エピソード

  • Gustavo Herrera from Knowme Global

    Gustavo Herrera from Knowme Global
    2026/04/07

    Alex Gras sits down with Gustavo Herrera, Founder and CTO of Knowme Global, to talk through why recruiting still sucks and how AI might actually fix it by doing the one thing everyone avoids, sitting down and talking to people for two hours. Gustavo breaks down why LLMs are just glorified averaging, how he picked Llama Index over LangChain by keeping it simple, and why the industry's obsession with automation is backwards when the hardest part is capturing who someone actually is. They dig into prompt versus context engineering, why philosophy matters more than people think, and how AI is forcing everyone back to the most human things we do.

    Click here to watch a video of this episode.


    Join the conversation shaping the future of energy.
    Collide is the community where oil & gas professionals connect, share insights, and solve real-world problems together. No noise. No fluff. Just the discussions that move our industry forward.
    Apply today at collide.io


    Click here to view the episode transcript.

    0:00 - Why Gustavo went back to school to learn AI for recruiting
    4:30 - Choosing tech stacks and avoiding bleeding edge tools
    9:15 - How Knowme ingests long-format interviews to build candidate profiles
    14:45 - Prompt engineering versus context engineering
    19:00 - Why LLMs will never be truly intelligent
    24:20 - Philosophy's role in shaping AI and society
    30:45 - How employers don't actually know what they want
    36:15 - Speeding up the 42-day hiring process with AI interviews
    42:00 - Why Knowme does the unscalable thing nobody wants to do
    47:30 - Preventing AI from embellishing candidate answers
    54:10 - Training interviewers to strike the right conversational tone
    58:45 - When to use deterministic solutions versus agentic AI
    1:01:40 - AI bringing us back to our humanity

    https://twitter.com/collide_io

    https://www.tiktok.com/@collide.io

    https://www.facebook.com/collide.io

    https://www.instagram.com/collide.io

    https://www.youtube.com/@collide_io

    https://bsky.app/profile/digitalwildcatters.bsky.social

    https://www.linkedin.com/company/collide-digital-wildcatters
    続きを読む 一部表示
    1 時間 3 分
  • Bowden Kelly from Nutiliti

    Bowden Kelly from Nutiliti
    2026/03/24

    Bowden Kelly from Nutiliti joins Alex on AI After Dark to explain why being a fast follower beats bleeding-edge adoption, particularly when your customers just want accurate utility bills paid on time and don't care about your tech stack. He breaks down why AI native companies often confuse magic demos with production reality, why replacing his entire engineering team with agents would be insane, and how the paradigm shift isn't about chat interfaces but building artifacts that make high school interns into experienced employees. The conversation reveals why Nutiliti spent years building PDF parsers only to discover LLMs weren't good enough yet, his controversial take that the agentic AI wave is overhyped, and how the hiring equation is flipping from one PM managing ten engineers to five PMs with two engineers because code execution is getting commoditized while business reasoning stays expensive.

    Click here to watch a video of this episode.

    Join the conversation shaping the future of energy.
    Collide is the community where oil & gas professionals connect, share insights, and solve real-world problems together. No noise. No fluff. Just the discussions that move our industry forward.
    Apply today at collide.io


    Click here to view the episode transcript.

    00:00 Fast follower versus bleeding edge adoption philosophy
    07:38 Paradigm shifts from blockchain to mobile to AI
    13:29 Tech debt concerns and why Google engineer stories mislead
    21:10 Cloud Code improvements and managing context windows
    28:04 MCP servers and the actual problem of data access
    34:47 Why non-technical users aren't ready for AI interfaces
    39:07 Building artifacts versus executing tasks as organizational design
    46:17 High school intern analogy for training agentic AI
    51:18 Controversial take on AI being overhyped despite being valuable
    55:02 Hiring shift from programmers to business-minded builders
    01;06;05 Fundraising strategy for non-unicorn companies
    01;08;05 Why companies force AI into products for the story

    https://twitter.com/collide_io

    https://www.tiktok.com/@collide.io

    https://www.facebook.com/collide.io

    https://www.instagram.com/collide.io

    https://www.youtube.com/@collide_io

    https://bsky.app/profile/digitalwildcatters.bsky.social

    https://www.linkedin.com/company/collide-digital-wildcatters
    続きを読む 一部表示
    1 時間 16 分
  • Matt Robillard from SmartAC

    Matt Robillard from SmartAC
    2026/03/10

    Matt Robillard from SmartAC joins Alex on AI After Dark to explain why he walked into a company with product-market fit but antiquated tech stack and immediately started rewriting everything with AI at its core. He breaks down the difference between horizontal process automation that Service Titan can eventually replicate versus vertical intelligence modeling the actual thermodynamics of HVAC systems that nobody else can touch, why clean energy is magical thinking constrained by the Shockley-Queisser limit and Betz limit, and how hiring the best AI talent in HVAC meant recruiting from SpaceX, Databricks and Perplexity by selling real-world impact over another CRM with AI slapped on. The conversation reveals why OEMs surprisingly embrace their solution instead of fearing disruption, his controversial take that LLMs are stochastic parrots not true intelligence, and how the marginal cost of code going to zero is pure hype when production systems at scale still require senior engineers who deeply understand customer pain points.

    Click here to watch a video of this episode.

