Free Form AI

Most AI systems fail before they scale because teams build the “final architecture” too early. Research spikes exist to expose what you don’t understand, not to prove you’re right. The fastest path to a working system is starting with something intentionally simple, validating invariants, and throwing away what doesn’t hold.00:00 What a research spike actually is04:15 The real problem: context overload13:29 Why you should not build the “final system” first21:42 The cathedral vs farmhouse mistake33:42 When NOT to use advanced tech like graph RAG39:04 The unsexy work that actually mattersIf your first version feels impressive, you probably built the wrong thing.

LLMs don’t fail loudly, they drift into undefined behavior and take your system with them. The only way to build stable AI systems is to enforce contracts at every boundary, especially when dealing with non-deterministic outputs. Modern Python tools like Pydantic, enums, and structured interfaces aren’t optional, they’re how you turn probabilistic generation into reliable software.00:00 Why LLMs behave like “chaos goblins”03:38 What a contract actually enforces14:56 Real bug caused by missing validation26:32 Why external APIs will break your system44:06 The worst mistake: putting logic in promptsIf you’re not validating every boundary, you’re not building software, you’re gambling.

AI has made it easier than ever to build fast, but also easier to fake progress. In this episode, we break down the difference between people who optimize for real value versus those who optimize for visibility, how this shows up in AI workflows, and why long-term thinking is the only way to build systems that actually matter.00:00 — Builders vs posers in tech05:00 — Why visibility can distort incentives12:00 — How AI enables shallow prototypes22:00 — How to identify real value35:00 — Building culture vs playing the gameAI doesn’t change the game, it exposes how you’re playing it.

AI coding agents can generate code faster than ever, but they also introduce new risks that most teams don’t understand yet. In this episode, we break down how software development is shifting from writing code to reviewing and directing AI systems, why poorly guided agents create fragile systems, and what skills actually matter as AI takes over execution.00:00 — AI writing code vs humans reviewing it06:00 — Why coding agents fail on complex systems15:00 — Tactical vs strategic use of AI in development25:00 — The role of humans as system designers and guardrails35:00 — What skills matter in an AI-driven workflowAI doesn’t remove engineers, it makes their judgment the most important part of the system.

The conversation delves into the use of AI in the legal industry, focusing on the Succession platform and its approach to will creation. It explores the collaboration between technical and non-technical stakeholders, the efficiency of billable practices, and the fundamentally human aspects of law, particularly in in-person negotiations.TakeawaysAI for sensitive topicsCollaboration between technical and non-technical stakeholdersEfficiency in billable practicesFundamentally human aspects of lawIn-person negotiations as a hallmark of a lawyer's jobChapters00:00 Introduction and Focus on AI in Law06:43 Comparison with ChatGPT Experience12:14 Efficiency Optimizations and Reception20:52 Technical Implementation and Collaboration28:00 Personal Relationship and Collaboration33:06 Disruption of Legal Industry by AI43:15 Fundamentally Human Aspects of Law

The conversation delves into the importance of data science in AI, the challenges with existing evaluation frameworks, the significance of looking at data, types of tests and prioritization, the evolution of evaluation solutions, overcoming psychological blocks, summary statistics and data exploration, and the future of evaluation and outsourcing.TakeawaysEvaluating AI products requires a deep understanding of how to test and measure stochastic systems.Looking at data is a crucial step in evaluating AI products and building intuition on what the failures are. AI is revolutionizing the process of instrumenting systems, reducing the toil and tedious tasks involved.AI can assist in exploring and analyzing data efficiently, providing insights and suggestions for improvement.Chapters00:00 Introduction and Background05:55 Importance of Looking at Data13:01 Overcoming Psychological Blocks18:50 Summary Statistics and Data Exploration29:36 Revolutionizing System Instrumentation with AI36:30 AI-Assisted Data Exploration and Analysis

The conversation covers topics related to software development, testing, and team culture. It emphasizes the importance of system-driven solutions, effective testing strategies, and fostering a positive and collaborative work environment. The discussion also delves into the impact of individual behavior on team dynamics and the significance of leading by example.TakeawaysSystem-driven solutionsEffective testing strategiesFostering a positive and collaborative work environmentImpact of individual behavior on team dynamicsLeading by exampleChapters00:00 System-Driven Solutions10:51 Team Culture and Dynamics18:12 Leadership and Mentoring25:56 Positive Work Environment

