Episodi

  • Stay Frosty: Wes McKinney, Kenn Software Founder & pandas Creator, on building software with agents

    Stay Frosty: Wes McKinney, Kenn Software Founder & pandas Creator, on building software with agents
    Sep 22 2026
    Wes McKinney is the founder of Kenn Software, the creator of pandas, and the co-creator of Apache Arrow. He went from an AI skeptic writing code in Emacs with no autocomplete two years ago to not writing code by hand at all. At API rates, his last 30 days of coding agent use would have cost $60,178. Listen in to hear how Wes, a top 1% developer, has architected his tech stack to get the most out of coding agents and work around their failure modes. Wes says agents are letting his team build far more software than they ever could before, but it comes with a lot of frustration. He makes the case that the frontier labs have little incentive to make their models write less code, since the more bloated the code base, the more tokens it takes to maintain. In his experience, 70 to 80 percent of agent turns introduce a bug, so agents can't be trusted to act on their own and every piece of work has to be double-checked. His team built accountability into the loop: they have agents commit after every single prompt, review each commit with roborev, and use AgentsView to track exactly where their tokens are spent. They even wrote and open sourced a Clanker Constitution, a set of house rules that bans agents from writing walls of text or spending tokens on work nobody asked for. We also get into which engineering skills still matter when agents write the code, why GitHub keeps going down, and what a developer's job might look like in 2050. 00:00 Introduction 01:04 From AI Skeptic to Believer 07:51 Why Models Get Bloated 18:11 Specs With Superpowers 20:08 Clanker Constitution Rules 26:42 Harnesses and Benchmark Games 36:32 The Real Cost 42:07 Kenn Vision and Tool Stack 48:09 Developers in 2050 Kenn Software: https://kenn.io/ Wes' LinkedIn: https://www.linkedin.com/in/wesmckinn Cagla’s LinkedIn: https://www.linkedin.com/in/caglakaymaz/ Cagla’s X: https://x.com/caglakaymazLearn more about Category Ventures: https://www.categoryvc.com/ Produced and edited by Eric Johnson from LightningPod: https://lightningpod.fm/ Theme music composed by DJ La Mano, provided by https://soundtaxi.com/
    Mostra di più Mostra meno
    51 min
  • Buy vs. Build on a $12B tech budget, with David Griffiths, Group Head of AI at Citi

    Buy vs. Build on a $12B tech budget, with David Griffiths, Group Head of AI at Citi
    Aug 24 2026
    David Griffiths is the Group Head of AI at Citi, a bank that operates in 160+ countries, serves 200 million+ customer accounts, runs a $12 billion annual tech budget, and employs 30,000+ developers. In this episode, we discuss how he decides what to buy versus build, what it takes for an AI startup to get deployed at a bank of this size and some of the wins from partnering with the likes of Cognition, xAI, Anthropic and Google Gemini. We also talk about why he thinks tokenmaxxing is a really bad idea and what he is doing to keep costs under control as usage takes off. David makes the case that there are now enough scaled use cases where smaller task-specific models will beat paying frontier prices for general intelligence, and that the timing has arrived for startups helping enterprises train or fine-tune their own models. 00:00 Introduction 01:03 Evaluating Enterprise AI Tools at Scale 10:05 Measuring Real ROI in AI 14:13 Citi's Multi-Model AI Strategy 18:55 Managing AI Risks & Cost Control 21:33 Small Language Models & Proprietary Data 24:57 Advice for AI Founders & Future Agents David's LinkedIn: https://www.linkedin.com/in/david-griffiths-17106b59/ Cagla’s LinkedIn: https://www.linkedin.com/in/caglakaymaz/ Cagla’s X: https://x.com/caglakaymazLearn more about Category Ventures: https://www.categoryvc.com/ Produced and edited by Eric Johnson from LightningPod: https://lightningpod.fm/ Theme music composed by DJ La Mano, provided by https://soundtaxi.com/
    Mostra di più Mostra meno
    28 min
  • How Agents Learn, Remember, and Improve Over Time, with João (Joe) Moura, CEO of CrewAI

    How Agents Learn, Remember, and Improve Over Time, with João (Joe) Moura, CEO of CrewAI
    Jul 13 2026
    João (Joe) Moura built CrewAI into one of the most popular agent frameworks out there, with over 50,000 GitHub stars and running inside more than half of the Fortune 500. Listen in as we get into whether frameworks are dead or just turning into harnesses, how agents learn and improve themselves, and why rising token costs are pushing enterprises toward open source and local models. 00:00 Introduction 01:45 What is CrewAI? + Frameworks vs Harnesses 05:42 AI Agents and the SaaSpocalypse 09:43 How AI Agents Use Memory and Self-Learning 14:41 Optimizing Context, Token Costs, and Local Models 19:44 Enterprise AI Use Cases and Agent Workshops 29:08 Why Use CrewAI Instead of Claude? 33:48 Community Building and the Vision for CrewAI 41:42 The Future of Open Source AI in 2050 Joe's LinkedIn: https://www.linkedin.com/in/joaomdmoura/ Cagla’s LinkedIn: https://www.linkedin.com/in/caglakaymaz/ Cagla’s X: https://x.com/caglakaymazLearn more about Category Ventures: https://www.categoryvc.com/ Produced and edited by Eric Johnson from LightningPod: https://lightningpod.fm/ Theme music composed by DJ La Mano, provided by https://soundtaxi.com/
    Mostra di più Mostra meno
    45 min
  • Arvind Jain’s AI Playbook: How The Glean CEO Built to $300M ARR

    Arvind Jain’s AI Playbook: How The Glean CEO Built to $300M ARR
    Jun 4 2026
    For the launch episode of Main Branch, I sit down with Arvind Jain, founder and CEO of Glean, who took a widely dismissed idea and turned it into one of the fastest-growing AI platforms, crossing $300M in ARR. We chat about how he built conviction when the enterprise search market was a graveyard, the three shifts (SaaS, cloud, and BERT-era transformers) that quietly made the problem solvable in 2019, and how the launch of LLMs and every step-function change from the labs since has been a lucky supercharger for Glean. We trace the product arc from a search box to an agentic coworker, and dig into the early operating playbook, including how he used to send 100 cold LinkedIn DMs a day. Arvind makes the case for how to compete when foundation labs move up the stack, and why the AI market is still 100x undersupplied. Chapters: 00:00 Introduction 02:12 Early Conviction in Enterprise Search 05:33 Why Previous Search Solutions Failed 07:42 From Transformers to Generative AI 14:24 The Scale of the AI Revolution 16:17 Competitive Moats and Strategic Focus 18:31 Advice for Founders Starting Today 21:24 The Multi-Year Stealth Journey 24:04 Early Sales Philosophy and Customer Discovery 29:38 Integrating AI at Glean 32:31 Hiring and Testing AI Aptitude 36:40 Embracing Market Challenges and Mission Arvind's LinkedIn Cagla’s LinkedIn Cagla’s XLearn more about Category Ventures Produced and edited by Eric Johnson from LightningPod Theme music composed by DJ La Mano, provided by Soundtaxi
    Mostra di più Mostra meno
    42 min