Intrepid Growth Partners' Substack

Managing AI Platforms for Healthcare (ep 27)

Intrepid Growth Partners — Tue, 28 Apr 2026 15:57:47 GMT

When the right answer is a matter of life and death, how do you ensure that your AI systems avoid fatal mistakes? Signal 1 is a software platform for healthcare that functions as a control panel for deployed AI systems in hospitals. Rather than building individual models, Signal 1 focuses on improving the safety, observability and governability of pre-existing AI models used throughout the patient experience. The platform enables health systems to track real-world performance, detect drift and risk, enforce approval workflows, and tie AI predictions to improved clinical and operational outcomes.

Tomi Poutanen is Signal 1’s CEO and co-founder. His earlier start-up, Layer 6, employed artificial intelligence to provide financial services companies with predictive analytics, and was acquired by TD Bank, where Tomi served as chief AI officer. In this episode, Tomi and the Derby Mill team discuss the future of healthcare, whether machine learning could make hospitals obsolete, and how to improve the management of systems that include numerous different AI agents working together.

GUESTS AND HOSTS
Tomi Poutanen, CEO, Signal 1
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist and CEO, Emergent Platforms
Suzanne Gildert, CEO, Nirvanic Consciousness Technologies, quantum physicist, co-founder of Sanctuary AI

LINKS
Signal 1 website
Media on Signal 1 from the Globe and Mail, Betakit and the University of Toronto
Derby Mill series website. Derby Mill is created by the team at Intrepid Growth Partners and produced by Ghost Bureau.
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms: YouTube // Spotify // Apple Podcasts // Substack

DISCUSSION POINTS
00:00 Cold open
00:44 Ajay’s tee up
01:22 About Signal 1
04:11 Tomi’s Signal 1 explainer
06:25 Mechanics of Signal 1
07:10 Tomi’s first-order question
08:45 Signal 1’s core value prop
09:29 Patient experience perspective
12:27 Taking MD out of the loop
13:56 Why Signal 1 focuses on AI
18:20 Evaluating AI systems
22:46 Rows
23:13 Ajay’s segue
24:42 Part 2: Signal 1 at the Limit
25:12 Niamh’s sci-fi hospital vision
29:04 Rich questions the premise
31:25 Sendhil: Division of labor for AI
34:50 Tomi: Niamh’s vision is real
38:22 Tomi: Competitive healthcare
43:55 Sendhil: Virtues of randomness
49:44 Niamh: Multiple agent drama
52:54 Tomi’s final remarks

NUGGETS
How Signal 1 evaluates AI algorithms
Sendhil Mullainathan asks Signal 1 CEO Tomi Poutanen how his company assesses quality and function of large language models and agentic AIs.

DISCLAIMER
The content of this podcast is for informational and educational purposes only and should not be construed as marketing, solicitation, or an offer to buy or sell any securities or investments. The opinions expressed in this video are those of the participants and do not necessarily reflect the views of Intrepid Growth Partners or its affiliates. Any discussion of specific companies, technologies, or industries is for illustrative purposes and does not constitute investment advice. Viewers are encouraged to consult with their own financial, legal, and tax advisors before making any investment decisions.

How Will OpenClaw Change Agentic AI? (ep 26)

Intrepid Growth Partners — Tue, 14 Apr 2026 15:46:07 GMT

OpenClaw, the latest AI chatbot to go viral, has some key differences from ChatGPT, Claude or Gemini. The agentic AI developed by Peter Steinberger is open source and hosted on your own computer. You also can talk to it via messaging apps like WhatsApp. So how does OpenClaw work? What are the implications for the AI industry overall? How will it affect the adoption of artificial general intelligence, or household robots? And what are the security concerns that OpenClaw may trigger? Host Ajay Agrawal explores these issues and more with the Derby Mill team.

GUESTS AND HOSTS
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist and CEO, Emergent Platforms
Suzanne Gildert, CEO, Nirvanic Consciousness Technologies, quantum physicist, co-founder of Sanctuary AI

LINKS
Y Combinator interviews OpenClaw creator Peter Steinberger
OpenClaw explained
Subscribe to The Derby Mill Series on YouTube, Spotify or Apple Podcasts
Derby Mill is created by the team at Intrepid Growth Partners and produced by Ghost Bureau.

DISCUSSION POINTS
00:00 Cold open
01:05 What is OpenClaw?
02:08 Difference from chatbots
03:44 OpenClaw’s abilities (Niamh)
04:49 WhatsApp interaction
06:14 Sendhil Mullainathan’s take
09:39 Suzanne Gildert’s take
12:50 Rich Sutton’s take
14:44 OpenClaw security concern
17:27 Niamh on OpenClaw implications
19:05 Final implications: Sendhil

NUGGETS

Will OpenClaw help make workflows more efficient? (2601)
MIT economics professor Sendhil Mullainathan weighs in

OpenClaw and the advent of artificial general intelligence (2602)
Turing Award winner Rich Sutton’s take

How OpenClaw will affect physical AI (2603)
According to robotics expert Suzanne Gildert

Big problem for OpenClaw early adopters (2604)
A warning from AI expert Niamh Gavin

DISCLAIMER

The content of this podcast is for informational and educational purposes only and should not be construed as marketing, solicitation, or an offer to buy or sell any securities or investments. The opinions expressed in this video are those of the participants and do not necessarily reflect the views of Intrepid Growth Partners or its affiliates. Any discussion of specific companies, technologies, or industries is for illustrative purposes and does not constitute investment advice. Viewers are encouraged to consult with their own financial, legal, and tax advisors before making any investment decisions.

Rethinking Humanoid Robots (The Derby Mill Series ep 25)

Intrepid Growth Partners — Tue, 07 Apr 2026 15:02:57 GMT

Are humanoid robots using the best possible form factor, or should we consider a wholesale redesign if we’re seeking the most useful mechanical helpers for daily living? Drawing on recent demonstrations at CES, China’s Spring Festival Gala and the India AI Impact Summit, the Derby Mill team explores the implications of ever-advancing robotics capabilities.

Ajay Agrawal and collaborators Rich Sutton, Sendhil Mullainathan, Niamh Gavin and Suzanne Gildert explore public hesitancy around in-home robots. They explain why dexterity and reliability in everyday settings remain unsolved problems, and discuss the technical realities of robot hands. Why is learning from trial and error so essential to advance the field? Plus: What’s with the obsession with human-like bodies? What about radically different robot forms inspired by nature, like the octopus?

LINKS
Snapshots on current-state robotics: The humanoid robots of the Chinese New Year. Recap of India’s AI Impact Summit. Every humanoid robot at CES 2026. Best robots at CES 2026.

Subscribe to The Derby Mill Series at our Substack (main site) or on YouTube, Spotify or Apple Podcasts.

Derby Mill is created by the team at Intrepid Growth Partners and produced by Ghost Bureau.

DISCUSSION POINTS
00:00 Current perceptions and evolving expectations for the future of robotics
00:44 Highlights from global robotics summits
01:45 Market penetration and the commercial realities of emerging robot types
02:58 Consumer sentiment and safety concerns regarding robotics in domestic environments
04:15 Sector-specific applications for robots in industrial, data centre and military settings
05:40 Roadblocks to general-purpose utility and the timeline for home adoption
06:25 Shifting from humanoid to specialized robotic designs in factories and warehouses
07:06 Technical limitations of robotic dexterity compared to human fine motor control
08:28 Mechanical hand design: tendon and motor placement trade-offs
09:25 The software bottleneck and the necessity of trial-and-error learning from experience
10:21 De-centering the human form factor in the exploration of robotic physicality
11:41 Infrastructure limits and the anthropomorphic design debate in human environments

NUGGETS

What’s Taking So Long for General Purpose Robots to Go Mainstream? (2501)
Suzanne Gildert, Rich Sutton and the Derby Mill team weigh in.

DISCLAIMER

Embedded Cybersecurity (The Derby Mill Series ep 24)

Intrepid Growth Partners — Tue, 24 Mar 2026 11:03:28 GMT

AI-enabled cyberattacks are increasing in sophistication and creativity. For example, recently, a vending machine became a potential entry point for attackers targeting a large financial institution in a major city. So how can manufacturers protect devices in the age of AI and LLMs? How can that effort be helped by open source software and hardware? And what does this mean for the future of connected devices?

All this and more is on the agenda in the latest episode of the Derby Mill Series, which sees our usual team of Ajay Agrawal, Rich Sutton, Sendhil Mullainathan, Niamh Gavin and Suzanne Gildert joined by Intrepid growth partner Ken Nickerson and our special guest, Gianni Cuozzo from Exein, as they explore embedded cybersecurity and how it protects billions of connected devices around the world.

GUESTS AND HOSTS

Gianni Cuozzo, founder and CEO, Exein
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist and CEO, Emergent Platforms
Suzanne Gildert, founder and CEO, Nirvanic Consciousness Technologies
Ken Nickerson, senior advisor, Intrepid Growth Partners, and founder and CEO, iBinary

LINKS

Exein website
Exein explanation video
Exein raises €100m in new funding to accelerate its global expansion
Mentioned in the episode: Read Anthropic’s report: Disrupting the first reported AI-orchestrated cyber espionage campaign
Subscribe to The Derby Mill Series at our Substack (main site) or on YouTube, Spotify or Apple Podcasts
Derby Mill is created by the team at Intrepid Growth Partners and produced by Ghost Bureau.

