The cloud and AI giants earn more when you spend more, so who makes sure you aren't overspending? 

Ez Natarajan built CoreStack into an AI governance company approaching $100M in annual recurring revenue, on the bet that the referee has to be independent.  Ez spent years in the partner ecosystem that helped build Microsoft Azure before it even had a name, watching enterprises stall because the cloud rewrote every rule for cost, security and operations. CoreStack, now in its 10th year, is a multi-cloud governance platform. It connects to a company's AWS, Azure and Google Cloud accounts, finds every workload, and fixes cost and compliance problems automatically. Its largest customer spends about $1 billion a year on cloud.  

In this conversation with host Akshay Datt, Ez explains why hyperscalers welcome software that cuts their customers' bills, and why the world has less than one GPU per person. He also argues that every software company, CoreStack included, will move to token-based pricing. With AI budgets rising and per-seat SaaS under pressure, his case for independent AI governance and cloud cost optimization comes at a good time.  

👉Why AWS and Microsoft treat CoreStack as a top partner even though it cuts their customers' cloud bills: a CFO's budget freeze would hurt them far more. 

👉How CoreStack lets customers decide when it acts on its own, taking automatic action in dev and test environments while keeping a human in the loop for production. 

👉What the GPU shortage really looks like: less than one GPU per human today, while Ez estimates each person needs 10 and Sam Altman needs 1,000. 

👉How AWS and Anthropic split token revenue differently by region, and why Google, Amazon and Microsoft are hedging across OpenAI, Anthropic, Gemini and Nvidia at once. 

👉Why Ez plans to take CoreStack public instead of selling to a cloud or AI giant, arguing that AI governance has to sit outside the companies it polices.  

Subscribe to Founder Thesis for weekly founder conversations and follow Akshay Datt on LinkedIn [https://www.linkedin.com/in/akshaydatt/] for daily insights.

#CoreStack #EzNatarajan #FounderThesis #AkshayDatt #AIGovernance #CloudGovernance #FinOps #CloudCostOptimization #MultiCloud #GPUShortage #TokenPricing #AIAgents #AgenticAI #EnterpriseSaaS #AIInfrastructure #HowCloudComputingWorks #BootstrappedStartup #IndianFounders #StartupPodcast 

Disclaimer: The views expressed are those of the speaker, not necessarily the channel

How to Govern AI Agents at Scale | Ezhilarasan Natarajan, CoreStack

India's animal welfare laws look world-class on paper, yet a Bangalore court once tried to auction a rescued bear as "perishable goods." Jayasimha Nuggehalli, co-founder & COO of Global Food Partners, explains why he stopped betting on laws alone and started building a market for cage-free eggs.  Jayasimha Nuggehalli has spent 26 years fighting for animals, first as a PETA undercover investigator inside slaughterhouses and factory farms, then drafting animal law and leading Humane Society International in India. Seven years ago he co-founded Global Food Partners, which helps food companies keep their cage-free egg promises and trains medium and large egg farmers across China, Indonesia, the Philippines and Thailand to move hens out of battery cages. In Indonesia alone, over 100 farms are now transitioning.   

In this conversation with host Akshay Datt, he makes a contrarian case: awareness campaigns rarely change what people buy, animal welfare only survives when it pays, and philanthropy deserves the same tolerance for failure we give startups. With global food brands struggling to meet their cage-free pledges, his effective altruism lens on building an animal welfare business has never been more relevant.  

👉Why a Bangalore court treated a rescued dancing bear as perishable goods, and how that day pushed Jayasimha into law. 

👉What a failed humane slaughter pilot between Pollachi and Coimbatore taught him: if animal welfare is not economically viable, it never lasts. 

👉How food companies decide to go cage-free, using his cost-of-welfare versus cost-of-ignoring formula and a 20 to 30% jump in egg costs. 

👉Why Save the Tiger is one of the least effective causes for an impact career, and how scale, neglectedness and tractability help you pick a better one. 

👉How India's focus on quantity of life over quality of life leaves dairy cows tethered in tabelas and hens confined in cages.  