The conversation delves into the philosophy of collaboration, the absurdity of technical work, the ethical imperative of mentorship, the role of a technical role model, and embracing the absurdity of the technical world. Key takeaways include the emphasis on collaboration over individual excellence and the importance of mentorship and knowledge transfer. The conversation delves into the power of knowledge as a force multiplier and the approach to leadership, emphasizing excellence, approachability, and humor. It explores the concept of embracing absurdity, controlling the controllables, and revolting against the absurd. The discussion also highlights the importance of self-validation, finding like-minded individuals, and aspiring to guide others in the professional journey.TakeawaysCollaboration over individual excellenceThe importance of mentorship and knowledge transfer Knowledge as a force multiplierLeading with excellence, approachability, and humorEmbracing absurdity and controlling the controllablesRevolt against the absurdChapters00:00 The Philosophy of Collaboration07:04 The Absurdity of Technical Work14:41 The Ethical Imperative of Mentorship26:25 Embracing the Absurdity34:08 Existentialism and Implementing Team-Based Knowledge Transfer43:19 Rebellion Against the Absurd and Psychological Calluses49:25 Finding Like-Minded Individuals and Aspiring to Guide Others

The conversation delves into the concepts of hygiene vs. motivation factors in the workplace and explores the methods of creating motivation within a team. It also touches on the role of a manager in maintaining team efficiency and morale. The conversation delves into the importance of mentorship and the builder ethic, emphasizing the value of humility and team cohesion. It also explores the concept of navigating a career, discussing the emergent vs. deliberate approach and the significance of humility and bias for action.TakeawaysHygiene vs. motivation factorsCreating motivation in the workplace Mentorship and the Builder EthicTeam Cohesion and HumilityChapters00:00 Creating Motivation in the Workplace29:45 Mentorship and the Builder Ethic37:31 Navigating a Career: Emergent vs. Deliberate

The conversation delves into the scale of finance and the concept of 'enough' in relation to wealth. It explores the value of money, safe withdrawal rates, financial mistakes, and the transition to wealth management as a career path. The conversation delves into the use of AI-based coding assistants, the importance of guardrails and testing focus, AI oversight and refactoring, workflow and quality assurance, code review dynamics, human-in-the-loop review, coaching and career development, and the metaphor of 'testing in prod' for human interactions. The key takeaways include the adversarial prompting approach for AI-based coding assistants, the significance of soft power and people skills for senior engineers, and the metaphor of 'testing in prod' for human interactions and relationships.TakeawaysFinance operates on a different scaleThe concept of 'enough' varies based on individual circumstances AI-based coding assistants benefit from an adversarial prompting approachSoft power and people skills are essential for senior engineersTesting in prod is a metaphor for human interactions and relationshipsChapters00:00 The Scale of Finance11:02 Defining 'Enough'17:54 Safe Withdrawal Rate22:56 Financial Mistakes and Overconfidence29:56 AI-Based Coding Assistance37:59 Human-in-the-Loop Review51:57 Testing in Prod: Human Interactions

The conversation delves into Victor Dibia's career journey, global experiences, transition to a PhD, and strategic career planning. It also explores his focus on AI tooling and frameworks, as well as the evolution of Autogen and the Microsoft Agent Framework. The conversation delves into the actor-first paradigm in multi-agent systems and the concept of ensembling in machine learning. It explores the benefits of the actor-first approach and the considerations for using multiple agents in complex tasks. Additionally, it discusses the power of ensembling in complementing the biases of individual models and the potential for mixture of experts in achieving better performance.TopicsCareer progression through diverse experiencesActor-first paradigm in multi-agent systemsAutogen and Semantic KernelChapters00:00 Career Journey and Global Experiences11:03 Focus on AI Tooling and Frameworks19:11 Evolution of Autogen and Microsoft Agent Framework28:59 Actor-First Paradigm in Multi-Agent Systems36:39 Ensembling in Machine Learning