DISCUSSION POINTS

00:00 — What is embedded cybersecurity?
01:23 — How IoT devices become attack vectors
02:20 — The vending machine breach at a major bank
06:00 — Lateral movement and malware explained
08:53 — Gianni’s background in cyber warfare
11:05 — What is embedded cybersecurity?
13:49 — What makes Exein unique in cybersecurity
16:06 — Reinforcement learning and hacking
17:29 — Physical risks in robotics
19:24 — Pattern recognition vs expected behaviour
25:56 — Historical lessons from computing vulnerabilities
31:05 — AI and cybersecurity at the limit
32:13 — Future equilibrium: open hardware and software
36:44 — Scaling attacks and defence in cyber-physical systems

NUGGETS

What Is Embedded Cybersecurity? (2401)
Exein CEO explains how devices stop cyberattacks in real time

The Problem with Modern Hardware Security (2402)
Why your computer isn’t really under your control

Is AI Making Cyberattacks Easier? (2403)
How AI cuts reverse engineering from months to minutes.

DISCLAIMER

Is the AI-Led Software Selloff Justified? (The Derby Mill Series ep 23)

Intrepid Growth Partners — Fri, 20 Feb 2026 10:01:59 GMT

Nearly a trillion dollars in market value vanished from software and services stocks after a Reuters headline framed AI as a potential existential threat to traditional software companies. In this episode of the Derby Mill Series, host Ajay Agrawal leads Rich Sutton, Niamh Gavin, and Suzanne Gildert in a focused discussion on whether that reaction reflects structural risk or market overreach.

Questions discussed:

—How much risk do traditional software firms face from AI?
—Was the AI-led selloff in software stocks justified?
—How will Gen AI change product cycles and the way organizations run?
—At the limit, will the idea of a company, with leadership and employees, even make sense anymore?
—How will the incorporation of AI technology into human physiology affect evolution?
—Are we heading toward a speciation event that see new classes of humans emerge?

GUESTS AND HOSTS

Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist and CEO, Emergent Platforms
Suzanne Gildert, founder and CEO, Nirvanic Consciousness Technologies

LINKS

Selloff wipes out nearly $1 trillion from software and services stocks as investors debate AI’s existential threat
Threat of New AI Tools Wipes $300 Billion Off Software and Data Stocks
Software Selloff Hits Industry’s Billionaires
The Software Rout Is Spreading Pain to the Debt Markets
Mentioned in the episode: The Innovator’s Dilemma
Subscribe to The Derby Mill Series at our Substack (main site) or on YouTube, Spotify or Apple Podcasts
Derby Mill is created by the team at Intrepid Growth Partners and produced by Ghost Bureau.

NUGGETS

Is the AI-led software selloff justified? (2301)
Three experts weigh in on risk, adaptability, and the future of traditional companies.

DISCUSSION POINTS
00:00 Introduction to software selloff and AI discussion
02:38 Niamh on AI-native approaches and incumbents
06:35 Wall Street reaction and software risk
08:37 Innovator’s dilemma and challenges
10:39 Data moats and Gen AI product cycles
11:47 Foundation models capturing end-user workflows
14:10 Rich on legacy systems and moats collapsing
16:29 Summary: selloff directionally correct
17:31 Suzanne on AI-human hybrids and future work
20:12 Speciation, societal impacts, and UBI discussion
24:41 Niamh and Rich on company adaptation
26:06 Episode wrap-up and key takeaways

DISCLAIMER

OpenAI & Anthropic Healthcare Announcements

Intrepid Growth Partners — Fri, 06 Feb 2026 10:42:14 GMT

As everyone has likely seen, OpenAI and Anthropic announced health-focused product releases in January. Below, we’ve summarized what was announced, what is actually new versus incremental, how ChatGPT Health and Claude Health are positioned differently, and the potential implications for healthcare AI more broadly.

1. Adoption & Baseline Reality

Health and wellness inquiries are already one of the dominant use cases for large language models, particularly ChatGPT. More than 230 million people globally ask health- or wellness-related questions each week, with ~40 million users per day engaging ChatGPT on health topics. In practice, LLMs are already being used as a first line of inquiry, an informal second opinion, and a pre-triage step before individuals interact with the healthcare system. This adoption has occurred organically and is consumer-led, rather than driven by formal deployment through health systems or clinicians. Patients are using LLMs to interpret symptoms, understand diagnoses, sanity-check treatment plans, and decide whether and when to seek care, regardless of whether these tools are clinically validated or formally regulated for such use cases. The release of ChatGPT Health and Claude Health are not introducing AI into healthcare but formalizing and productizing behavior that already exists at scale.

2. What’s Actually New in the January 2026 Announcements

The January 2026 announcements from OpenAI and Anthropic primarily relate to product structure, data integrations, and privacy framing, rather than changes in underlying model capability.

ChatGPT Health

Introduces a dedicated health space within ChatGPT, separate from general conversations.
Health interactions within this space are positioned with enhanced privacy protections, including a stated commitment that data shared in ChatGPT Health will not be used to train foundation models.
Supports ingestion of personal health information, including medical records and connections to consumer health and wellness data sources such as Apple Health, Function Health, and MyFitnessPal, as well as access to longitudinal medical records via b.well Connected Health, which aggregates EHR data using FHIR-based interoperability across U.S. providers.
Continues to rely on the same general-purpose LLMs, with health-related performance evaluated using HealthBench, a physician-designed benchmark developed over a two-year period with input from more than 260 physicians across 60 countries and dozens of medical specialties. HealthBench is built around 5,000 realistic health conversations, each graded using clinician-written rubrics focused on safety, clarity, appropriate escalation of care, and use of patient context.

Claude Health

Does not introduce a separate health-dedicated interface or a health-specific model.
Supports ingestion of medical records and connections to consumer health and wellness data sources, including Apple Health and Function Health, and allows users to connect EMR/EHR data via HealthEx, which aggregates records across tens of thousands of providers using national interoperability infrastructure.
In addition to personal health data, Claude emphasizes integration with structured medical reference frameworks, including standardized clinical taxonomies (e.g., ICD codes) and CMS-related reference data.
Claude’s health-related performance is evaluated using benchmarks such as MedAgentBench and MedCalc, which assess the model’s ability to work with structured patient information and correctly reason through common medical calculations and metrics.

Across both products, the substantive change is not new reasoning capability, but clearer scoping of health data, more explicit health-focused integrations, and (in ChatGPT’s case) a formal privacy and UX boundary around health use. These releases consolidate health use cases that already existed into more structured, governed product experiences without crossing into diagnosis, treatment recommendation, or autonomous clinical decision-making.

Positioning: Consumer vs. Enterprise

While both OpenAI and Anthropic support similar underlying health data integrations, their go-to-market focus around health differ. ChatGPT Health is positioned primarily as a consumer-facing health assistant. It builds on ChatGPT’s large existing user base and emphasizes individual use cases such as understanding symptoms, interpreting personal records, and incorporating wellness data. The dedicated health space and explicit privacy framing are designed to build trust with individual users already engaging with the product for health-related questions at scale.

Claude Health, by contrast, is positioned more clearly toward enterprise and health system adoption. Anthropic’s messaging, early deployments, and evaluation approach emphasize use cases relevant to clinicians and healthcare organizations, including chart summarization, care coordination, prior authorization, and administrative workflows. This orientation is reflected in Claude’s benchmarking and integrations: its health-related benchmarks focus on provider-adjacent tasks such as answering questions about a patient’s chart, reasoning over structured patient information, and calculating common medical metrics, rather than consumer conversational quality alone. Claude also places greater emphasis on integration with structured medical reference frameworks (e.g., ICD clinical taxonomies, CMS-related reference data) relevant to providers and healthcare organizations. Overall, the distinction is less about model capability and more about who the product is built for and how success is measured: ChatGPT Health centers the individual patient or consumer, while Claude Health centers healthcare organizations and provider-facing tasks.

3. Regulatory Posture: Why These Tools Remain Largely Unregulated

Despite their increasing use in health-related contexts, ChatGPT Health and Claude Health are currently positioned outside traditional medical device regulation. This is not because regulators are unaware of these tools, but because of how they are framed and what they explicitly do not claim to do.

Regulators such as the FDA, Health Canada, and EU authorities do not regulate “AI” in the abstract; they regulate medical devices, defined by intended use. Tools typically require clinical validation and regulatory clearance if they diagnose conditions, recommend treatments or dosing, replace clinician judgment, or make claims about clinical outcomes. Both ChatGPT Health and Claude Health are framed as general-purpose health information and support tools.