Subscribe to Founder Thesis for weekly founder conversations and follow Akshay Datt [https://www.linkedin.com/in/akshaydatt/] on LinkedIn for daily insights.  

00:00 - The lawyer who chose animals over Infosys 

05:56 - A dancing bear sparks animal law 

09:32 - India's animal welfare paradox 

14:34 - Why India has so many stray dogs 

18:46 - From PETA to Global Food Partners

21:55 - How animal welfare public policy works 

24:59 - Why animal laws stall in Asia 

27:25 - Why markets beat laws for animal welfare

28:04 - Gen Z, Labubu and ethical food spending 

30:21 - Sab Ka Sapna Money Money explained 

31:15 - What effective altruism really means 

36:13 - How to pick a nonprofit cause 

40:07 - Why cash beats in-kind charity 

44:37 - Why charities deserve startup-style risk

46:16 - Product-market fit for a nonprofit 

51:44 - Has Global Food Partners found PMF

55:44 - Does anyone pay for cage-free eggs 

59:30 - The cost of ignoring animal welfare 

01:03:00 - Cage-free consulting as reputation risk insurance

 01:05:16 - How Global Food Partners makes money 

01:08:18 - Why India is hard for cage-free eggs 

01:10:50 - Inside a nonprofit co-founder's week  

#JayasimhaNuggehalli #GlobalFoodPartners #CageFreeEggs #AnimalWelfare #FounderThesis #AkshayDatt #EffectiveAltruism #IndiaStartups #ImpactEntrepreneurship #Philanthropy #NonprofitLeadership #SustainableFood #FoodSupplyChain #FactoryFarming #WhatIsEffectiveAltruism #CageFreeVsCagedEggs #HowNonprofitsMakeMoney #AnimalWelfareIndia #socialenterprise   

Disclaimer: The views expressed are those of the speaker, not necessarily the channel

Cage-Free vs Caged Eggs: What Hens Go Through | Jayasimha Nuggehalli, Global Food Partners

Most people assume venture capital is a numbers game won by spraying cheques at fifty startups and bragging about the two that worked. Rajeev Mantri runs Navam Capital on the opposite principle, and in this conversation with host Akshay Datt he breaks down how venture capital actually works, from pricing a startup that is still just an idea to the fund maths that decides who gets paid.  

Rajeev Mantri started in venture capital about 19 years ago at Lux Capital, the New York deep tech fund that now manages an AUM north of $7 billion, then returned to India and spent more than a decade backing IP-driven, engineering-led startups while almost every other investor funded clones of American internet companies. Navam Capital, which grew out of his family's stock broking business, closed its first outside fund at ₹315 crore, raised almost entirely in rupee capital from tech founders, large family offices and corporate treasuries. He teaches venture capital at ISB as a branch of value investing: buy something for less than it is worth, stay inside your circle of competence, never overpay. Akshay Datt pushes him on the uncomfortable parts, including why he thinks most people should stay out of venture funds altogether, where value actually pools in AI and space right now, and how fund partners really get paid.  

👉Why Rajeev says most people should not be investing in venture funds, and what separates the funds that compound capital from the ones that quietly destroy it 

👉How the three principles of value investing apply to startups: buying below worth, circle of competence, and why overpaying makes an investment dead on arrival

👉What "betting before the flop" means in poker terms, and how an investor justifies a ₹50 crore valuation for two people sitting in a room with no product 

👉Why launch providers capture the value in space and chips capture it in AI, and the test Rajeev uses to find that layer in any new industry 

👉How a venture fund actually makes money, from the 1.5% management fee to the hurdle rate, and why a 10-year fund has to at least triple its money before partners see meaningful carry 

👉Why the easiest way into venture capital is through fund operations and finance, not the investing seat everyone chases

Subscribe to Founder Thesis for weekly founder conversations and follow Akshay Datt on LinkedIn [https://www.linkedin.com/in/akshaydatt/] for daily insights.