The conversation delves into the importance of knowledge transfer within an organization, highlighting the benefits of a collaborative culture, the value of building replacements, and the demonstration of one's value through knowledge sharing. It emphasizes a kindness-oriented approach and the need for frequent touch points when teaching and learning. The discussion covers the minimum requirements for knowledge transfer, the approach to teaching someone senior and junior, and the curation of work and experiences to facilitate learning and growth.TakeawaysCollaborative cultureBuilding replacementsDemonstrating valueKindness-oriented approachFrequent touch pointsChapters00:00 Introduction and UI Learning06:07 Why Help Others?18:31 Teaching Someone Senior to You26:02 Teaching Someone Junior to You33:03 Curating Work and Experiences39:24 Final Thoughts on Knowledge Transfer

The conversation covers the topics of diversifying vs. focusing, structuring your day for productivity, the evolution of AI training, AI leadership training, building community and support, and navigating career transitions. Key takeaways include the importance of structuring your calendar for productivity, the value of diverse hiring practices, and the concept of social entrepreneurship as a North Star for career navigation.TakeawaysStructuring your calendar for productivityDiverse hiring practices for better outcomesSocial entrepreneurship as a North StarChapters00:00 Diversifying vs. Focusing07:58 AI Training Evolution29:08 Building Community and Support33:52 Navigating Career Transitions

What happens when creativity is treated not as intuition, but as a system that can be studied and scaled?In this episode of Free Form AI, Michael and Ben sit down with Nicolas Douard, Lead Data Scientist at the Virtue Foundation, to explore how AI and data science are being used to automate innovation itself. Drawing from Nicolas’ PhD research, the conversation examines TRIZ — a systematic framework for inventive problem solving — and how it can be augmented with modern AI techniques to connect ideas across disciplines.The discussion moves through biomimicry as a model for interdisciplinary discovery, the use of knowledge graphs to represent and traverse complex domains, and the role AI may play in accelerating scientific insight. Along the way, this conversation unpacks deeper questions about creativity, discovery and whether innovation can be meaningfully formalized without losing its human essence.Tune into episode 26 for a wide-ranging conversation about:TRIZ as a structured methodology for inventive problem solvingBiomimicry as a blueprint for cross-disciplinary innovationHow knowledge graphs enable new forms of scientific reasoningThe role of AI in discovery, not just automationWhether creativity can be systematized without being diminishedWhether you work in data science, engineering or applied research, this episode offers a thoughtful look at how AI innovation itself might become a computable process.Note: This episode was released first on YouTube as part of Free Form AI’s video-first relaunch.

Ever wondered what senior engineers actually talk about behind closed doors?In this episode of Free Form AI, Michael and Ben open up the conversations developers usually only hear behind closed doors. We're talking how real engineering teams review code, manage dependencies, keep tests reliable and prevent their codebases from turning into chaos.Live and in real time, they break down the habits and workflows that make software durable: using reusable components to avoid reinvention, building integration tests that catch silent failures, choosing versioning strategies that won’t break downstream users, and writing documentation that actually accelerates collaboration.Tune into episode 25 for a wide-ranging conversation about: • What code reviews really accomplish • Why reusable components reduce long-term friction • How dependency management goes wrong (and how to keep it stable) • Why integration tests are the backbone of reliable software • How versioning choices shape releases • The role of clear documentation in team velocity • Why internal utilities need user-centric design • How clean codebases speed up onboarding and feedbackIf your work touches code, this episode gives you the kind of insight you’d normally only get sitting next to seasoned engineers at the office.

What happens when AI stops generating answers and starts deciding what’s true?In this episode of Free Form AI, Michael Berk and Ben Wilson dive into GPT-5’s growing role as an interpreter of information — not just generating text, but analyzing news, assessing credibility, and shaping how we understand truth itself.They unpack how reasoning capabilities, source reliability, and human feedback intersect to build, or break trust in AI systems. The conversation also examines the ethical stakes of explainability, the dangers of “sycophantic” AI behavior and the future of intelligence in a market-driven ecosystem.Tune in to Episode 24 for a wide-ranging conversation about: • How GPT-5’s reasoning is redefining “understanding” in AI • Why explainability is critical for trust and transparency • The risks of AI echo chambers and feedback bias • The role of human judgment in AI alignment and evaluation • What it means for machines to become arbiters of truthWhether you build, study, or rely on AI systems, this episode will leave you questioning how far we’re willing to let our models think for us.