While both products include disclaimers stating they are not intended for diagnosis or treatment, they are designed to deliver confident, personalized, and highly engaging responses. This positioning requires a careful balance: disclaimers frame the tools as non-clinical, while product design optimizes for engagement and perceived usefulness, increasing the risk that users place more trust in the outputs than their formal positioning would suggest. This tension, between legal framing and user perception, underscores the challenge of operating in health-related domains without crossing into regulated clinical use.

Regulation may also play a more active role in shaping where health AI adoption occurs. While general-purpose tools like ChatGPT Health attract consumers organically, governments and regulators can explicitly endorse or deploy alternative platforms for regulated use cases, steering patients and providers toward approved pathways. Early examples, such as state-level partnerships enabling AI-assisted prescription renewal, illustrate how public-sector adoption can coexist alongside consumer AI rather than being displaced by it.

4. Benefits of Continued Adoption

ChatGPT Health and Claude Health address practical gaps in today’s healthcare system around access, availability, and understanding. For many users, these tools offer immediate, low-cost guidance that complements traditional care.

Key benefits

Bridges access gaps: Users turn to LLMs when appointments are hard to get, wait times are long, or follow-up questions go unanswered.
Always available: 24/7 access at little to no cost, compared with traditional healthcare touchpoints.
Second-opinion utility: Most effective as a preparatory or sanity-check layer, helping users understand symptoms, interpret records, and engage more productively with clinicians.
Higher perceived empathy: Conversational interfaces can make users feel heard and supported, particularly when navigating anxiety, uncertainty, or complex information.
Greater personalization: The ability to reference individual context (e.g., uploaded records or connected health data) allows responses to be tailored in a way that static content or traditional tools often cannot.
Enterprise productivity: For health systems, tools like Claude Health target administrative and clinical-adjacent workflows (e.g., chart summarization, care coordination), where incremental efficiency gains can materially reduce clinician burden.
AI normalization: Familiarity at the consumer and enterprise level lowers resistance to broader adoption of AI-assisted healthcare tools over time.

These tools should be evaluated against the healthcare system as it exists today, not against an idealized standard. In many cases, the realistic alternative to AI-assisted guidance is delayed care, incomplete information, or no guidance at all. Notably, algorithmic errors tend to attract disproportionate scrutiny relative to human error, even when average performance exceeds existing baselines, a dynamic that can slow the adoption of technologies with the potential to materially improve access and efficiency.

5. Risks & Limitations

Despite clear benefits, the use of general-purpose LLMs in health contexts carries meaningful risks, particularly as adoption deepens and personalization increases.

Key risks

Over-trust by users: Confident, fluent responses and personalized context can lead users to treat outputs as authoritative, even when disclaimers state the tools are not intended for diagnosis or treatment. Publicly reported incidents illustrate that this risk is not theoretical: when incorrect or unsafe guidance from LLMs is over-trusted (particularly in acute or high-risk situations), it can conflict with clinical judgment and contribute to patient harm.
Lack of clinical evidence: These tools have not been validated through clinical trials or prospective studies demonstrating improved patient outcomes or safety in real-world use.
Hallucination risk: General-purpose LLMs can generate plausible but incorrect medical information, particularly in edge cases or less common conditions.
Safeguard degradation over long interactions: Safety mechanisms tend to work more reliably in short, common exchanges and can become less reliable in long, multi-turn conversations. For example, models may appropriately escalate to crisis resources early in an interaction but later provide responses that contradict earlier safeguards.
Evaluation gaps: Benchmarks such as HealthBench, MedAgentBench, and MedCalc focus largely on text-based and do not fully capture real-world complexity or evaluate performance on multimodal inputs such as images.
Language and population bias: Most health benchmarking and evaluation has been conducted in English, raising concerns about performance and safety across different languages.
Unclear accountability: Responsibility for harm remains ambiguous when outputs are framed as informational but influence real health decisions, creating liability and trust challenges for developers, health systems, and users.

Taken together, these risks reflect the gap between rapid adoption and the slower development of evidence, safeguards, and accountability frameworks appropriate for health-related decision support.

6. Implications for the Health AI Landscape

ChatGPT Health and Claude Health do not make most healthcare AI categories obsolete, but they do change which parts of the stack face the greatest competitive pressure.

Likely pressured

Generic symptom checkers and triage chatbots
(e.g., WebMD Symptom Checker)
Low-moat “AI front doors” and intake/navigation layers without deep workflow or regulatory insulation
(e.g., early-stage conversational intake tools)
Standalone patient engagement and messaging assistants primarily focused on reminders, routing, and education
(particularly where differentiation is limited to UX)

These products increasingly compete against horizontal platforms with massive distribution, strong trust signals, and improving personalization.

More insulated / advantaged

Regulated clinical AI with evidence and clearance
(e.g., Skin Analytics, HeartFlow, Cleerly, Viz.ai, Aidoc)
Workflow-embedded enterprise AI, particularly tools natively integrated into core EHRs
(e.g., Nuance / Microsoft DAX, Epic-native AI modules)
Verticalized health AI with proprietary data and clear ROI
(e.g., Iterative Health in clinical trials, Tempus in oncology)

ChatGPT Health and Claude Health operate as horizontal health intelligence layers that power conversational interfaces across consumer and enterprise use cases. They commoditize low-differentiation conversation while increasing the importance of evidence quality, workflow integration, proprietary data, and regulatory alignment in healthcare AI.

7. Conclusion & Key Takeaways

The launch of ChatGPT Health and Claude Health reflects how far large language models have already penetrated healthcare decision-making, rather than a sudden inflection in technical capability. These products formalize behavior that already exists at scale and make it easier for both consumers and enterprises to engage with health-related information through conversational interfaces. Low-differentiation conversational layers, particularly symptom checkers, intake chatbots, and standalone engagement tools, face increasing competitive pressure. By contrast, durable value in healthcare AI will continue to accrue to companies that combine strong UX with evidence, deep integration into systems of record, proprietary data and alignment with regulatory and operational realities, rather than to standalone conversational intelligence alone. The central tension is not whether LLMs belong in healthcare (they are already there) but how they are governed, integrated, and trusted. For now, both OpenAI and Anthropic are walking a narrow line: offering highly capable, personalized health assistance while avoiding explicit clinical claims that would trigger regulatory oversight.

Personal Finance and AI with Cleo founder and CEO Barney Hussey-Yeo (The Derby Mill Series ep 22)

Intrepid Growth Partners — Thu, 05 Feb 2026 11:02:36 GMT

Barney Hussey-Yeo is the founder and CEO of Cleo, the company behind an AI-powered financial companion that is transforming the way millions of people manage their money. Founded in London in 2016, the company has an annual recurring revenue of $350 million. It has doubled its subscriber base every year since 2021, and now has more than 300 employees in offices in London, New York and San Francisco.

Here, host Ajay Agrawal and the Derby Mill panel of Rich Sutton, Sendhil Mullainathan and Niamh Gavin brainstorm with Barney Hussey-Yeo on where Cleo may go, at the limit. How can the agentic AI help its subscribers avoid costly mistakes while offering advice that is both reliable and tailored to the individual? What challenges arise when working with incomplete, inconsistent financial data? And how will Cleo’s machine learning model transform the way regular people manage their finances? Barney, Ajay and the Derby Mill team discuss it all—and more.

GUESTS AND HOSTS
Barney Hussey-Yeo, founder and CEO, Cleo
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist and CEO, Emergent Platforms

LINKS
Subscribe to The Derby Mill Series at our Substack (main site) or on YouTube, Spotify or Apple Podcasts
Watch Cleo’s one-minute demo video. And here’s the Cleo website.
Derby Mill is created by the team at Intrepid Growth Partners and produced by Ghost Bureau.

DISCUSSION POINTS
00:00 Future of financial products and AI in banking
01:30 What is the Cleo AI personal finance assistant
04:04 Cleo growth active users and scaling an AI startup
05:41 How Cleo builds trust during onboarding
07:30 Using large language models for personal finance AI
12:04 Optimizing financial health with AI and Cleo
18:33 How conversational AI captures user context in finance
25:49 Cleo monetization strategy subscriptions and financial products
32:18 Pushing AI to the limit: next-gen finance applications
33:40 Holistic financial management with AI guidance
36:33 AI helping users achieve financial goals
44:41 Future of finance AI and market transformation

NUGGETS

Cleo at the Limit (2201)
AI pioneer Rich Sutton's take on the AI-enabled personal finance firm.

How Agentic AI Will Transform Personal Finance (2202)
According to MIT economist Sendhil Mullainathan

How to Monetize Personal Finance Advice? (2203)
MIT economist Sendhil Mullainathan asks Cleo founder Barney Hussey-Yeo

DISCLAIMER

AI and eCommerce with Shopify CEO Tobias Lütke (The Derby Mill Series ep 21)

Intrepid Growth Partners — Thu, 18 Dec 2025 21:01:17 GMT

The Derby Mill regulars host Shopify CEO Tobias Lütke on the heels of the ecommerce giant’s release of its Winter ‘26 Edition, aka The RenAIssance Edition, a significant artificial intelligence-enabled refresh of the company’s products and services. With more than 150 new and updated products, the update aims to help entrepreneurs, merchants and small businesses use AI to amplify their human creativity.