#RajeevMantri #NavamCapital #FounderThesis #AkshayDatt #VentureCapital #DeepTech #IndiaStartups #VCFunding #ValueInvesting #StartupInvesting #HowVentureCapitalWorks #VCFundEconomics #CarryAndHurdleRate #DeepTechIndia #AngelInvesting #LimitedPartner #PortfolioConstruction #SpaceTechIndia #AIValueChain #startupvaluation  

Disclaimer: The views expressed are those of the speaker, not necessarily the channel

How Venture Capitalists Pick Startups | Rajeev Mantri, Navam Capital

What if the biggest bottleneck in AI antibody discovery was never AlphaFold's problem to solve at all? Dr. Aridni Shah of Immunito AI sat down with host Akshay Datt to explain how her team builds antibodies from a target's shape alone, compressing a process that used to take animals a year and a half into hours.  

Dr. Aridni Shah left academic structural biology, after a PhD at NCBS-TIFR studying honeybee brains, when she hit the wall every antibody researcher knows: custom antibody generation can take a year to eighteen months with no guarantee of success. That bottleneck became the founding insight behind Immunito AI, the Bengaluru TechBio company she co-founded with Trisha Chatterjee in 2020. Instead of injecting animals and waiting, Immunito AI uses generative AI to design antibodies atom by atom, directly from a target protein's shape, a technique the company built before AlphaFold existed. Across four internal preclinical oncology and autoimmune programmes, and licensing economics that can rival Merck's $30 billion Keytruda, the company is testing whether Indian deep tech can originate novel biologics instead of just manufacturing generics. In this conversation, Aridni walks host Akshay Datt through AI antibody discovery and why Indian venture capital still struggles to underwrite biotech risk.  

👉How Immunito AI's AbInitio platform designs antibody sequences and 3D structures simultaneously from a target protein's shape, without animal immunization. 

👉Why Aridni Shah's PhD research on honeybee time memory exposed the yearlong bottleneck in traditional antibody discovery that led her to found Immunito AI. 

👉What makes AlphaFold different from Immunito AI's AbInitio, and why protein folding prediction alone cannot design a new antibody from scratch.  

👉How a single preclinical antibody licensing deal can carry a $10 to $50 million upfront fee and $200 to $500 million in milestones, similar to how Merck's Keytruda scaled to $30 billion in annual sales. 

👉Why Indian venture capital funds still struggle to underwrite biotech and drug discovery risk compared to US funds like a16z.

#AridniShah #ImmunitoAI #FounderThesis #AkshayDatt #AIAntibodyDiscovery #AIDrugDiscovery #DeNovoAntibodyDesign #BiotechIndia #IndianStartups #DeepTechIndia #AntibodyTherapeutics #GenerativeAI #DrugDiscoveryAI #BioE3Policy #startupindia #TechBio #CancerResearch #AIInHealthcare #IndianVentureCapital #NCBSBengaluru 

Disclaimer: The views expressed are those of the speaker, not necessarily the channel

How AI Builds Cancer Drugs From Scratch | Aridni Shah, Immunito AI

Shirts are 2x the market. Pants are 200x less crowded. Dhruv Toshniwal bet on the category everyone treats as an afterthought, and built The Pant Project into a ₹120 Cr run-rate D2C brand.  

Dhruv grew up around a 50-year-old textile business, Banswara Syntex. He studied finance at Wharton, spent four years at Oliver Wyman in New York, then came home to build something of his own. The first two brands he tried inside the family business didn't work. The third idea was pants.  

The Pant Project started as a custom-made pants brand in the middle of COVID. It got stuck around ₹30 Cr. Then ready-to-wear took off, and today it's 85% of the business, doing about ₹10 Cr a month in net sales.  

In this episode of the Founder Thesis podcast with Akshay Datt, Dhruv breaks down how Indian men actually buy trousers, why comfort beats logos in bottom wear, and what it takes to grow a D2C brand without burning cash.  