    Join the conversation shaping the future of energy.
    Collide is the community where oil & gas professionals connect, share insights, and solve real-world problems together. No noise. No fluff. Just the discussions that move our industry forward.
    Apply today at collide.io


    Click here to view the episode transcript.

    00:00 CES takeaways and hardware semiconductor relationships
    01:53 Joining SmartAC post product-market fit with legacy tech
    07:15 Data platform challenges and ML model development
    12:14 Vertical AI defensibility versus horizontal automation
    18:26 Alignment with Josh and commercial-product pairing
    24:17 Replatforming effort and organizational design overhaul
    29:27 Recruiting top AI talent from SpaceX and Databricks
    36:03 Development velocity changes and junior engineer growth
    41:01 AGI possibility and stochastic parrots debate
    48:21 Controversial clean energy physics constraints opinion
    56:42 Smart home promises and Tesla-style behavioral nudges
    01;01;10 VC organizational design mistakes over two years
    01;05;13 Who else should be on AI After Dark

    https://twitter.com/collide_io

    https://www.tiktok.com/@collide.io

    https://www.facebook.com/collide.io

    https://www.instagram.com/collide.io

    https://www.youtube.com/@collide_io

    https://bsky.app/profile/digitalwildcatters.bsky.social

    https://www.linkedin.com/company/collide-digital-wildcatters
    続きを読む 一部表示
    1 時間 5 分
  • Akhil Mantripragada from Pie

    Akhil Mantripragada from Pie
    2026/02/24

    Akhil Mantripragada from Pie joins Alex on AI After Dark to trace his journey from coding without a computer in eighth grade India to building a $100M+ revenue AI-native marketing platform for SMBs. He shares how changing seven schools growing up forced him to become adaptable, why he spent hours writing multiple code versions on paper with only one hour per week at a computer lab, and how selling his first ed-tech company to UTSA funded grad school at Columbia. The conversation reveals why Pi's first three hires were designers instead of engineers, his controversial belief that chat is the "lousiest way to interact with AI," and how meeting users where they're at - not where you want them to be - became the core design philosophy that lets three engineers do what used to take thirty. Plus, the framework for building modular systems where perfection lies in the architecture's inputs and outputs, not the code inside the black box.

    Click here to watch a video of this episode.


    Join the conversation shaping the future of energy.
    Collide is the community where oil & gas professionals connect, share insights, and solve real-world problems together. No noise. No fluff. Just the discussions that move our industry forward.
    Apply today at collide.io


    Click here to view the episode transcript.

    00:00 Growing up moving between seven schools in India
    01:39 Learning to code without a computer access
    07:13 Starting an events management company as an introvert
    12:18 The arrogance and humility combination that drives builders
    15:12 First website job at UTSA and learning by doing
    18:01 Building quantitative literacy tracking system for professors
    24:17 Why school being too easy created entrepreneurial opportunities
    26:45 Columbia grad school and starting Edu Link collaboration tool
    35:14 The two acquisition experiences and MIT licensing lessons
    40:56 Blockchain land registry in India and creator platform experiments
    43:35 Why education will become skill-based not credit-based
    46:15 Joining Toast's New Ventures program and building benchmarking
    56:07 Recruiting A-players through bias for action and ownership
    01;02;40 Why Pi hired three designers before the first engineer
    01;06;54 Meeting users where they're at versus forcing chat interfaces
    01;12;48 The black box modularity philosophy for AI-native startups

    https://twitter.com/collide_io

    https://www.tiktok.com/@collide.io

    https://www.facebook.com/collide.io

    https://www.instagram.com/collide.io

    https://www.youtube.com/@collide_io

    https://bsky.app/profile/digitalwildcatters.bsky.social

    https://www.linkedin.com/company/collide-digital-wildcatters
    続きを読む 一部表示
    1 時間 17 分
  • Canisius Rozario from collide.

    Canisius Rozario from collide.
    2026/02/18

    In this kickoff to a new series, AI After Dark, Alex Gras (VC @ Mercury Fund) talks with Canisius Rozario (CTO @ Collide) about how he built a top-tier AI team in oil & gas without any industry background. It’s a sharp, candid look at hiring, learning fast, and building AI-first systems, packed with practical takeaways for founders, CTOs, and early builders working in high-stakes, data-heavy industries.

    Click here to watch a video of this episode.


    Join the conversation shaping the future of energy.
    Collide is the community where oil & gas professionals connect, share insights, and solve real-world problems together. No noise. No fluff. Just the discussions that move our industry forward.
    Apply today at collide.io


    Click here to view the episode transcript.

    0:00 Introduction and series kickoff
    1:50 Canisius career background and journey through tech
    9:30 Joining Collide and the decision to go AI-native
    20:00 First 90 days as CTO and the "sniff test"
    33:00 Why senior engineers beat junior hires in AI builds
    37:00 Incorporating oil and gas experts into AI model training
    41:45 The Forward Deploy Engineer model and how it works
    46:00 Prioritization process for POCs and customer feedback
    52:50 What to look for in the next generation of AI CTOs
    56:30 What does "AI native" actually mean and audience Q&A

    https://twitter.com/collide_io

    https://www.tiktok.com/@collide.io

    https://www.facebook.com/collide.io

    https://www.instagram.com/collide.io

    https://www.youtube.com/@collide_io

    https://bsky.app/profile/digitalwildcatters.bsky.social

    https://www.linkedin.com/company/collide-digital-wildcatters
    続きを読む 一部表示
    1 時間 1 分