In one of the first conversations to happen with Tobi Lütke after the Shopify update, AI legends Rich Sutton, Sendhil Mullainathan, Niamh Gavin and Suzanne Gildert join Intrepid’s Ajay Agrawal to examine where artificial intelligence, machine learning and reinforcement learning may take ecommerce at the limit. How can AI help ecommerce merchants? What can machine learning do for small business? Could Shopify Sidekick’s agentic AI help merchants optimize their path to profitability? And will Shopify SimGym empower small businesses with the testing capability of much larger companies? It’s all on the agenda, and more, in our latest episode.

About Shopify CEO and co-founder Tobias Lütke:

Tobias Lütke is CEO and co-founder of Shopify, the marquee shopping cart system of the e-commerce industry, which he co-founded in 2006 after encountering difficulties trying to create an online snowboard retailer. Today, the company has a market capitalization of $210 billion USD, with customers in 175 countries around the world. FY2024 revenue was $8.88 billion US and transactions on the Shopify platform can amount to 10% of all US commerce.

GUESTS AND HOSTS

Tobias Lütke, CEO and co-founder, Shopify
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist and CEO, Emergent Platforms
Suzanne Gildert, founder and CEO, Nirvanic Consciousness Technologies

LINKS

Subscribe to The Derby Mill Series at our Substack (main site) or on YouTube, Spotify or Apple Podcasts.
Shopify’s Winter ‘26 Edition presentation and summary press release.
Mentioned in the pod: Susan Athey’s co-written journal paper is Artificial Intelligence, Competition, and Welfare, published by the National Bureau of Economic Research.
Derby Mill is created by the team at Intrepid Growth Partners and produced by Ghost Bureau.

DISCUSSION POINTS

00:00: Cold open with Shopify CEO Tobi Lütke saying, the goal is not to be the most powerful AI company, but to make AI gifts from labs maximally valuable to people.

01:01: Guest introductions, including Tobi Lütke, CEO of Shopify; Turing award winner Rich Sutton, who pioneered reinforcement learning; MacArthur Genius recipient Sendhil Mullainathan; applied AI scientist Niamh Gavin; and robotics and AI expert Suzanne Gildert.

01:50: Tobi discusses Shopify’s scale—operating close to six million storefronts and serving close to a billion customers purchasing about $30 billion a month in gross merchandise value.

03:17: Shopify as a counter-example to machine intelligence amplifying the power of large companies, instead using it to significantly boost smaller companies.

04:28: Toby Lütke provides an overview of Shopify, a Canadian company started 20 years ago that powers millions of merchants, often the websites customers buy from if it’s not Amazon.

06:57: Introduction of the three specific AI applications to be discussed, starting with “SimGym” for launching with confidence without real consumer testing.

07:38: Description of SimGym, a simulator with AI shoppers that predict customer behaviour and reflect the archetype of a merchant’s customers.

08:51: Discussion on the data backbone for SimGym’s personalized prediction, which includes transactional history, browsing behaviour matched to personas through standard clustering, and demographics from deliveries.

10:30: The goal of SimGym is to help small businesses get to conviction faster with their testing, as traditional AB testing takes a very long time for them.

12:58: Sendhil Mullainathan discusses how Shopify deploys scale economies to artisan producers, providing small businesses with data to make consequential decisions.

14:09: Introduction of “Sidekick,” Shopify’s agentic co-pilot, and the feature “Sidekick Pulse,” which delivers insights based on a store’s data, economic trends, and Shopify’s commerce knowledge, serving the non-sophisticated, time-and-money-constrained entrepreneur.

15:55: Sidekick is described as an assistive technology that automates tasks, finds factors to benchmark a business’s success, and provides insights in a human-like way, contrasting with the “very autistic” nature of typical software.

18:02: Shopify as a bridge between incredible research and the global network of commerce, bringing valuable morsels back to “entrepreneurship land”.

19:31: How merchants complained when Sidekick was temporarily taken down, with some referring to it as their “employee of the month”.

22:11: Shopify’s business model is fully aligned with customers; it does not charge for services like Sidekick because it benefits from bigger businesses, allowing the value of the AI to be absorbed in the existing model.

24:24: “Shop slop”—the concern that fully automated store production and drop shipping might push out small business owners.

25:48: Tobi Lütke argues that e-commerce is different from content generation because it has two governors: atoms must be assembled, and a transaction involves money, which is a rivalrous resource, meaning a purchase validates the value.

28:02: Rich Sutton asks how SimGym works and how it can be better than a shop owner’s intuition. Tobi Lütke’s response explains that it involves parameterized agents using a vision model browser loop to browse the website.

30:44: Discussion of Shopify’s advantage in having end-goal data (the sale) for its Reinforcement Learning (RL) system, providing true ground truth for the goal.

34:10: Ajay speculates that Shopify may become the most powerful AI company due to its access to vast data, the end goal (sale), and the large number of independent merchants, enabling a high degree of experimentation crucial for RL.

36:39: Introduction of “Shopify Product Network,” which uses machine intelligence to fill in product gaps for small merchants, like a skateboard store selling compatible helmets, thereby removing a scale economy disadvantage.

39:18: Introduction of the third AI product, “Sidekick Pulse,” which provides “next best action” predictions to merchant owners, advising on the most ROI- or sales-increasing action to take.

40:57: Niamh Gavin’s vision is that this technology enables a new age of affordable mass personalization by levelling the playing field for merchants and leveraging the community in a win-win network effect.

41:41: Suzanne Gildert questions the long-term objective function, asking if optimizing only for purchase volume could lead to a “dopamine addiction system” and suggests including consumer happiness.

42:34: Sendhil Mullainathan presents a positive future vision where Shopify’s architecture pushes AI in a different, decentralized direction, focusing on innovations that decision-makers (small merchants) find helpful.

44:32: Clarification of the two AI trajectories: autonomous decision-making (large organizations) versus human-machine “centaur” optimization (Shopify’s small merchants), where local information and the shop owner’s power are key.

47:40: The discussion notes that the centaur model would require a different set of performance benchmarks, focusing on improving human performance aided by AI.

49:54: Sendhil Mullainathan compares autonomous coding to co-pilots, suggesting that the centaur model focuses on making AI errors more transparent to humans and optimizing for diversity/variance rather than correctness.

53:38: Tobi Lütke reiterates that Sidekick and the other products function as “assistive technology with human in the loop,” aligning with the philosophical view that computers should work for humans and handle computing/data transfers.

56:08: Mention of the HSTU architecture, developed with Liquid AI and Nvidia, which has been “extremely game-changing.”

57:23: Rich Sutton discusses the limits of e-commerce, questioning whether decentralization should be around the merchant or the customer, suggesting a future where Shopify supports both.

01:02:30: The question is raised: what new and surprising thing will make online commerce different a year from now.

01:03:13: Niamh Gavin predicts that Sidekick Pulse’s ability to generate insights and automatically execute next best actions (like drafting a win-back email) will be the most surprising change in online commerce.

01:04:39: Suzanne Gildert expresses interest in consumers delegating agency to their own AI assistants, which could use simulation tools like SimGym to make buying choices from artisan merchants.

NUGGETS

Should AI Optimize for Correctness or Variance? (2101)

MacArthur Genius Sendhil Mullainathan to Shopify CEO Tobi Lütke: Should AI Optimize for Correctness or Variance?

The Most Powerful AI in the World (2102)

Could Shopify’s Winter ‘26 Edition Make It the World’s Most Powerful AI Company? Ajay Agrawal, Rich Sutton and Tobi Lütke discuss.

DISCLAIMER

Canada's AI Advantage (The Derby Mill Series ep 20)

Intrepid Growth Partners — Wed, 19 Nov 2025 16:00:30 GMT

Does the Canadian AI community have a communications problem? Too often, AI investors feel they have to go outside of the country to find great targets for deals. Similarly, domestic AI companies find it difficult to attract dollars from Canadian sources of capital. Too few investors and companies actually talk to one another. And fewer still have the kind of trusted relationship required to get deals done.

So in this episode, Derby Mill host Ajay Agrawal, a co-founder and partner at Intrepid Growth Partners, gathers some of the key figures working to create the Canadian AI community, to discuss how to improve things. We’re excited to welcome Canada’s first Minister of Artificial Intelligence, Evan Solomon, in a discussion that also includes one of the driving forces behind Canadian growth equity, Mark Shulgan, also a co-founder and partner at Intrepid, as well as Adam Keating, the co-founder and CEO of CoLab, a software platform that uses AI to accelerate and improve engineering design processes, based in St. John’s, Newfoundland.

Their discussion highlights the special moment in which Canadian AI finds itself—as well as the challenges the country must overcome to achieve international success.