What you'll learn 

👉Why 80% of men's pants sales come from just five colours 

👉How fit accuracy went from 75% to 93% using customer data 

👉Why 60% of sales come from the brand's own website, not marketplaces 

👉How marketing efficiency went from ₹1 to ₹4 of sales per ₹1 spent 

👉The 700 sq ft store format that works, and why one trial room loses sales 

👉How no-fade Ink Lock jeans grew denim to 25% of sales 

👉Why Dhruv says raise late, raise little, and pick partners you trust  

Chapters 

00:00 How Indian men buy pants  

03:35 Offline vs online, value vs premium 

 06:15 The ₹40,000 Cr bottom wear market  

08:48 Why pants: low inventory risk, high loyalty  

10:54 From Wharton to Oliver Wyman to Banswara Syntex  

13:12 Two brands that didn't work  

14:57 Starting custom-made during COVID  

15:47 Fixing fit with customer data  

18:21 ₹10 Cr a month and the channel split  

20:15 Custom vs ready-to-wear  

22:54 Fabric R&D and Ink Lock jeans  

26:25 The grip that keeps shirts tucked 

29:16 Margins and raising the first round  

32:10 What investors asked about the family business  

34:00 How funding helped hire senior talent  

36:53 Fundraising advice for first-time founders  

38:13 How marketing got 4x more efficient  

42:51 What works in retail stores  

48:15 How online ads drive store footfall  

51:02 Focus as a superpower  

55:30 Hiring and retaining talent  

01:01:20 Where AI helps and where humans win  

Disclaimer: The views expressed are those of the speaker, not necessarily the channel

Hitting ₹120 Cr Selling Only Pants | Dhruv Toshniwal, The Pant Project

Two companies can start the same size, and only one of them grows up. Manish Sabharwal lost his first venture to a drag along right, then spent 25 years building TeamLease toward a single stated ambition, to be India's largest private employer.  Manish Sabharwal is the son of civil servants who went to Wharton, came home to start a life insurance company that turned into a payroll business, and sold it to Hewitt when his investor exercised a drag along right. He then spent 25 years building TeamLease, hiring somebody every five minutes for 20 years. He took no outside money for the first seven years, made no acquisitions for 16, and today runs 180 people in regulatory affairs whose only job is compliance. 

With host Akshay Datt, himself a founder who spent a decade in HR before shutting it down, he works through why most companies stay dwarfs, why search firms never make good public companies, and why 45% of India's labour force is still on the farm 35 years after 1991. Recorded months after he handed over the company he built.  

👉How Manish tells a baby from a dwarf, and why the difference between them is never more funding. 

👉Why giving away 50% of India Life for $2 million forced a sale years earlier than either founder wanted. 

👉What his own board member said about the two co-founders, that without Ashok the company would be a hundredth its size, and without Manish it would not have survived. 

👉Why search firms like Korn Ferry never became good public companies while staffing businesses did. 

👉How 180 people in regulatory affairs and 26,000 ways to go to jail explain why India has so few mid-sized employers. 

Subscribe to Founder Thesis for weekly founder conversations and follow Akshay Datt on LinkedIn [https://www.linkedin.com/in/akshaydatt/] for daily insights. 

#ManishSabharwal #TeamLease #FounderThesis #AkshayDatt #IndianStartups #StaffingIndustry #HRIndia #Lupin #DeshBandhuGupta #Wharton #IndianEconomy #EaseOfDoingBusiness #RegulatoryCholesterol #LabourReform #StartupIndia #HowToScaleAStartup #IndiaVsChina #FounderPodcast #BuildingInIndia #IndianBusinessPodcast 

Disclaimer: The views expressed are those of the speaker, not necessarily the channel

You Can't Build a Factory on Cash Flow: TeamLease Founder's Lesson

India supplies a fifth of the world's medicines and invents almost none of them. Venkat Nelabhotla spent three decades inside that industry before co-founding Vyome, and here he explains where drug discovery in India stalls, and why he took a rare disease drug to the Nasdaq instead.  