How to fight complexity creep and build software that stays simple, even as it grows.Every engineer knows the struggle: a simple system slowly buried under complexity.In this episode of Free Form AI, Michael Berk and Ben Wilson break down how complexity creeps into code, dependencies and design. And why simplicity almost always wins. They cover how iteration, testing and mentorship can keep software maintainable. So where Gen AI can (and can’t) help reduce friction?Tune in to Episode 23 for a wide-ranging conversation about: • Complexity shows up in code, dependencies and design decisions • Incremental iteration helps map the solution space more effectively • Testing isn’t just QA, it’s how we preserve maintainability • Gen AI can simplify coding tasks, but it still needs human oversight • Mentorship remains one of the best ways to fight chaos in codeIf you’ve ever wrestled with “complexity creep,” this one’s for you.

In this episode of Free Form AI, Michael and Ben draw surprising parallels between military strategy and consulting practice. They break down how principles like competence, trust-building and ethical engagement translate into stronger client relationships and long-term success.From showing quiet confidence to making clients aware of their pitfalls, the discussion explores how consulting is as much about empathy and communication as it is about analysis. Along the way, they highlight why being “nice” is not a soft skill but a strategic one.Tune in to Episode 22 for a wide-ranging conversation about: • Why competence and confidence create credibility in client work • How to earn trust by showing clients their blind spots • Why ethical guidelines are non-negotiable for lasting relationships • The role of empathy, communication, and culture in consulting successIf you’ve ever wondered how to approach consulting with both rigor and humanity, this episode is for you.

Passion for the product, not compensation, is what keeps the best engineers engaged.In this episode of Free Form AI, Michael and Ben sit down with Aleksandr Patrushev, Head of DevRel at Nebius, to explore the evolving role of Developer Relations and the culture that sustains technical talent. From modernizing hardware in data centers to weighing the trade-offs between self-hosted and managed services, Alexander shares how knowledge sharing, process and passion drive both DevRel and product management.Tune in to Episode 21 for a wide-ranging conversation about: • Why DevRel is built on knowledge sharing and community impact • The ongoing challenges of modernizing hardware in cloud + AI systems • How culture influences technical staff retention and motivationIf you’re interested in the future of cloud + AI services and how culture shapes technology, this episode is for you.

Behind every robust AI system is a solution accelerator that makes complex development smarter.In this episode of Free Form AI, Michael and Ben dive into the role of solution accelerators in bridging the gap between production-ready software and simpler resources like blogs or cookbooks. They discuss frameworks vs. tools, the challenges of maintainability and how forward deployed engineers can build reusable components.Tune in to Episode 20 for a wide-ranging conversation about:• Why solution accelerators streamline development• How to balance abstraction and usability in frameworks and APIs• Testing strategies that support quality and extensibility• Where AI tools like GitHub Copilot fit into modern workflowsIf you’re focused on building software that lasts while delivering business value, this episode is for you.

In this episode of Free Form AI, Michael and Ben explore the evolution of prompt engineering and the rise of auto-optimization in AI. We discuss how agents can execute tasks autonomously, why human oversight remains essential and how defining success criteria shapes model performance.Tune in to Episode 19 for a wide-ranging conversation about: • Why prompt engineering still matters in an era of automation • How auto-optimization could replace traditional data annotation • How AI is augmenting business analytics and software workflowsIf you’re interested in where AI agents and automated optimization are heading, click through!

In this episode of Free Form AI, we sit down with Tomu Hirata (Databricks), whose journey from medicine to software engineering offers a unique lens on building AI systems. From automating construction machinery to optimizing job postings at Indeed, Tomu has worked at the edge of AI data collection and model design.Tune in to Episode 18 for a wide-ranging conversation about:• The challenges of gathering real-world data for AI in construction and beyond• Why human feedback is critical for prompt optimization• How context and memory augmentation will shape the future of AI agentsIf you’re curious about where AI agents are heading, and what it takes to make them truly effective, this episode is for you.