GUESTS AND HOSTS (extended bios below)
Evan Solomon, Canada’s Minister of AI and Digital Innovation
Adam Keating, CEO & co-founder, CoLab
Mark Shulgan, co-founder and partner, Intrepid Growth Partners
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners

LINKS
Derby Mill series website. Derby Mill is created by the team at Intrepid Growth Partners.
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms: YouTube // Spotify // Apple Podcasts

DISCUSSION POINTS
00:00 Cold open
01:35 Context for episode
02:19 What is CoLab?
04:21 Role of AI
06:21 AI beyond hotspots
07:44 Canada’s AI potential
17:09 AI in St. John’s
24:22 CoLab’s innovation
29:04 Canada’s greatest risk
37:36 Final remarks

Evan Solomon is Canada’s first Minister of Artificial Intelligence and a Member of Parliament representing Toronto Centre. Before entering politics, he was one of Canada’s most recognized journalists for more than 25 years, known for his incisive interviews and deep coverage of national and global issues. He co-founded Shift, an award-winning international magazine exploring the rise of the digital age, and is the author of two best-selling books, Fueling the Future and Feeding the Future. Today, Evan leads Canada’s efforts to build a responsible and ambitious AI future — one that reflects Canadian values and strengthens the country’s digital sovereignty.

Mark Shulgan is the co-founder and Partner of Intrepid Growth Partners, a growth-stage investment fund. Previously, Mark founded and led OMERS Growth Equity, which he launched in 2018. During his time at OMERS, Mark invested $1 billion in private North American software and healthcare companies and served as the chairman of the investment committee. Prior to joining OMERS, Mark co-founded and then led the Thematic Investing team (now called Venture and Growth Equity) at CPP Investments.

Adam Keating is a mechanical engineer who co-founded CoLab out of sheer frustration when he saw how engineers were being held back by inadequate tools for working together. He led development of one of the world’s first Hyperloop vehicles (taking home 2nd place internationally at SpaceX’s 2017 competition), he’s invented an electric propulsion system for large-scale aircraft, designed systems for biology-guided radiotherapy, and managed elements of multi-billion dollar energy projects—just to name a few achievements!

NUGGETS

Evan Solomon on Canada’s AI Problem (2001)
Many Canadian tech companies struggle to gain recognition and funding at home, says Canada’s Minister of AI and Digital Innovation, Evan Solomon.

Evan Solomon on Canada’s AI Potential (2002)
Canada’s Minister of AI and Digital Innovation Evan Solomon says Canadian talent and innovation are the “lowest-hanging fruit” for global AI leadership.

Canada’s Greatest AI Risk (2003)
Intrepid co-founder and Derby Mill Series host Ajay Agrawal asks Canadian AI Minister Evan Solomon about the biggest risks AI poses for the country.

Generative Design (The Derby Mill Series ep 19)

Intrepid Growth Partners — Wed, 12 Nov 2025 16:01:04 GMT

Engineering is growing more complex—but design reviews still drag through email screenshots and PowerPoints.

In this episode of the Derby Mill Series we welcome Adam Keating, CEO & co-founder of CoLab, whose platform uses AI to accelerate and improve engineering design reviews. One client achieved a 40% reduction in the cost of poor quality in a single year.

With 160 employees and clients like Ford, Hyundai, GE, Johnson Controls and Lockheed Martin, CoLab is headquartered in St. John’s, Newfoundland.

This week we’re also proud to note that Intrepid Growth Partners, the Derby Mill Series’ parent firm, led a US$72 million Series C financing round in CoLab, marking a major step in scaling the company’s AI work for engineering.

So what would that scaling look like? What’s the future of AI and engineering? And how can machine learning improve generative design? These topics and more are explored in today’s episode by our hosts Ajay Agrawal, Rich Sutton, Sendhil Mullainathan, and Niamh Gavin, along with special guest Suzanne Gildert. They ask: what if AI didn’t just assist engineers, but fundamentally changed how design decisions are made—faster, smarter, with fewer errors?

GUESTS AND HOSTS

Adam Keating, CEO & co-founder
Suzanne Gildert, co-founder & CEO, Nirvanic Consciousness Technologies
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist, CEO, Emergent Platforms

LINKS

Intrepid leads the Series C investment in CoLab
CoLab secured US$72 million in venture capital funding.
Series C round press from Axios and The Globe and Mail
Adam Keating’s LinkedIn post announcing the Series C round, which features a cool video that provides some great context
CoLab website
Video explainer of what CoLab does
Video explainer of CoLab AutoReview
Mentioned in the episode: genetic algorithms to design radio antennas
Derby Mill series website
Derby Mill is created by the team at Intrepid Growth Partners
Rich Sutton’s home page. Follow Rich on X
Sendhil Mullainathan’s website. Follow Sendhil on X
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms:
YouTube // Spotify // Apple Podcasts // Substack

DISCUSSION POINTS

00:00 Cold open
01:29 Context for episode
02:59 About CoLab
05:25 Niamh: ML techniques
07:54 Suzanne: Training data
11:25 Rich: Language & application
18:30 Niamh: Open vs. closed foundations
22:52 CoLab customer base
24:34 Sendhil: ML similarity model
30:49 Protein model for parts
33:26 CoLab at the limit
39:50 Rich: Value functions
45:44 Feedback cycles
52:35 Adam Keating responds
56:05 Final remarks

NUGGETS

Why Are People in the Loop At All? (1901)
CoLab CEO and Co-founder Adam Keating talks about designing a waterbottle. MacArthur Genius Award recipient Sendhil Mullainathan responds with why are humans in the loop at all?

The Future of Collaborative Design (1902)
Why does Suzanne Gildert, CEO of Nirvanic, worry about the future of collaborative design?

Automotive Design and AI (1903)
Derby Mill host Ajay Agrawal and co-host Niamh Gavin debate the limitations of automotive design.

Are LLMs Bitter Lesson Pilled? (The Derby Mill Series ep 18)

Intrepid Growth Partners — Thu, 09 Oct 2025 17:30:40 GMT

A trillion-dollar clash of ideas is roiling the artificial intelligence community. Today, in a special episode, our host Ajay Agrawal leads Rich Sutton, Sendhil Mullainathan and Niamh Gavin, and special guest Suzanne Gildert, in a fascinating exploration of the issue: Are Large Language Models (LLMs) sufficiently “bitter lesson pilled” to live up to their hype?

“Bitter lesson pilled” is the AI community’s term of art for scaling with the constantly falling cost of compute (e.g., search and learning). The term arises from Rich Sutton’s 2019 essay, The Bitter Lesson.

As he recently told independent journalist Dwarkesh Patel on the Dwarkesh Podcast, Rich Sutton does not believe that LLMs are sufficiently “bitter lesson pilled.” In other words, Rich believes LLMs suffer from a key vulnerability: A limit exists on their ability to improve – and it’s much closer than we’ve been led to believe.

GUESTS AND HOSTS

Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist, CEO, Emergent Platforms
Suzanne Gildert, founder and CEO, Nirvanic Consciousness Technologies

LINKS

The Dwarkesh Podcast episode featuring Rich Sutton. The computer scientist Andrej Karpathy’s take. Rich’s original Bitter Lesson essay.
Meta machine-learning engineer Chris Hayduk’s tweet about the debate on X, retweeted by Rich and referenced in this episode by Sendhil.
Good description of the train-fly problem that Sendhil mentioned, from Presh Talwalkar.
Derby Mill series website. Derby Mill is created by the team at Intrepid Growth Partners.
Rich Sutton’s home page. Follow Rich on X
Sendhil Mullainathan’s website. Follow Sendhil on X
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms:
YouTube // Spotify // Apple Podcasts

DISCUSSION POINTS

00:00 Cold open
00:39 Context for episode
01:39 The bitter Lesson
02:49 Supervised learning
04:30 Challenge of RL
09:49 Discussing a Tweet
13:30 Rich’s opinion on the big lesson
21:28 Tension in the LLM space
23:25 Behaviour and extrapolation
25:27 What is considered AI
26:05 Final remarks

NUGGETS

Why Squirrels Still Outthink Supervised AI (1801)
Derby Mill Series host Ajay Agrawal asks co-host Suzanne Gildert, why can’t AI learn like a squirrel?

Addressing Rich’s Tweet (1802)
MacArthur Genius Award recipient Sendhil Mullainathan responds to a tweet that underscores a key difference between LLMs and humans.

What Happens if LLMs Don’t Pay Off Soon (1803)
The Bitter Lesson says, “look out if you’re putting all your eggs into the basket of human knowledge,” according to Turing Award recipient Richard Sutton.

DISCLAIMER

Automating Manufacturing (The Derby Mill Series ep 17)

Intrepid Growth Partners — Thu, 25 Sep 2025 11:02:53 GMT

The Derby Mill Series hosts are back to kick off Season 2 with an episode about automating factories—an extension of a discussion we began in the series’ first-ever episode. Here, hosts Ajay Agrawal, Rich Sutton, Sendhil Mullainathan and Niamh Gavin sit down with Vention founder and CEO Etienne Lacroix and CTO Francois Giguere. Vention’s mission: to become the default operating system for factory automation, combining modular hardware, intuitive design software, and low/no-code programming tools to speed deployment and enhance performance. The team asks, What if AI could go beyond design assistance and run fully autonomous, self-optimizing factories from concept to deployment?