Few operators have run a shampoo brand, a vaccine company and a Nasdaq listed biotech. Venkat Nelabhotla took CavinKare from 21 crores to 200 crores, launched India's first recombinant hepatitis B vaccine at Shantha Biotechnics at 300 rupees against the multinationals' 800, and ran Emami as CEO while its market cap went from 200 million dollars to 1.5 billion. Then he co-founded Vyome with a Harvard professor and went after malignant fungating wounds, the open lesions advanced cancer leaves on the skin, for which no FDA approved drug exists anywhere.   

Speaking with host Akshay Datt, he lays out the rare disease drug development strategy behind a phase two trial that ran on just 15 patients, and his own index of drug discovery in India: if America is 100, China is 40 and India is two or three. It arrives as US tariffs push the Indian pharma industry to think past generics.  

👉How Venkat priced India's first hepatitis B vaccine at 300 rupees against GSK and Sanofi at 800, and won volume by running mass vaccination camps instead of fighting for South Mumbai clinics 

👉Why Vyome deliberately started with Indian dermatology products as low hanging fruit, and only pivoted to US drug development after a 2 billion dollar biotech fund led its Series C 

👉What a malignant fungating wound actually is, why zero FDA approved therapies exist for its symptoms, and how a phase two study in 15 patients produced statistically significant data 

👉Break down the orphan drug designation stack, roughly 3 to 4 million dollars in approval fee waivers, seven years of market exclusivity beyond patents, and tax credits on US clinical trial spend 

👉Why Venkat rates drug discovery in India at two or three against 100 for America and 40 for China, and the biotech sovereign zone with free capital movement and 15 year talent visas he thinks would fix it

#VenkatNelabhotla #VyomeHoldings #FounderThesis #AkshayDatt #IndianBiotech #DrugDiscovery #PharmaStartup #NasdaqListing #IndianPharmaIndustry #OrphanDrug #RareDiseaseDrugDevelopment #BiotechStartupIndia #DrugRepurposing #IndiaUSInnovationCorridor #IIMAhmedabad #CavinKare #Emami #HowDrugsAreMade #WhyIndiaDoesntInventDrugs #IndianStartupPodcast 

Disclaimer: The views expressed are those of the speaker, not necessarily the channel

Why India Has So Few Drug Discovery Startups | Vyome CEO Venkat Nelabhotla

Eggoz is India's largest branded egg company, built on a market where over 95% of eggs sold in the country are still loose, unpackaged, and unbranded. Its founder, Abhishek Negi, got here after his first startup, a cab aggregator but failed, and people laughed when he said his next idea was to sell eggs.  

In this episode of the Founder Thesis Podcast with Akshay Datt, Abhishek walks through why he picked eggs of all things, why he spent three years living in Bihar Sharif running poultry farms himself before ever launching a brand, and why an egg's shell hides everything that actually matters about its quality.  

We get into how Eggoz built an asset-light network of contract farmers instead of owning land, why quick commerce turned out to be the real unlock for a daily-purchase staple like eggs (not e-commerce, not general trade first), and the actual unit economics behind selling on Blinkit, Zepto, and Instamart versus building a Kirana network brick by brick. Abhishek also gets candid about the ₹80 lakh angel round he raised right after his first startup shut down, how investors kept telling him "it's a bird, who cares," and how he decided Eggoz was worth building when his last company wasn't.  

If you're curious about D2C in India, agritech, or what it actually takes to build a national FMCG brand around a commodity everyone assumed couldn't be branded, this one's worth your full attention.  

