Episode 21: Why Great Data Still Leads to Bad Decisions (And How to Fix It)

Lance Martin of LangChain joins High Signal to outline a new playbook for engineering in the AI era, where the ground is constantly shifting under the feet of builders. He explains how the exponential improvement of foundation models is forcing a complete rethink of how software is built, revealing why top products from Claude Code to Manus are in a constant state of re-architecture simply to keep up. We dig into why the old rules of ML engineering no longer apply, and how Rich Sutton's "bitter lesson" dictates that simple, adaptable systems are the only ones that will survive. The conversation provides a clear framework for leaders on the critical new disciplines of context engineering to manage cost and reliability, the architectural power of the "agent harness" to expand capabilities without adding complexity, and why the most effective evaluation of these new systems is shifting away from static benchmarks and towards a dynamic model of in-app user feedback. LINKS Lance on LinkedIn (https://www.linkedin.com/in/lance-martin-64a33b5/) Context Engineering for Agents by Lance Martin (https://rlancemartin.github.io/2025/06/23/context_engineering/) Learning the Bitter Lesson by Lance Martin (https://rlancemartin.github.io/2025/07/30/bitter_lesson/) Context Engineering in Manus by Lance Martin (https://rlancemartin.github.io/2025/10/15/manus/) Context Rot: How Increasing Input Tokens Impacts LLM Performance by Chroma (https://research.trychroma.com/context-rot) Building effective agents by Erik Schluntz and Barry Zhang at Anthropic (https://www.anthropic.com/engineering/building-effective-agents) Effective context engineering for AI agents by Anthropic (https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) How we built our multi-agent research system by Anthropic (https://www.anthropic.com/engineering/multi-agent-research-system) Measuring AI Ability to Complete Long Tasks by METR (https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/) Your AI Product Needs Evals by Hamel Husain (https://hamel.dev/blog/posts/evals/index.html) Introducing Roast: Structured AI workflows made easy by Shopify (https://shopify.engineering/introducing-roast) Watch the podcast episode on YouTube (https://youtu.be/2Muxy3wE-E0) Delphina's Newsletter (https://delphinaai.substack.com/)

Paras Doshi (Head of Data, Opendoor; former data leader at Amazon) joins High Signal to unpack the playbook for building an indispensable data function. He shares his experience tackling the classic scaling challenge of fragmented data at Opendoor, where rapid growth led to inconsistent metrics across the business, and turning the data function into a centralized strategic asset. We dive deep into how to earn a true seat at the table, why he believes AI is creating the "100x individual contributor," and how the principles of agency, autonomy, and adaptability are the new essentials for data careers. The conversation also explores the pragmatic divide between batch and real-time ML, how to identify a truly data-led company, and why leaders must shield their top talent to unlock disproportionate impact. LINKS Paras Doshi on LinkedIn (https://www.linkedin.com/in/doshiparas/) Insight Extractor, Paras' blog on analytics, data science, and business intelligence (https://insightextractor.com/) Watch the conversation on YouTube (https://youtu.be/DDSKxL_JeLc) Delphina's Newsletter (https://delphinaai.substack.com/)

Vishnu Ram Venkataraman (Generative AI Executive & Entrepreneur; former AI Leader at Credit Karma and Intuit) joins High Signal to unpack the true cost of generative AI. Having scaled AI solutions impacting over 140 million users, Vishnu reveals why the ease of shipping Gen AI prototypes often masks significant operational and engineering debts, challenging the conventional wisdom of rapid deployment. We dive deep into the strategic shift from traditional ML to Gen AI, discussing why the shelf value of code is dramatically falling, how to design new organizational triads for continuous iteration, and the critical differences in testing probabilistic AI systems. The conversation also explores how to manage risk with sensitive data, the power of synthetic data in early development, and which mature ML practices remain indispensable in the new AI era. LINKS Vishnu on LinkedIn (https://www.linkedin.com/in/vishnuvram/) Fei-Fei Li on Generative AI as a Civilizational Technology (https://high-signal.delphina.ai/episode/fei-fei-on-how-human-centered-ai-actually-gets-built) Tim O'Reilly on The End of Programming As We Know It (https://high-signal.delphina.ai/episode/tim-oreilly-on-the-end-of-programming-as-we-know-it) Watch the conversation on YouTube (https://youtu.be/vDQdCl_EOKg) Delphina's Newsletter (https://delphinaai.substack.com/)