About Vention

Vention is a vertically-integrated manufacturing automation platform. Its primary AI application today is predicting optimal component selection and system design. When a manufacturer specifies their automation needs, Vention’s AI recommends compatible parts, layouts, and configurations from its proprietary dataset of 400,000 labelled designs, with real-time pricing and compatibility checks. Vention serves more than 4,000 factories across more than industries, including facilities belonging to Tesla, L’Oréal, Amazon and Lockheed Martin.

GUESTS AND HOSTS

Etienne Lacroix, founder and CEO, Vention
Francois Giguere, CTO, Vention
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist, CEO, Emergent Platforms

LINKS

Vention CEO Etienne Lacroix explains the mission at Nvidia GTC 2025
Vention website
Vention’s video tutorials mini-site
Derby Mill series website. Derby Mill is created by the team at Intrepid Growth Partners.
Rich Sutton’s home page. Follow Rich on X
Sendhil Mullainathan’s website. Follow Sendhil on X
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms:
YouTube // Spotify // Apple Podcasts

DISCUSSION POINTS

00:00 Cold open
01:52 Automating manufacturing with Vention
03:45 Factory assembly tasks
05:40 AI for design
07:48 Faster and cheaper
10:28 When automation reaches its limits
10:43 Pragmatic control system design
12:02 AI training datasets
12:58 Vention’s end-to-end platform
15:20 Hybrid AI model approaches
17:47 AI spotting unmet needs
21:28 Manual versus automated processes
27:46 Full process of factory automation
37:20 Customer interfaces
40:59 Data feedback and improvement
45:58 Distribution shift in AI
1:00:17 Adaptive AI in factories
1:04:59 Final thoughts

NUGGETS

Why Automating Factories Is Becoming Faster and Cheaper (1701)
Intrepid’s Ajay Agrawal asks Vention founder and CEO Etienne Lacroix why automating factories is becoming faster to do, and cheaper to implement.

Automating Automation (1702)
Vention CTO Francois Giguere describes the future of AI-driven workflows, which he says includes the counterintuitive tagline of “automating automation.”

ML Can Fix the Black Box Model Challenges (1703)
Why MIT’s Sendhil Mullainathan believes machine learning can do what physical models can’t.

DISCLAIMER

Drug Discovery (The Derby Mill Series ep 16)

Intrepid Growth Partners — Tue, 19 Aug 2025 13:15:22 GMT

Our hosts chat with Liran Belenzon, CEO and co-founder of BenchSci. Based in Toronto, BenchSci has raised more than $215-million to date, and is backed by such funds as former U.S. vice-president Al Gore’s Generation Investment Management, private and public markets investment giant TCV, Google-backed Gradient Ventures and F-Prime Capital Partners. More than half of the world’s largest pharmaceutical companies are clients of BenchSci, which is officially known as Scinapsis Analytics Inc.

The company’s mission is to accelerate the speed and quality of life-saving R&D to improve patient health. This episode touches on the challenges and potential of using AI in drug discovery, emphasizing the importance of understanding disease biology and the need for significant investment in data collection and analysis. The name, BenchSci, is a reference to “bench science,” the fundamental laboratory research that uncovers the biological mechanisms underlying diseases and forms the foundation for drug discovery.

With machine intelligence, BenchSci seeks to automate hypothesis generation and experiment design by deeply analyzing scientific publications, preprints, and pharma data. Central to their approach is building a comprehensive knowledge graph that maps bio-entities such as genes, proteins, and diseases, along with their complex relationships.

GUESTS AND HOSTS
Liran Belenzon, co-founder and CEO, BenchSci
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist, CEO, Emergent Platforms

LINKS
BenchSci explanation video
BenchSci website
BenchSci ranked #29 on Deloitte’s 2024 Technology Fast 500™
BenchSci named to The Globe and Mail’s Canada’s Top Growing Companies 2024 list
Liran’s 2023 TechTO talk about fundraising
Mentioned by Sendhil in this episode: Don R. Swanson, a pioneer in information science
Derby Mill show website
Rich Sutton’s home page. Follow Rich on X
Read Sendhil’s co-written journal on Machine Learning as a Tool for Hypothesis Generation
Sendhil Mullainathan’s website. Follow Sendhil on X
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms:
YouTube // Spotify // Apple Podcasts

DISCUSSION POINTS

00:00 Cold open and introductions
01:33 R&D for drug discovery and BenchSci
02:07 A shocking number of drug trials fail
04:33 What BenchSci does and doesn’t do
09:50 What kind of feedback is sent to BenchSci?
14:09 Where does BenchSci fall on these extremes?
16:39 Is BenchSci too ambitious?
21:20 Niamh’s take
25:37 Rich’s take
27:15 Hypothesis generation
29:31 What Niamh loves about AI
34:47 Final remarks

NUGGETS

Small Changes in Drug Research Matter (1601)
Intrepid's Sendhil Mullainathan explains why even a 1% improvement in drug trial success can be worth millions.

AI for Discovery (1602)
Intrepid's Niamh Gavin shares how AI’s "global sweep" could unlock science’s blind spots.

AI’s Biggest Scientific Breakthrough (1603)
Intrepid’s Sendhil Mullainathan explains the hidden obstacle holding back AI’s biggest scientific breakthroughs.

DISCLAIMER

Business Productivity (The Derby Mill Series ep 15)

Intrepid Growth Partners — Thu, 31 Jul 2025 18:18:33 GMT

Joining the usual Derby Mill team of Ajay Agrawal, Rich Sutton, Sendhil Mullainathan and Niamh Gavin are two experts in the automation of business workflows: AppliedAI CEO and founder Arya Bolurfrushan, and member of the technical staff Phillip Kingston.

AppliedAI closed a $55 million USD Series A round of financing in February 2025 led by G42 and with backing from Palantir and McKinsey, among others. With a pre-investment valuation of $300 million, the UK-founded, Abu Dhabi-based firm develops software to enhance the efficiency of businesses by automating their back-office processes, particularly in highly regulated industries such as healthcare, insurance, and pharmaceuticals. For example, AppliedAI processed more than four million pages of U.S. medical records in 2024. On its client list are such firms as Abu Dhabi’s M42 Healthcare Group, U.S. law firm Morgan & Morgan and UK-based drug safety firm Qinecsa.

In this discussion, Arya and Phillip join the Derby Mill hosts to discuss the technicalities of automating workflows, such as medical coding for hospitals. They explore the challenges and opportunities of integrating AI and human intelligence to optimize things at the limit, and conclude by speculating how business could change when automation is fully integrated into every step of the process.

GUESTS AND HOSTS

Arya Bolurfrushan, founder and CEO, AppliedAI
Phillip Kingston, member of the technical staff, AppliedAI, and Visiting Professor at State University Kyiv Aviation Institute, Kyiv, Ukraine
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist, CEO, Emergent Platforms

LINKS

AppliedAI’s Series A press release
AppliedAI website
Arya Bolurfrushan on McKinsey’s Faces of Disruption
Phillip Kingston’s personal webpage
Rich Sutton’s home page. Follow Rich on X
Sendhil Mullainathan’s website. Follow Sendhil on X
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms:
YouTube // Spotify // Apple Podcasts // Substack
Thumbnail image is a detail from a mural by Diego Rivera, Man at the Crossroads

DISCUSSION POINTS

00:00 Cold open and introductions
01:10 Business productivity workflows and Applied AI
02:37 How most workflows are 80% similar
06:39 An example from the healthcare industry
10:21 AppliedAI’s commercial approach
12:50 Niamh asks Philip to get technical on their process
16:32 What is "supervised automation"?
25:21 Sendhil’s take
32:56 Rich’s take
40:15 How AppliedAI may change things at the limit

NUGGETS

How Will AI Algorithms Change Human Workflows? (1501)
MIT Economist Sendhil Mullainathan asks, if we knew there was an AI algorithm underneath most business processes, would the entire workflow be different?

Fewer Human Hours Per Case (1502)
AppliedAI’s Phillip Kingston describes how the company chooses which workflows to automate.

Human Auditors, Not Processors (1503)
AppliedAI’s Arya Bolurfrushan explains why the cost of auditing AI workflows may increase over time.

DISCLAIMER

Humanoid Robots (The Derby Mill Series ep 14)

Intrepid Growth Partners — Tue, 08 Jul 2025 10:15:29 GMT

Nvidia CEO Jensen Huang recently described physical AI, a category that includes robots that can perceive, understand and act in the real world, as the next wave in artificial intelligence. So in the last episode before Derby Mill’s summer break, and our first-ever in-person recording session, the team of Ajay Agrawal, Rich Sutton, Sendhil Mullainathan and Niamh Gavin welcome Suzanne Gildert, the CEO and founder of Nirvanic Consciousness Technologies in Vancouver, BC.

Gildert is a pioneering figure in the humanoid robotics community. Here, she discusses with the Derby Mill team such questions as: Why now for humanoid robots? What are the advantages and disadvantages of the bipedal human form factor? What makes humanoid robots difficult to create? The episode concludes with Ajay asking the team what they want listeners to think about during our two-month summer break. See you in September!