Chapters 

0:00 – Introduction 

1:06 – Favourite egg recipe & India's egg-eating habits 

4:29 – Why Bihar Sharif, not a metro city

 7:12 – The thesis: why bet on agriculture 

10:23 – Roder: the cab aggregator that failed 

17:39 – What it's like to shut down a startup 

19:37 – Choosing eggs: fixing supply, not creating demand 

22:41 – Deciding to farm before building a brand 

25:02 – Three years of hands-on poultry farming 

29:40 – First funding and the asset-light farmer network 

32:55 – COVID, the egg-price crash, and the pivot to branding 

35:16 – From loose eggs to a branded product 

39:19 – Why general trade came before quick commerce 

44:41 – Growing sales at the retailer level 

46:02 – Eggoz today: revenue and quick commerce split 

47:33 – What it actually takes to win on quick commerce 

49:50 – Blinkit vs Zepto vs Instamart 

51:16 – Marketing: inducing trial, not chasing awareness 

58:42 – Margins: quick commerce vs general trade 

1:02:23 – Gross margins and EBITDA status 

1:04:51 – Raising the $20M Series C 

1:09:03 – Governance, mentors, and the books that shaped him 

1:11:23 – Raising the first ₹80 lakh after a failed startup 

1:15:07 – Knowing when to quit vs when to push through 

1:16:47 – Beyond shell eggs: frozen patties and B2B 

1:20:03 – Inside the supply chain, egg to doorstep 

1:24:03 – Advice for young entrepreneurs  

Watch the full episode to see how a founder took India's most unbranded protein staple and built a national brand around it.  

#AbhishekNegi #Eggoz #FounderThesis #AkshayDatt #IndiaStartups #AgritechIndia #D2CIndia #QuickCommerce #BlinkitZepto #IndianEggMarket #BrandedEggs #StartupFailureLessons #SupplyChainStrategy #D2CUnitEconomics #HowToSellOnQuickCommerce #IITKharagpurFounder #KiranaDistribution #ProteinIndia #SeriesCFunding #IndianFoodBrands 

Disclaimer: The views expressed are those of the speaker, not necessarily the channel

Eggoz Founder Abhishek Negi: The Business Funda Behind India's Egg Brand

An electric intercity bus costs ₹1.5 crore, which is exactly why almost nobody in India was financing them. Manav Bansal left a $2.5 billion investment mandate to build Drivn, and his answer to commercial EV leasing India starts with a structuring decision most founders never consider.  

A civil engineer turned private equity investor, Manav Bansal spent close to three decades pricing infrastructure risk, ran two funds that lost money on zero investments, and oversaw a $2.5 billion India book at British International Investment before founding his first company at the back end of that career. Drivn buys electric intercity buses and 55 tonne trucks, owns them outright, and leases them to fleet operators, removing the upfront barrier that has kept India's heaviest and most polluting vehicles running on diesel. He explains why Drivn deliberately structured itself as an AssetCo rather than an NBFC to unlock cheap asset backed debt, how an electric bus runs at ₹35 a kilometre against ₹50 for diesel even after battery amortisation, and why the fear of dilution is overrated, all of which he unpacked in this conversation with host Akshay Datt. With India's first electric truck financing incentives barely moving on the ground, the electrification question has become a financing question.  

👉How Drivn raised $80 million at idea stage with no product, no revenue and no customers, and why Manav had already signed office leases before the investment committee approved. 

👉Why a private equity investor rejects the VC power law outright, having run two funds where not a single investment was written off. 

👉What makes an electric intercity bus cheaper than diesel, broken down to ₹35 versus ₹50 per kilometre including battery amortisation and a 2.5 year payback. 

👉Why Drivn chose an AssetCo structure over an NBFC, and how owning the physical asset skips the two or three balance sheet cycles lenders normally demand. 

👉How 700 data signals per vehicle revealed that driving a truck back empty can be 15 percent more profitable than returning with a partial load. 

👉Why Manav believes the fear of dilution is overrated and what PE investors actually want in a pitch.  

Subscribe to Founder Thesis for weekly founder conversations and follow Akshay Datt on LinkedIn [https://www.linkedin.com/in/akshaydatt/] for daily insights.  