In this episode of Free Form AI, we go behind the scenes with co-host Ben Wilson to explore how his early experiences in the Navy shaped his approach to engineering, risk, and creativity.From high-stakes problem solving to the importance of failure-friendly environments, Ben and Michael unpack what it takes to build resilient systems and resilient people. Along the way, they discuss why fundamentals matter, how psychology plays into technical decisions, and why hobbies might be the most underrated engineering tool.Tune in to Episode 17 for a wide-ranging conversation about:• How high-pressure environments foster innovation• Why testing and validation are non-negotiables in good engineering• How hobbies support creative problem-solving• Why failure (with forgiveness) can unlock new approachesMore than a career story, this episode is a deep dive into how experience shapes intuition. The best engineers don’t just know what works. They know why it works.

Behind every major leap in AI is a wave of experimentation. And GPT-5 is no exception.In this episode of Free Form AI, Michael and Ben unpack what makes the latest generation of large language models different, from reasoning improvements and reduced hallucinations to the open-source revolution reshaping the field. They explore how the industry is moving beyond accuracy metrics to deeper forms of evaluation, where curiosity and real-world testing drive meaningful progress.Tune in to Episode 16 for a forward-looking conversation about: • How GPT-5 represents a step change in reasoning and contextual understanding • Why open-source AI models are accelerating global research collaboration • The ethical questions surrounding the path toward super-intelligenceWhether you’re building with open models or studying AI’s evolution, this episode will leave you rethinking how we measure progress.

In this episode of Free Form AI, we sit down with Lara Rachidi and Maria Zervou, two AI professionals whose paths defy the traditional tech blueprint.Lara began her career in law and economics before pivoting into data science. Maria started at the British Library, digitizing manuscripts and unlocking forgotten datasets. Today, they’re building GenAI systems, leading strategy across regions, and shaping the next chapter of machine learning at OpenAI and beyond.Tune in to Episode 15 for a wide-ranging conversation about:•What actually makes a technical career resilient•Why network-driven learning accelerates growth in AI•How tools like DSPy are revolutionizing prompt optimization for agentic systems•What the future of LLMs may look like More than a career story, this episode is a reminder: some of the most effective voices in tech began outside of it.If you’re looking to pivot into the AI field and want some role models, this episode’s for you.

In this episode of Free Form AI, we sit down with Puneet Jain, Senior Applied AI Engineer at Databricks, to explore what’s really holding back GenAI from reaching production at scale, and why infrastructure, not intelligence, might be the missing piece.From toy clustering algorithms to national-scale LLMs in healthcare, we unpack how secure data access, agentic tooling, and long-term system design are reshaping the GenAI landscape. We also discuss what “AI agency” really means. More importantly, is the road to AGI shorter than we think?Tune in!

In this episode of Free Form AI, we sit down with Mary Moore-Simmons, VP of Engineering at Kibo (formerly GitHub), to explore how career pivots, leadership, and AI are reshaping engineering.From drying Kona coffee beans with radiant heat to leading complex orgs at GitHub, Mary shares how curiosity, efficiency, and a bias for action can accelerate your career. We discuss the hidden cost of inefficiency in scaling teams, why “just doing the thing” often opens new doors, and how AI is transforming cost and performance optimization at scale.

In this episode of Free Form AI, we explore how AI coding assistants are reshaping the development workflow. From spotting bugs that slip past human review to writing context-aware tests and reviewing entire PRs, these tools are changing how engineers build software. We break down the “secret sauce” behind how they scan a repo with targeted search commands, keeping context focused and results accurate without overwhelming the model. The conversation also tackles the trade-offs, from productivity gains to the risk of skill decay and what this means for the future of engineering.

In this episode of Free Form AI, Michael sits down with Gino Ferrand, CEO of Tecla, to explore how traditional tech hiring misses the mark, especially when it comes to LATAM engineering talent. We break down flawed “rockstar” hiring, AI-proofing your interview process and what successful onboarding actually looks like in a distributed, GenAI-driven world.Music: “Intro” by DJ RADIK, used under CC BY-SA 4.0. Edited for duration and volume by Free Form AI.

In this episode, Michael and Ben discuss the intricacies of building a video clipping service and recommendation engine. They explore the challenges of defining interesting clips, the importance of feature requirements, and the cold start problem when launching without user data. The conversation delves into tagging and metadata extraction, hierarchical models for recommendations, and the significance of user telemetry data. A/B testing is highlighted as a crucial method for optimizing user experience and recommendations.