GUESTS AND HOSTS

Suzanne Gildert, co-founder & CEO, Nirvanic Consciousness Technologies

Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist, CEO, Emergent Platforms

LINKS

Nirvanic Consciousness Technologies homepage
The Jenson Huang / NVIDIA presentation Ajay references early in the episode. Reuters story
Both Sendhil and Rich love the sci-fi novels of Iain Banks
Derby Mill show website
Rich Sutton’s home page. Follow Rich on X
Sendhil Mullainathan’s website. Follow Sendhil on X
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms:
YouTube // Spotify // Apple Podcasts // Substack

DISCUSSION POINTS

00:00 Cold open

01:52 Welcome and intro to humanoid robots and Suzanne Gildert

04:20 What’s so hard about building a humanoid robot?

06:15 The complexity of the human body

07:55 So why bother making a humanoid robot?

09:30 Why are humanoid robots so hot right now?

10:50 Why now: AI software

15:49 Rich Sutton explains why humanoid robots are so intriguing

17:05 Can we code robots the same way we approached LLMs?

20:30 Teaching robots with reinforcement learning in simulation

21:47 Sendhil: How important are humanoid robots?

29:10 Niamh: Is the bipedal form factor the best all-around solution?

37:15 Sendhil: What about hybrid human-robot creatures?

41:15 Agent architecture and humanoid robots

44:35 The idea that we explore by random action selection

48:00 Suzanne on types of decision making

52:13 Decision making as centrepiece of economics, and AI

57:19 Quantum physics and self-aware AI

1:00:50 Defining consciousness

1:05:00 Lightning round: Niamh on cost of experimentation

1:06:33 LR: Ajay on what’s RLable

1:07:37 LR: Rich on AI disillusionment

1:08:40 LR: Suzanne on AI consciousness

1:10:20 LR: Sendhil on the “what is AI” turf war

1:18:18 Ajay wraps up season 1

NUGGETS

Nugget 1 - Cost of Experimentation

Intrepid's Ajay Agrawal asks AI scientist Niamh Gavin to name one topic for listeners to reflect on over Derby Mill's summer break.

Nugget 2 - AI Disillusionment and Turf Wars

Intrepid's Ajay Agrawal asks Turing Award winner Rich Sutton to name one topic for listeners to reflect on over Derby Mill's summer break.

Nugget 3 - Consciousness and Empathy

Intrepid's Ajay Agrawal asks Nirvanic CEO Suzanne Gildert to name one topic for listeners to reflect on over Derby Mill's summer break. Her response is to appeal to viewers to question any fearful reaction they have to the notion of conscious AI.

DISCLAIMER

Data Privacy (The Derby Mill Series ep 13)

Intrepid Growth Partners — Tue, 17 Jun 2025 22:01:02 GMT

Headquartered in Toronto, Private AI detects and removes personally identifiable information (PII) from data using large language models (LLMs), all without compromising individual or institutional privacy. In this episode, Private AI co-founder and CEO Patricia Thaine offers a behind-the-scenes look at the company’s technical strategy, including the scalability challenge inherent in protecting confidential company information, and the growing threat of re-identification. With more than 30,000 hours invested in building their PII detection system, Private AI now operates in seven countries and partners with organizations such as the Business Development Bank of Canada, MaRS, and the University of Toronto.

This episode features the Intrepid team exploring such questions as:

How can organizations effectively protect personally identifiable information (PII) and confidential company information in large language models?
What are the risks of re-identification, even after attempting to anonymize data?
How can companies balance data utility with privacy preservation?
How can privacy protection be approached as a dynamic, evolving challenge rather than a static solution?
What role can technology play in helping organizations understand and control their data privacy?

GUESTS AND HOSTS

Patricia Thaine, co-founder & CEO, Private AI
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist, CEO, Emergent Platforms

LINKS

Private AI website, explainer video
Private AI demo, PrivateGPT
Read the NYT article, A Face Is Exposed for AOL Searcher No. 4417749
Pymetrics, a company that pioneered the use of AI and behavioural science to improve workforce decisions, was acquired by Harver.
Rich Sutton’s home page. Follow Rich on X
Sendhil Mullainathan’s website. Follow Sendhil on X
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms:
YouTube // Spotify // Apple Podcasts

DISCUSSION POINTS

00:00 Introduction
01:22 Meet Patricia Thaine of Private AI
01:41 About Private AI
02:54 How Private AI redacts and protects data
04:18 What would scalability look like for confidential company info?
08:14 Deconstructing NYT’s article: A Face is Exposed for AOL Searcher No. 4417749
10:12 Can your digital footprint be an identifier?
15:35 Solving the synthetic data problem
19:15 How data minimization can help with privacy
21:24 Mapping out the future of data privacy
25:56 What would a reward function look like?
27:12 Final comments

NUGGETS

Nugget 1 - The Challenge of De-Identification Concerning Data Privacy

Private AI CEO and co-founder Patricia Thaine describes the challenge of data privacy and de-identification.

Nugget 2 - Consumer Market Demand and Regulation

Intrepid’s Sendhil Mullainathan explores the challenge of creating a start-up in the “personally identifiable information” space.

Nugget 3 - Different Types of CII and PII

Private AI’s Patricia Thaine discusses the nuances of removing personally identifiable information, as even a piece of jewellery in an X-ray can compromise anonymity.

DISCLAIMER

Hospital Care (The Derby Mill Series ep 12)

Intrepid Growth Partners — Tue, 03 Jun 2025 21:26:02 GMT

Hospitals in jurisdictions around the world face tight budgets and staff shortages. Artisight offers an AI-powered suite of software services designed to help hospitals do more with the limited resources they have available. Based in Chicago and led by a medical doctor who also has an MBA, Dr. Andrew Gostine, Artisight’s mission is to improve quality metrics and financial outcomes with the help of computer vision, IoT sensors and vital-sign monitoring.

For example, one hospital system that used Artisight’s technology, Northwestern Medicine, saw a 52% reduction in nursing overtime and a 76% reduction in nursing turnover alongside improved nursing and patient satisfaction scores. At the start of 2024, Artisight announced that it raised US$42 million in a funding round that was oversubscribed by 2.4x and included NVIDIA as an investor.

On the agenda in today’s discussion: What’s the potential for AI-enabled healthcare administration? How can AI be of assistance to the healthcare industry? What can be done to increase efficiency in the near term, and where does the technology go at the limit? The Derby Mill team talks to Artisight CEO and co-founder Dr. Andrew Gostine, and chief science officer and co-founder Tim Koby, to discuss the future of healthcare.

GUESTS AND HOSTS

Dr. Andrew Gostine, Co-Founder & Chief Executive Officer, Artisight
Tim Koby, Co-Founder & Chief Science Officer, Artisight
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist, CEO, Emergent Platforms

LINKS

Derby Mill show website
Artisight’s website, explainer video
Learn more about Artisight’s Series B funding round
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms:
YouTube // Spotify // Apple Podcasts

DISCUSSION POINTS

00:00 Introduction
01:07 Meet the Artisight team
02:00 Core value of Artisight
03:44 Sensor suite: What data is collected
05:42 Artisight’s top 3 AI predictions
09:39 Why fall prevention matters most
16:12 AI in hospital care—why now?
24:05 Raising the ceiling in healthcare
25:47 Improving models without moving data
30:19 Smarter AI vs. smartest doctor?
40:14 Are there limits to trusting AI?
44:54 Numbers that prove AI trust
51:01 Next big AI-driven interventions
55:05 Rewards for in-hospital problem-solving
59:57 AI vs. human default in hospital care
01:09:35 Final remarks

NUGGETS

Nugget 1 - Using AI to Maximize Hospital Care

AI can now recognize procedures like IV insertions without staff saying a word, thanks to AI using voices as a timestamp and training computer vision with synthetic images.

Nugget 2 - "Why now?" How Real-World Problems Held AI Back Until Today

For years, AI has promised to transform healthcare. So why is it only working now? Emergent Platform CEO Niamh Gavin uses her healthcare expertise to describe the real-world hurdles that held earlier AI solutions back. Artisight CEO and co-founder, Dr. Andrew Gostine, explains the innovation that was needed to reach the quality of AI and get it to where it is today.

Nugget 3 - Stopping Sepsis Before It Starts

AI could predict sepsis up to 18 hours before it strikes—early enough that patients may never meet the clinical criteria at all. Artisight CEO and co-founder Dr. Andrew Gostine explains how predictive intelligence is transforming preventative care, and why fall prevention was just the beginning.

DISCLAIMER

Digital Advertising (The Derby Mill Series ep 11)

Intrepid Growth Partners — Tue, 20 May 2025 10:30:25 GMT

StackAdapt is an advertising platform that leverages AI to optimize digital ad campaigns across multiple channels, including display, video, native, and connected TV. With an auction system, it evaluates millions of ad opportunities each second using predictive analytics to maximize ROI and enhance audience targeting. By integrating customer data and providing privacy-conscious, scalable solutions, StackAdapt provides advertisers with data-driven insights and automated ad placement.

So how can AI enhance discovery and shape awareness of digital advertising solutions that people may not yet realize they need? And what reward systems might be most effective for RL in optimising ad campaigns? The Derby Mill team talks to StackAdapt CTO and co-founder Yang Han to discuss potential answers.