Sergey Fogelson (VP of Data Science, Televisa Univision) joins High Signal to reveal how the world’s largest Spanish-language media company built a sophisticated data engine from the ground up. This transformation fueled a tenfold expansion of its digital streaming business by redefining how the company connects with 300 million viewers worldwide. At the heart of this success is a proprietary household graph that creates a single, privacy-first view of a massive and culturally diverse audience. We dig into the journey from basic data unification to building production-ready recommendation engines, how his team uses embeddings on user behavior to uncover surprising connections in content consumption, and the trade-offs between investing in internal data tools versus direct revenue-driving products. The conversation also explores a pragmatic framework for AI adoption, showing how foundational machine learning often outperforms chasing the latest trends and where LLMs can deliver real, measurable value. LINKS Sergey Fogelson on LinkedIn (https://www.linkedin.com/in/sergeyfogelson/) Watch the conversation on YouTube (https://youtu.be/f9R8mGcwygU) Delphina's Newsletter (https://delphinaai.substack.com/)

Andrés Bucchi (Chief Data Officer, LATAM Airlines) joins High Signal to unpack how a century-old airline reinvented itself with data and AI—and how that transformation is unlocking value from fuel efficiency to fraud detection. LATAM has built a massive data operation, experimenting across everything from pricing to operations, while customers benefit from a more reliable and secure travel experience. We dig into how LATAM fostered an experimentation culture, why existing data infrastructure is a critical asset, and how the biggest bottleneck in AI adoption isn't the technology itself, but human decision-making. The conversation also looks ahead to the future of generative AI as a software engineering problem, and the organizational changes needed to unlock its full potential. LINKS Andrés Bucchi on LinkedIn (https://www.linkedin.com/in/bucchi/) Tim O'Reilly on The End of Programming As We Know It, High Signal (https://high-signal.delphina.ai/episode/tim-oreilly-on-the-end-of-programming-as-we-know-it) Watch the conversation on YouTube (https://youtu.be/U_eaOmt-Rw4) Delphina's Newsletter (https://delphinaai.substack.com/)

Anu Bharadwaj (President, Atlassian) joins High Signal to unpack how humans and AI agents will work together across the enterprise, and how that shift could change the very nature of teamwork. Atlassian employees have already built thousands of agents across product, marketing, engineering, and HR teams, while customers like HarperCollins are cutting manual work by 4x as industries from publishing to finance rethink their workflows. We dig into how Atlassian’s culture enables bottom-up experimentation, why grounding and reliability are critical for adoption, and how non-technical teams are often the ones creating the most useful agents. The conversation also looks ahead to the frontiers of multiplayer agent collaboration, proactive and ambient workflows, and the governance and compliance challenges enterprises will face as agents move from tools to teammates. LINKS Anu on LinkedIn (https://www.linkedin.com/in/anutthara/) Building effective agents by Erik Schluntz and Barry Zhang at Anthropic (https://www.anthropic.com/engineering/building-effective-agents) How we built our multi-agent research system by Anthropic (https://www.anthropic.com/engineering/multi-agent-research-system) Watch the podcast episode on YouTube (https://youtu.be/898M86sKIi8?si=YGoekFzVJ0UH6pCJ) Delphina's Newsletter (https://delphinaai.substack.com/)

