadas, he sits down in Medellín with Joshua Eidelman (his friend and direct competitor in the outsourcing world) for a conversation two rival CEOs almost never have on the record.

They get into the thing the industry avoids saying out loud: what AI actually does to a business built on billable humans. Matt breaks down the "AI Paradox" his team wrote a white paper on (everyone's using it, nobody's getting better), why he thinks it's a force multiplier that makes bad operators worse faster, and the moment in a Las Vegas parking lot that convinced him his whole industry was in trouble. Along the way: why he bet his company against its most repeatable accounts, why he chose Colombia on pure data, and what the guy on the phones knows that the CEO forgets.

Highlights:

How Matt climbed from tier-one support agent to CEO in 13 years and why BPO rewards that path

The "AI Paradox": why the work gets better while the people don't

Why AI is a force multiplier that amplifies bad operators, not a fix for them

The farming-to-tractor analogy that reframes who survives automation

The Vegas parking-lot moment that made him wean the company off repeatable work

Why he picked Colombia to scale a "purely data-driven decision"

Why the founder and the CEO can never be two different people in the human business

Subscribe for more founder conversations over empanadas. New episodes on YouTube, Spotify, and Apple Podcasts.

At 26, Yusif Gurbanli has already worked at CERN, traded at J.P. Morgan, built aircraft for NASA, and finished a PhD reconstructing human knees with AI. Most people would call that scattered. Yusif calls it the whole strategy.

Yusif is the co-founder and CFO of AIdMD, a clinical AI platform tackling one of healthcare's most stubborn problems: the hours of documentation doctors take home every night, a habit so common that physicians have a name for it — pajama time. Before AIdMD, his path ran through a robotics world championship that got him recruited by NASA, a research project designing aircraft for Mars's near-nonexistent atmosphere, and a pivot into AI-driven bone reconstruction that became the seed of his company.

In this conversation, Yusif makes the case for the generalist: why range plus the ability to go deep on demand beats narrow expertise, especially now that AI has made depth-on-demand available to anyone willing to ask the right questions.

Highlights:

How a high school robotics championship led to a NASA recruitment offer

Why he turned down Cornell for a full ride at the University of Miami

The pivot from designing aircraft for Mars to reconstructing knees with AI

What 200+ physician interviews taught him about building healthcare software

The investor lesson that changed how he sees his own startup (call it "the ugly kid")

Why real founder confidence looks like a bridge across a canyon, not blind belief

The advice from a NASA director that got him through the hardest stretch of his PhD

Follow Founders & Empanadas for more conversations with founders who took the long way around. New episodes weekly on YouTube, Spotify, and Apple Podcasts. Connect with Yusif on LinkedIn or learn more about AIdMD at aidmdusa.com.

Hilary Mason was one of the most recognizable names in data science: chief scientist at bitly, founder of Fast Forward Labs, co-author of Ethics and Data Science. Then she walked away to build Hidden Door, a roleplaying platform where you step inside licensed story worlds with an AI narrator running the game. The catch? She calls herself a skeptic of most of the generative AI market, and she's building one of its most ambitious products.

Her resolution to that contradiction is the business model itself. Authors license their worlds directly, 30% of subscription revenue flows back to creators, and Hidden Door never trains models on their work. She founded the company the week before the pandemic and spent over five years building before charging a single player.

Host Joshua Eidelman sits down with Hilary over empanadas to get into ethics as strategy, why most AI products are designed wrong, and what it takes to win over authors and game developers who distrust the entire category.

In this episode:

Why she says ChatGPT is "the original sin of bad AI product design"

How aligning your business model with your principles has to happen on day one

The decision to route 30% of revenue back to the authors whose worlds power the platform

Why she runs a strongly synchronous remote team ("demos, dilemmas, and aspirationally, donuts")

The accidental pivot from a kids' product to grown-up fans

How Hidden Door moderates open-ended roleplay without an OpenAI model in the stack

Her one belief 99% of founders would disagree with

Subscribe for new episodes of Founders & Empanadas every week.
Watch on YouTube, listen on Spotify and Apple Podcasts.

Dr. Myra Ahmad graduated from medical school in 2022, looked at the system she'd just trained inside, called it dystopian, and decided to build something outside of it instead of starting a residency. Three years later, Mochi Health has served more than 500,000 patients across all 50 states.

In this episode of Founders & Empanadas, Myra and Joshua get into the strategy behind the build: why weight-loss medication was only ever a wedge into primary care, how she designed a three-sided marketplace where the best providers rise and the weakest wash out, and why she thinks the entire incentive structure of American medicine is broken.

She also shares a contrarian take that lands well beyond healthcare: in a post-AI company, stop hiring narrow specialists and hire generalists who can move between teams, because the tooling makes picking up new skills nearly free.

Highlights:

• Why GLP-1s were a go-to-market, not the actual business
• The post-AI hiring philosophy: generalists over specialists, and roles that aren't fixed
• Why doctors never get a raise, and what that taught her about incentive design
• How she draws the line between what AI handles and what only a provider should
• Why a $250 surprise bill is a symptom of healthcare's pricing black box
• How she reads marketplace reviews when everything defaults to five stars
• Building three software products at once without starving any side of the market

Subscribe to Founders & Empanadas for new founder conversations every week.

Rand Fishkin built Moz from a struggling web-design agency into a $60 million software company and watched venture capital quietly turn it from a business that served customers into one that served investors. His verdict, over pineapple empanadas: raising VC was the second-biggest mistake of his career. The biggest? Stepping down as CEO.

Rand co-founded SparkToro after Moz and now builds "customer-shaped" companies funded by friends, not funds, including a brand-monitoring tool with no AI in its marketing and a 1960s-Italy video game. He's candid about founder depression, the gap between promised and delivered equity, and why he once wrote $700,000 in personal checks to his former employees.

This is a founder-to-founder conversation about doing it differently and why that might be the only honest way left.

What you'll take away:

• Why Rand calls stepping down as CEO his single biggest career mistake and what he'd do instead

• How VC funding shifted 80% of his energy from product to pitching

• The "true lifetime value" math VCs ignore (and how it warps what you build)

• Why he wrote $700K in personal checks when Moz finally sold

• The SparkToro clickstream data showing search still outgrowing AI

• His low-risk, friends-as-investors funding model that repaid backers in 5 years

• The belief 99% of founders would disagree with and how it changes how he runs companies

Subscribe to Founders & Empanadas for candid, founder-to-founder stories. Watch on YouTube, listen on Spotify and Apple Podcasts.