GUESTS AND HOSTS

Yang Han, CTO and co-founder, StackAdapt
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist, CEO, Emergent Platforms

LINKS

Derby Mill show website
StackAdapt’s website and explainer video
Read Rich Sutton’s latest paper Welcome to the Era of Experience
Rich Sutton’s home page. Follow Rich on X
Sendhil Mullainathan’s website. Follow Sendhil on X
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms:
YouTube // Spotify // Apple Podcasts

DISCUSSION POINTS

00:00 Introduction
02:00 Welcome, Yang Han, CTO and Co-Founder of StackAdapt
02:45 How advertising on StackAdapt works
09:10 How StackAdapt thinks about ROI
11:30 Yang on ad competition and who gets the credit.
13:55 Niamh on what StackAdapt will look like at the limit.
18:23 Sendhil on how we can surface decision-making in advertising.
24:20 Rich on the advantages of assistance-based shopping.
29:13 Becoming customer-focused with the rise of AI
31:53 What executives lack when perfecting the matching problem.
26:47 What’s one thing investors should pay attention to in this industry?

Nugget 01 - Alternative Customer-First Business Model

Nugget 02 - Educating Customers with Personalized Ads

Nugget 03 - Disrupting the Ad Market with Agent Discovery

DISCLAIMER

Intrepid GP is an investor in StackAdapt. The content of this podcast is for informational and educational purposes only and should not be construed as marketing, solicitation, or an offer to buy or sell any securities or investments. The opinions expressed in this video are those of the participants and do not necessarily reflect the views of Intrepid Growth Partners or its affiliates. Any discussion of specific companies, technologies, or industries is for illustrative purposes and does not constitute investment advice. Viewers are encouraged to consult with their own financial, legal, and tax advisors before making any investment decisions.

Welcome to the Era of Experience (The Derby Mill Series ep 10)

Intrepid Growth Partners — Tue, 06 May 2025 11:30:15 GMT

Derby Mill co-host Richard Sutton and his former student, David Silver, recently published a paper about the future of artificial intelligence, called Welcome to the Era of Experience. So in this episode the show’s other hosts—Ajay Agrawal, Sendhil Mullainathan and Niamh Gavin—take their chance to interview Rich about the essay, and provide their take on its implications.

Today’s large language models (LLMs) are trained on human-generated data. So far, this has led to the development of incredible capabilities, such as mastering complex games like backgammon or chess, or absorbing content created by humans and creating fascinating new iterations of art.

While the evolution of LLMs—from AlphaZero (2017) to ChatGPT (2022) to DeepSeek (2025) and beyond—can make it seem as though their possibilities are endless, the agents remain constrained by the scope of the data they are given. In the paper, Silver and Sutton write that “in key domains such as mathematics, coding, and science, the knowledge extracted from human data is rapidly approaching a limit.” Consequently, AI agents will have to be trained on other data, such as their own experiences, which could lead to rapid innovation and superhuman capabilities—a time period which Silver and Sutton refer to as the “age of experience.”

This episode, a roundtable discussion, focuses on the following quotes pulled from the paper:

Why now? "This will become possible, as outlined above, when agents are able to autonomously act and observe in streams of real-world experience, and where the rewards may be flexibly connected to any of an abundance of grounded, real-world signals."
Why science? "Perhaps most transformative will be the acceleration of scientific discovery."
Human-like vs superhuman AIs. "This era of experience will likely be characterised by agents and environments that, in addition to learning from vast quantities of experiential data, will break through the limitations of human-centric AI systems... Furthermore, the pursuit of this agenda by the AI community will spur new innovations in these directions that rapidly progress AI towards truly superhuman agents.”

From the paper: Figure 1: A sketch chronology of dominant AI paradigms. The y-axis suggests the proportion of the field’s total effort and computation that is focused on RL.

GUESTS AND HOSTS

Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist, CEO, Emergent Platforms

LINKS

Derby Mill show website
Read Rich Sutton’s latest paper Welcome to the Era of Experience
Rich Sutton’s 2019 paper The Bitter Lesson
Co-founder of OpenAI, Ilya Sutskever, says AI reasoning power will become less predictable
Listen to our previous episode about DeepSeek
Check out co-author David Silver’s website
Rich Sutton’s home page. Follow Rich on X
Sendhil Mullainathan’s website. Follow Sendhil on X
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms: YouTube // Spotify // Apple Podcasts

DISCUSSION POINTS

00:00 Introduction
01:34 Context about the paper
02:56 Chronology of AI paradigms
03:10 Why now?
06:09 Niamh’s chronology of AI development
14:10 Why science?
20:36 Sendhil on scientific research and AI
27:07 Grounded vs. ungrounded rewards
29:21 Rich on RL temporal difference errors
31:10 Human-like vs. superhuman AIs
36:40 Final comments

Nugget 01 - AI for Scientific Discovery

Nugget 02 - Is Science like RL?

Nugget 03 - The Value of Experience

DISCLAIMER

Cancer Detection (The Derby Mill Series ep 09)

Intrepid Growth Partners — Wed, 16 Apr 2025 08:01:54 GMT

Skin Analytics is a UK company using AI to automate the diagnosis of serious skin conditions, starting with skin cancer. Its core product, DERM, is the only Class III CE mark AI medical device for autonomous dermatology in the UK’s health system. Used on more than 150,000 real-world patients, DERM achieves 99.8% negative predictive value, outperforming dermatologists. The company is expanding into general dermatology and launching in the EU and US.

In the future, Skin Analytics intends to create a dermatology AI platform that is able to diagnose and treat a broader range of conditions. Based on a diverse sampling of low-cost data, the company intends its platform to transition from self-supervised to unsupervised learning, enabling ubiquitous, low-friction health monitoring.

This episode features the Intrepid team exploring such questions as:

What would it take to build healthcare around AI abundance, not human bottlenecks?
How might one frame an approach to reach 99% automation in dermatological triage?
What are the tradeoffs between sensitivity, specificity, and health system efficiency?
How could reward systems (RL or pathway-based optimization) be introduced?
What’s the potential of self-supervised learning across multiple medical modalities?

GUESTS AND HOSTS

Neil Daly, founder and director, Skin Analytics
Jack Greenhalgh, AI director, Skin Analytics
Ajay Agrawal, co-founder and partner, Intrepid Growth Partners
Richard Sutton, senior advisor, Intrepid Growth Partners, 2024 Turing Award recipient, pioneer of reinforcement learning and professor, University of Alberta
Sendhil Mullainathan, senior advisor, Intrepid Growth Partners, MacArthur Genius grant recipient and professor, MIT
Niamh Gavin, senior advisor, Intrepid Growth Partners, Applied AI scientist, CEO, Emergent Platforms

LINKS

Derby Mill show website: insights.intrepidgp.com/podcast
Skin Analytics website and explainer video
Rich Sutton’s home page. Follow Rich on X
Sendhil Mullainathan’s website. Follow Sendhil on X
Be sure to catch every episode of The Derby Mill Series by subscribing on the following platforms: YouTube // Spotify // Apple Podcasts

DISCUSSION POINTS

00:00 Introduction
01:24 Meet the team: Skin Analytics
06:12 The lead-up to image recognition
10:29 Patient drop-off post-referral
14:03 Getting classification right
18:47 Integrating into the healthcare system
22:36 Cancer detection in the limit
27:55 At-home cancer detection
34:10 Making dermatology RL-able
45:00 Using data as proxies for other diagnoses
50:21 Early detection vs. overdiagnosis
55:07 Higher rates of cancer detection advantages
57:00 What took so long?
59:07 Final remarks

Nugget 01 - Sensors Reveal Hidden Data in the Skin

Traditionally, dermatology has been rate-limited by the human eye and optical sensors. So incorporating a variety of additional sensors to collect more diverse and comprehensive data can open the door to a new kind of pre-primary care, potentially revealing more information about internal conditions like hypertension or liver disease.

Nugget 02 - The Economic Model Behind At-Home Diagnoses

There's a massive direct-to-consumer interest in skin health, which opens the door to a potential expansion of at-home skin-monitoring apps that could be used beyond only in primary care settings. But overdiagnoses risk overwhelming the healthcare system. In order to avoid case buildup, these apps require an economic model that leverages medical systems and consumer trust.

Nugget 03 - Redesigning the Treatment Delay

What prevents people from accessing treatment is not the diagnostic delay (which often involves a lengthy wait for results), but rather the delay in seeking help: People tend to wait for a reason to address an issue, which increases the risk of lowering the survival rate as a disease spreads.

DISCLAIMER

Intrepid GP is an investor in Skin Analytics. The content of this podcast is for informational and educational purposes only and should not be construed as marketing, solicitation, or an offer to buy or sell any securities or investments. The opinions expressed in this video are those of the participants and do not necessarily reflect the views of Intrepid Growth Partners or its affiliates. Any discussion of specific companies, technologies, or industries is for illustrative purposes and does not constitute investment advice. Viewers are encouraged to consult with their own financial, legal, and tax advisors before making any investment decisions.