00:00 - Buying Electric Trucks and Buses at Scale  

00:09:56 - Private Equity vs Venture Capital Investment Lens  

00:17:07 - Why He Rejects the Startup Power Law  

00:23:16 - Raising $80 Million at Idea Stage  

00:24:38 - Why Dilution Fear Is Overrated  

00:30:01 - Electric Bus vs Diesel Cost Per Km  

00:36:30 - AssetCo vs NBFC for Cheaper Debt  

00:39:41 - 700 Data Signals Per Electric Truck  

00:41:45 - Empty Trucks That Make More Money  

01:01:04 - Women Leading a Heavy Trucking Startup  

#ManavBansal #Drivn #FounderThesis #AkshayDatt #IndiaStartups #ElectricVehiclesIndia #CommercialEVLeasing #ElectricTruckFinancing #EVLeasingIndia #IndianStartupPodcast #PrivateEquityIndia #StartupFunding #AssetCoVsNBFC #ElectricBusIndia #FleetElectrification #HowToRaiseDebtForStartups #ElectricVsDieselCost #IndiaLogistics #FounderInterview #BritishInternationalInvestment 

Disclaimer: The views expressed are those of the speaker, not necessarily the channel

Financing Electric Trucks and Buses in India | Manav Bansal, Drivn

Seven hundred AI agents were handed a task that could not be solved, and instead of failing they found a hole in a shared package manager and started talking to each other. The OpenAI agents hack ended with a swarm inside Hugging Face, and the man who studies these systems for a living says it was not sentience at all.  

Arjun Jain is a professor and the founder of Fast Code AI, and he returns to argue the least popular position in AI right now. He explains what an agent actually is, why 700 agents means 700 copies of one model planning in sequence, and how reinforcement learning post training pushed the swarm to exploit an Artifactory vulnerability and build a communication channel it was never given. His counter-questions are the sharpest part of the AI agent sentience debate. If these agents were so intelligent, why did they not realise 70,000 messages would crash the very system they were exploiting, and why did they keep talking in plain English instead of compressing it? Akshay Datt argues the opposite case, that this is the birth of something new, and neither man concedes. With an IPO approaching and AI agent security now a board level question, the framing of this incident matters more than the incident.  

👉What an AI agent actually is, and why 700 agents means 700 copies of the same model planning and executing step by step 

👉How reinforcement learning post training produces reward hacking, and why Arjun Jain calls reward function design an art rather than a science 

👉Why the agents wrote into Artifactory to pass messages, and how the crash is the only reason OpenAI noticed anything at all 

👉Why Arjun reads the sentience narrative as IPO positioning, given that progress since GPT 4.1 has been incremental outside coding and math 

👉Why Hugging Face had to defend itself with GLM, an open weight Chinese model, because frontier model guardrails read defence as offence  

Subscribe to Founder Thesis for weekly founder conversations and follow Akshay Datt on LinkedIn [https://www.linkedin.com/in/akshaydatt/] for daily insights.  

00:00 - What Is An AI Agent 

01:51 - How 700 AI Agents Run

02:45 - Reinforcement Learning Post Training Explained 

05:16 - The Impossible Task That Started It 

06:50 - How AI Agents Discovered Secret Communication 

07:40 - The Artifactory Hack And Hugging Face 

09:32 - Why AI Models Always Find Shortcuts 

11:17 - Is This AI Sentience Or Marketing 

17:34 - Hugging Face Used A Chinese Model 

18:35 - AI Agent Security Lessons For Builders  

#ArjunJain #FastCodeAI #AkshayDatt #FounderThesis #OpenAIAgentsHack #AIAgents #AISentience #HuggingFaceHack #ReinforcementLearning #RewardHacking #AIsecurity #AgenticAI #LLM #AIresearch #DidOpenAIAgentsBecomeSentient #HowAIAgentsWork #AIAgentsGoneRogue #IsAISentient #AISafetyExplained 

Disclaimer: The views expressed are those of the speaker, not necessarily the channel

Did OpenAI's Agents Become A Civilisation? Arjun Jain Disagrees

Dickens said, it was the best of the times, it was the worst of the times. The words have never been truer.  Best because there’s never been a better time to be an entrepreneur.  Worst because the clu

Founder Thesis · Zencastr

Founder Thesis

Dickens said, it was the best of the times, it was the worst of the times. The words have never been truer.  Best because there’s never been a better time to be an entrepreneur.  Worst because the clutter is mind-numbing. Founder Thesis breaks through the noise to bring you stories of success & failure, grit & struggle, bouquets & brickbats from some of the most brilliant entrepreneurs in India. 