In this episode, Michael Berk and Ben Wilson discuss the various levels of a data science career, from internships to senior leadership roles. They explore the expectations, responsibilities, and best practices for each level, emphasizing the importance of collaboration, continuous learning, and maintaining a positive team culture. The conversation provides valuable insights for both aspiring data scientists and seasoned professionals looking to navigate their careers effectively.

In this episode, Michael and Ben speak with Mark Fussell, CEO of Diagrid, about Dapr, a runtime for building distributed applications. Mark shares insights on the challenges of microservices development, the importance of durable workflows, and the actor model. The conversation also touches on API design, the future of AI in software engineering, and how Dapr can accelerate development processes. Mark emphasizes the need for best practices in the rapidly evolving AI landscape and how Dapr aims to provide a robust solution for developers.

In this episode, Michael and Ben explore the world of open source projects in AI, focusing on MLflow and its various flavors. They discuss popular machine learning libraries like Scikit-learn, SparkML, and tree-based algorithms such as XGBoost and LightGBM. The conversation also delves into time series forecasting techniques and the rapidly evolving landscape of generative AI, emphasizing the importance of understanding the strengths and weaknesses of different tools in the open source community.

In this episode, Michael and Ben discuss the challenges of decision-making in software engineering, particularly focusing on the sunk cost fallacy, emotional attachment to work, and the importance of prototyping. They explore how to recognize when to pivot or bail on a project, the complexities of ROI calculations, and how to effectively communicate with stakeholders about project scope and trade-offs. The conversation emphasizes the need for a balanced approach to emotional investment in work, the value of iterative development, and the biases that can affect decision-making.

In this episode, Michael Berk and Ben Johnson discuss the essential traits of effective engineering leadership, the importance of customer-centric software development, and the challenges of integrating AI into business processes. They emphasize the need for clear communication, trust, and vulnerability in leadership, as well as the potential risks and ethical considerations surrounding AI technology. The conversation also touches on the evolving nature of human relationships in the age of AI and the importance of fostering a creative and empowered team culture.

In this episode, Michael Berk and Ben discuss the intricacies of transferring code, project ownership, and knowledge within teams, particularly in the context of machine learning and software engineering. They explore the importance of documentation, bridging skill gaps, and ensuring effective knowledge transfer to maintain and extend codebases. The conversation also touches on the challenges of generative AI and the need for continuous upskilling in a rapidly evolving field.

In this episode, Michael Berk and Ben Wilson discuss their experiences at Databricks, focusing on project management, team dynamics, and the importance of product alignment. They explore the challenges of onboarding new team members, the significance of effective communication with stakeholders, and the balance between upskilling and meeting project deadlines. The conversation emphasizes the need for early integration of components, maintaining code quality, and the role of a tech lead in delegating tasks and managing team dynamics.

In this episode, Michael and Ben Wilson discuss the concept of knowledge acquisition, focusing on how to effectively acquire knowledge for practical action rather than for its own sake. They explore the value of note-taking, the role of curiosity in learning, and the importance of hands-on experience in making knowledge intuitive. The conversation highlights the balance between curiosity-driven exploration and just-in-time learning, emphasizing that both approaches can lead to valuable insights and problem-solving skills. In this conversation, Ben and Michael explore the significance of practical experience in learning, the balance between curiosity and just-in-time learning, and the importance of physical fitness in the tech industry. They discuss the nature versus nurture debate in engineering skills and how AI is evolving to reflect societal changes in knowledge acquisition. The dialogue emphasizes that knowledge should be applied for action rather than mere recall, and highlights the value of general knowledge in fostering innovative ideas.

In this episode of Freeform AI, hosts Michael and Ben discuss the importance of understanding ROI in feature development, emphasizing the need for effective prioritization and the implementation of a Minimum Viable Product (MVP). They explore how to analyze demand, identify root causes of inefficiencies, and utilize the MoSCoW method for prioritization. The conversation also touches on the challenges of iterative development, legacy code, and the significance of communication between stakeholders and developers. The hosts share personal experiences and lessons learned from past mistakes, highlighting the value of simplicity in solutions and the importance of continuous improvement.