Tomasz Tunguz (Theory Ventures) joins High Signal to unpack why a trillion dollars of market cap is up for grabs as AI reshapes enterprise software. He explains why workflows are now changing faster than packaged software can keep up, how “liquid software” is redefining CRM and marketing automation, and why background agents will require a new kind of “agent inbox.” We discuss the compounding errors that arise when tools are chained too finely, the hidden AI technical debt accumulating in today’s systems, and why modular stacks—mixing local and cloud models—will beat monolithic apps. The conversation also surfaces early memory architectures, what breaks when one IC manages 100 agents, and how these shifts change the real bottlenecks in scaling AI. LINKS Tomasz' Website (check out his blog!) (https://tomtunguz.com/) Tomasz on LinkedIn (https://www.linkedin.com/in/tomasztunguz/) Building effective agents by Erik Schluntz and Barry Zhang at Anthropic (https://www.anthropic.com/engineering/building-effective-agents) How we built our multi-agent research system by Anthropic (https://www.anthropic.com/engineering/multi-agent-research-system) Tim O'Reilly on The End of Programming As We Know It (https://high-signal.delphina.ai/episode/tim-oreilly-on-the-end-of-programming-as-we-know-it) Delphina's Newsletter (https://delphinaai.substack.com/)

Amy Edmondson (Harvard Business School) and Mike Luca (Johns Hopkins) join High Signal to unpack what actually drives good decisions in data‑rich organizations. Using contrasts like the Bay of Pigs vs. the Cuban Missile Crisis and product cases such as Airbnb’s work on measuring discrimination, they show how decision quality tracks conversation quality—framing options, surfacing uncertainty, and challenging assumptions. We cover common failure modes (correlation vs. causation, anchoring, hierarchy, false precision), practical meeting designs that raise the signal, and where algorithms and LLMs help or hinder human judgment. LINKS Amy on LinkedIn (https://www.linkedin.com/in/amycedmondson/) Mike on LinkedIn (https://www.linkedin.com/in/profluca/) Where Data-Driven Decision-Making Can Go Wrong: Five pitfalls to avoid by Michael Luca and Amy C. Edmondson (https://hbr.org/2024/09/where-data-driven-decision-making-can-go-wrong) Many Analysts, One Data Set: Making Transparent How Variations in Analytic Choices Affect Results (https://journals.sagepub.com/doi/10.1177/2515245917747646) Trillion Dollar Coach by Eric Schmidt, Jonathan Rosenberg, and Alan Eagle (https://www.trilliondollarcoach.com/) Delphina's Newsletter (https://delphinaai.substack.com/)

Daragh Sibley, Chief Algorithms Officer at Literati and former Director of Data Science at Stitch Fix, joins High Signal to unpack how machine-learning moves from slide-deck promise to bottom-line impact. He walks through his shift from academic research on how kids learn to read to owning inventory and personalization algorithms that decide which five books land in every child’s box. We dig into the moment a data leader stops advising and starts owning P&L-critical calls, why some problems deserve simple analytics while others need high-dimensional models, and how to design workflows where human judgment and algorithmic predictions share accountability. Along the way we talk incentive design, balancing exploration and exploitation in inventory, and measuring success in dollars—not dashboards. LINKS Daragh on LinkedIn (https://www.linkedin.com/in/daragh-sibley-2111835/) Eric Colson on Why 90% of Data Science Fails—And How to Fix It (https://high-signal.delphina.ai/episode/why-90-of-data-science-fails-and-how-to-fix-it-eric-colson) Sudarshan Seshadri on High-Stakes AI Systems and the Cost of Getting It Wrong (https://high-signal.delphina.ai/episode/high-stakes-ai-systems-and-the-cost-of-getting-it-wrong) Delphina's Newsletter (https://delphinaai.substack.com/)

Lis Costa, Chief of Innovation and Partnerships at the Behavioural Insights Team, joins High Signal to explore how behavioral science is reshaping public policy, digital platforms, and machine learning. She explains how defaults influence behavior at scale, why personalization and chatbots are unlocking new kinds of interventions, and what happens when AI systems meet real-world complexity. We also discuss the limits of nudging, the promise of boosting, and why building for human decision-making requires more than just good models. LINKS The Behavioral Insights Team (https://www.bi.team/) Lis Costa on LinkedIn (https://uk.linkedin.com/in/elisabeth-costa-6a5b35248) High Signal podcast (https://high-signal.delphina.ai/) Delphina's Newsletter (https://delphinaai.substack.com/)